Transcript PowerPoint
GRS LX 700
Language Acquisition
and
Linguistic Theory
Week 4. Null subjects
(and some more root infinitives)
Null subjects (in English)
Until after around 2 years old, kids will
often omit subjects:
Drop bean.
Fix Mommy shoe.
Helping Mommy.
Want go get it.
Why?
Null subjects
Lots of languages allow
you to drop the subject.
Italian, Spanish: the verb
generally carries enough
inflection to identify the
person, number of the
subject.
Chinese: where the subject is
obvious from context it can
be left out.
Not in English though: Let’s
talk about Bill. *Left. *Bought
groceries. *Dropped eggs.
On the view that kids
know language, but are
just trying to figure out
the specific details
(principles and
parameters), one
possibility is that they
always start out speaking
Italian (or Chinese) until
they get evidence to the
contrary.
Null subjects are
grammatical for kids
Null subjects
Kids do tend to speak in
short sentences. There
seem to in fact be
identifiable stages in
terms of the length of the
kids’ sentences (oneword stage, two-word
stage, multi-word
stage…), often measured
in terms of MLU (mean
length of utterance) which
roughly corresponds to
linguistic development.
Perhaps the kid’s just
trying to say a threeword sentence in a
two-word window, so
something has to go.
That is, some kind of
processing limitation.
Subject vs. object drop
Percentage of Missing subjects and Objects
from Obligatory Contexts
70
60
A
Subjects
Objects
50
E
S
Subject 57
61 43
Object
7
40
30
20
10
0
Adam
Eve
Sarah
8
15
Null subjects
Subjects (in a non-null
subject language like
English) are way more
likely to be dropped than
objects. There’s
something special about
subjects.
Makes a processing
account more difficult to
justify.
Bloom (1990) made some
well-known proposals
about how the null
subject phenomenon
could be seen as a
processing issue, and
tried to explain why
subjects are the most
susceptible to being
dropped.
See also Hyams & Wexler
(1993) for a reply.
Null subjects vs. time
Null subjects seem to be
pretty robustly confined to
a certain portion of
linguistic development.
There’s a pretty sharp
dropoff at around 2.5 or 3.
Hamann’s Danish kids
illustrate this well.
Why can’t English kids really
be speaking Italian?
In Italian, subjects can be
dropped (but need not
be), in English, they can’t
be dropped at all.
So since having subjects
is consistent with Italian,
what’s going to signal to
the kid that they’ve got
the wrong kind of
language?
A “subset” problem.
Possible solution?
Expletive it and there.
In Italian, null subjects
are allowed wherever a
subject pronoun would
be, including embedded
finite clauses (“I know
that [he] has left”) and
finite root questions
(“What has [he]
bought?”).
In Kid English, null
subjects never show up in
these environments. It
doesn’t seem so much
like Italian.
Ok, maybe these kids are
speaking Chinese…
In adult Chinese, subjects
can also be omitted.
In Italian, Spanish, the
allowability of null
subjects was taken to be
tied to the verbal
agreement. Something
about the rich agreement
licenses null subjects.
In Chinese, there is no
agreement morphology,
so that isn’t what’s
allowing null subjects.
Proposal: What allows
argument omission in
Chinese is a form of topic
drop. They are allowed
roughly when they are
“old information”,
recoverable.
Speaking Chinese?
Suppose that these
are parameters.
±Pro-drop for the
Italian/English
difference.
±Topic-drop for the
Chinese/English
difference.
Kid English isn’t +Prodrop.
In +Topic-drop
languages, subjects
aren’t particularly
privileged.
Subjects are often old
information, but when
objects are old
information, they too
can be dropped.
Not speaking Chinese
We’ve already seen
that Kid English
overwhelmingly drops
subjects, not objects.
33% subjects,
4% objects
(Wang et al 1992)
Kid English looks like
English with some
extra null subjects.
But Kid Chinese
drops even more
subjects and lots
more objects.
47% subjects,
23% objects.
Kid Chinese looks like
Chinese with maybe
some extra null
subjects.
Parameters are quick
And recall that Italian
allows null subjects in
embedded clauses,
wh-questions, etc.
Kid Dutch and French
have practically no null
subjects in whquestions.
Kid Italian has
something like 56%
null subjects in whquestions.
If Chinese/Dutch is
distinguished by
[±topic-drop] and
Italian/English is
distinguished by
[±pro-drop], the kids
already know what
they’re trying to speak
by the time we’re
testing them.
Processing accounts…?
Kids have severely limited processing
power, and so they leave off subjects to
ease the load. (Bloom 1990)
In favor:
Length limitations even in imitations
Kids omit things other than subjects
Some kids don’t eliminate subjects, only
reduce their frequency.
Processing accounts…
Contra? Hyams points out:
Build house…Cathy build house
Go nursery…Lucy go nursery
Kathryn want build another house.
Bloom: So, no absolute limit on length,
only a tendency to reduce length.
Bloom (1990)
Bloom (1970) found:
negated sentences tend to lack subjects more
frequently then non-negated sentences.
Bloom (1990):
Hypothesis: sentences without subjects will
have longer VPs than sentences with
subjects.
Looked at past tense verbs and cognitive
states (need) to avoid any confusion with
imperatives.
Bloom (1990)
VP length (words from verb to the end)
counted for sentences with and without
subjects.
Results: Mean length of VP in sentences
with subjects were (statistically)
significantly shorter than those without.
E.g., Adam 2.333 with, 2.604 without.
Bloom (1990)
In fact, “long subjects” (lexical subjects), “short subjects”
(pronouns), and null subjects correlated with an increase
in VP length as well.
3
2.5
2
1.5
Adam
Eve
Sarah
1
No subject
Pronoun
Lexical
Bloom (1990)
And why are subjects dropped more
frequently than objects?
Two possibilities?
Subjects tend to be given (old) information
(low “informativeness”, more expendable)
Maybe processing “saves the heaviest load
for last”
Hyams & Wexler (1993)
Bloom’s (1990) approach (processing)
can’t be right either.
The difference between subjects and
objects is big, and only rate of subject drop
changes.
Adam & Eve both drop around 40-50% of
their subjects in an early stage, and in a later
stage are down to 15-30%—meanwhile their
rate of object drop stays around 5-10%.
Hyams & Wexler (1993)
“Informativeness”?
All else being equal, the ratio of missing
subjects to specific subjects should be equal
to the ratio of missing objects to specific
objects.
Turns out that kids drop specific subjects
about twice as often (Adam 52%) as they
drop specific objects (Adam 21%).
Hyams & Wexler (1993)
Considering Italian adults, we find exactly the same correlation
Bloom reported for English kids: VP seems to be longer where
there is null subject, shorter with a pronoun, and shorter still with
a lexical subject.
3
5
4.75
2.5
4.5
4.25
2
1.5
4
3.75
Adam
Eve
Sarah
Italian adult
3.5
3.25
1
3
No subject
Pronoun
Lexical
No subject
Pronoun
Lexical
Hyams & Wexler (1993)
Regardless of why the correlation holds, if it is a
processing deficiency in kids, what is it for the
Italian adults?
Seems like kids act like they’re speaking a
language where the null subject is a
grammatical option. Note: might be slightly
different from a “null subject language” though.
Point: dropping subjects is grammatical for these
kids, not an error.
Hyams & Wexler (1993)
“Output omission” model predicts ratio of overt
lexical subjects to overt pronouns should
increase over time.
Pronouns are easier, they’ll survive. Lexical subjects
are harder, they’ll be dropped. Initial advantage to
visible pronouns.
Grammatical omission model predicts ratio of
overt lexical subjects to overt pronouns should
decrease over time.
If null subjects are a form of pronoun for kids, they will
“dilute the pool”, putting visible pronouns at an initial
disadvantage.
Hyams & Wexler (1993)
We find: Ratio of overt lexical subjects to overt
pronouns decreases over time…
Adam goes from about 3:1 in favor of lexical subjects
(during subject drop stage) to 1:2 (after subject drop
stage).
When he’s dropping subjects, they are coming
out of the “pronoun” pile—the number of lexical
subjects is staying about the same across
development.
Hyams & Wexler (1993)
Ok, so maybe pronouns are more difficult
than lexical nouns? (Doesn’t fit well with
the length of VP result, but maybe…?)
Problem is: kids show a steady level of
object pronouns throughout this time
period—and output omission model
doesn’t have anything to say about subject
vs. object.
Hyams & Wexler (1993)
Basic conclusion:
Null subjects don’t seem to arise in child
language solely due to processing difficulty.
Rather, they seem to be allowed in the child
grammar.
This allows a distinction between subject (high rate
of omission) and object (low rate of omission)
Explains the tradeoff between null subjects and
pronouns (and the VP length/subject correlation) if
the principles governing availability of subject drop
are similar to those at work in Italian.
So what allows null subjects?
Here’s where we start to tie in to other
properties of that age.
Notice that in English (a non-null subject
language) you can have a grammatical
null subject in one context:
I want [Ø to have a fire drill]
[Ø to have a fire drill] would make my day.
So what allows null subjects?
Subjects of infinitives can be null.
Kids at the age where subjects are often
missing often use infinitive verb forms.
Perhaps that’s the key: Since kids can use
infinitives where adults can’t (main clause
main verb), this allows them to use null
subjects in those sentences as a side
effect.
Proportion of null subjects in
finite and non-finite clauses
null finite
null nonfinite
100
90
80
70
60
50
40
30
20
10
0
Flem
GermS GermA
FrP
FrN
DutchH EngA
Null subjects…
Null subject parameter(s) is/are not initially misset (kids don’t all start off speaking Italian or
Chinese—contra Hyams 1986, 1992); rather,
child null subjects are (at least in part) due to the
availability of non-finite verbs (the OI stage).
Most null subjects are licensed by being the
subject of a nonfinite verb (i.e. PRO)
But there are still some null subjects with finite
verbs… We’ll return to this.
Null subjects and C
Crisma (1992): French kids typically (1/114 =1% vs.
407/1002=41%) do not produce null subjects with a whphrase.
Valian (1991): English kids typically (9/552=2%) do not
produce null subjects with a wh-phrase.
Poeppel & Wexler (1993): German kids typically exclude
null subjects from post-V2 position.
Null subjects and C
It looks like: If the kid shows evidence of CP
(wh-words, V2), then the kid also does not drop
the subject.
Rizzi’s idea:
A discourse-licensed null subject is available only in
the highest specifier in the tree (topic-drop).
Axiom: CP=root
Kids don’t “get” the axiom until between 2-3 years old.
Truncated trees
The result (of not having CP=root) is that
kids are allowed to have truncated
structures—trees that look like adult trees
with the tops chopped off.
Importantly: The kids don’t just leave stuff
out—they just stop the tree “early.” So, if
the kid leaves out a functional projection,
s/he leaves out all higher XPs as well.
Truncation
If kid selects anything lower than TP as the
root, the result is a root infinitive—which
can be as big as any kind of XP below TP
in the structure.
Note in particular, though, it can’t be a CP.
So: we expect that evidence of CP will
correlate with finite verbs.
Truncation
Pierce (1989) looking at French observed
that there are almost no root infinitives
with subject clitics—this is predicted if
these clitics are instances of subject
agreement in AgrS; if there is no TP, there
can be no AgrSP.
Truncation
There is some dispute in the syntax
literature as to whether the position of
NegP (the projection responsible for the
negative morpheme) is higher or lower
than TP in the tree.
If NegP is higher than TP, we would expect
not to find negative root infinitives.
Truncation and NegP
But we do find negative Root Infinitives—
(Pierce 1989): in the acquisition of French,
negation follows finite verbs and preceds
nonfinite verbs (that is—French kids know
the movement properties of finiteness, and
thus they have the concept of finiteness).
Truncation and NegP
So, is TP higher than NegP?
Hard to say conclusively from the existing
French data because there are not many
negative root infinitives—but further study
of child language could lead to a
theoretical result of this sort about the
adult languages, if we assume a truncation
analysis of child language.
S O Vfin = AgrSP?
Usually (Poeppel & Wexler 1993) German
kids put finite verbs in second position, and
leave nonfinite verbs at the end.
Occasionally one finds a finite verb at the
end.
Rizzi suggests we could look at this as an
instance of a kid choosing AgrSP as root,
where CP is necessary to trigger V2.
P&W had to basically consider these “noise”.
Truncation and null subjects
As for null subjects:
If the tree is just a VP, the subject can be
omitted in its base position—it’s still in the
specifier of the root.
If the tree is just a TP, the subject can be
omitted from the normal subject position—
note that this would be a finite verb with a
null subject.
If the tree is a CP and SpecCP is filled (like
in a wh-question) we expect no null
subjects.
Null subject languages vs.
root infinitives
Italian seems to show no (or very very
few) root infinitives. If this is maturation of
“Root=CP” how could languages vary?
Rizzi suggests:
In English, V doesn’t move
In French, tensed verbs move to AgrS (I),
untensed verbs may move to AgrS
In Italian, all verbs move to AgrS
Null subject languages vs.
root infinitives
The idea is that a verb in Italian needs to
get to AgrS—it has a feature/property
(parametric) that marks it as needing to
get to AgrS in a grammatical sentence.
Hence, the kid needs AgrS.
English verbs have no such need, so the
English kids have to rely on Root=CP to
tell them to keep going.
Null subject languages vs.
root infinitives
Rizzi and Wexler capture NS/OI similarly:
Wexler: AgrS does not “need” a subject in its
specifier in Italian, so there is no competition
between AgrS and T, and thus no need for
root infinitives. AgrS and T are always both
there.
Rizzi: AgrS can never be omitted in Italian,
because the verb needs AgrS to be there.
Having AgrS implies T. AgrS and T are always
both there.
Back to null subjects vs. ±Fin
Bromberg & Wexler (1995) promote the idea that
null subjects with finite verbs arise from a kind of
“topic drop” (available to adults in special
contexts).
Proposal (Bromberg & Wexler)
Topic-drop applies to Very Strong Topics
Kids sometimes take (in reality) non-VS topics to
be VS topics (a pragmatic error)
Prediction about NS
RI’s have two ways of licensing NSs:
PRO (regular licensing of null subject)
Topic drop
Finite verbs have one way to license a NS:
Topic drop
So: We expect more null subjects with root
infinitives (which we in fact see).
Cf. Rizzi: Subject in highest specifier can always
be dropped, and RI’s also allow PRO. Same story,
basically.
Bromberg, Wexler, whquestions, and null subjects
If topic drop is something which drops a topic
in SpecCP…
…and if wh-words also move to SpecCP…
…we would not expect null subjects with
non-subject (e.g., where) wh-questions
where the verb is finite (so PRO is not
licensed).
Cf. Rizzi: Same prediction; if you have a CP, a
subject in SpecTP won’t be in the highest
specifier, so it can’t be dropped. One difference:
Rizzi predicts no nonfinite wh-questions at all,
hence no null subjects at all.
Bromberg, Wexler, whquestions, and null subjects
Finiteness of null/pronominal subjects, Adam’s whquestions (Bromberg & Wexler 1995)
Finite
Nonfinite
Null
2
118
Pronoun
117
131
*Truncation
Rizzi’s “truncation” theory predicts:
No wh-questions with root infinitives
wh-question CP, but
CP IP, and
IP finite verb
And of course we wouldn’t expect null
subjects in wh-questions if null subjects
are allowed (only) in the specifier of the
root.
*Truncation?
Guasti points out that although Bromberg
& Wexler did find null subjects in whquestions in English, English is weird in
this respect.
Arguably, null subjects are precluded from
wh-questions in most other languages.
V2 and wh-null subjects…
German and Dutch have extremely few root
infinitives when there is anything in SpecCP.
This does go with Rizzi’s prediction…
But they are V2 languages—finite verbs are
what you find in C, and when SpecCP is filled,
there must be something in C. Hence, Wexler’s
prediction seems to be:
V2 language no wh-question root infinitives
And this seems closer to accurate, given English.
V2 and wh-null subjects…
And yet, Crisma’s (1992) findings and Hamann
& Plunkett’s (1998) findings suggest that French
(not V2) also shows almost no null subjects in
wh-questions.
So what’s different about English?
French, Dutch, German basically never have null
subjects in wh-questions.
English allows them readily.
Adult null subjects
(“diary drop”)
Both Rizzi and Bromberg & Wexler appeal to
properties of adult language to justify the child
null subjects.
B&W suggest that topic drop is available in English,
but only for Very Strong topics, and what kids are
doing wrong is identifying far too many things as VS
topics.
Rizzi suggests that the ability to drop a subject in the
highest specifier is available in certain registers
(“diary drop”) (where presumably Root=CP is
disregarded, or at least relaxed to allow Root=IP).
Saw John today. Looked tired.
Hamann & Plunkett (1998)
Finite null subjects. Hamann discussed this
question: If null subjects are licensed by RIs,
what should we say about the null subjects with
finite verbs? W had previously said “topic drop”,
but H showed that Danish kids’ use of null
subjects with finite verbs correlated highly with
the use of RIs in general.
That’s a problem because “topic drop” according to
B&W is due to kids mistaking what can be a VS topic,
and should be independent of Tense/Agr. For
truncation, though, the same basic mechanism is at
work creating both finite null subjects and RIs.
Root infinitives vs. time
The timing on root
infinitives is pretty
robust, ending around
3 years old.
Wexler (2000)
Are there really lots of null subjects with finite verbs
in Danish?
Idea: køb-er looks like present tense finite, but it
could be missing T (hence legitimately license NS).
[+Agr, +Tns] køb-er (present) (adult)
[-Agr, +Tns] køb-e (infinitive) no NS allowed
[-Agr, -Tns] køb-e (infinitive) NS allowed
[+Agr, -Tns] køb-er (“present”) NS allowed.
Predicts: No NS’s with past tense verbs like køb-de
(since unambiguously +Tns, which is the thing that
prevents NS). True?
Hamann (2002) vs. Wexler
Well, not really vanishingly small…
Jens (20-34 mos.s) 14/42 (33%) NS past.
Anne (18-30 mos.) 13/33 (39%) NS past.
Hamann herself prefers a truncation story to
account for these; finite NS corresponds to
truncating at TP.
Yet, don’t forget about Swahili, and the
apparently visible effects of ATOM.
Interpretation and functional
categories
A basic premise of Hoekstra & Hyams (1998) is
that tense is a means of connecting between the
structural meaning and the discourse. Tense
anchors a sentence in the discourse.
They propose that the relation between
discourse (CP) and T must be signaled (to
ground an utterance), and is signaled by
different things in different languages.
Dutch: number morphology only these have RIs?
Japanese: tense morphology
Italian, Spanish, Catalan: person morphology
Underspecification of
number?
H&H propose in light of this that what’s
wrong with kids has to do with number
specifically. OI languages are those where
number is crucial in the finite inflection.
H&H picked up on something about when
these RIs seem to be used. It seemed that
there are certain verbs that showed up in
the nonfinite form, but others that didn’t.
Eventivity Constraint
In particular, it seems that RIs show up only with verbs
referring to events —not with verbs referring to states,
not with auxiliary verbs. Finite verbs seem to have no
such restriction. Original research on Dutch on French,
also Russian.
Eventivity Constraint
RIs are restricted to event-denoting predicates.
Modal Reference Effect
The other thing is that RIs often have a “modal”
meaning (can, will, must, want to..) (pretty
dramatic in Dutch, German, French).
Poeppel & Wexler (1993) did give a German example
from Andreas that showed an RI with seemingly no
modal meaning (Thorsten Ball haben); if H&H are
right, this was “noise”.
Modal Reference Effect
With overwhelming frequency, RIs have modal
interpretations.
English = weird
English doesn’t seem to conform to the
pattern. Ud Deen (1997) found:
plenty of bare stative verbs (*EC)
Man have it
Ann need Mommy napkin
Papa want apple
plenty of non-modal bare verbs (*MRE)
Dutch: 86% of RIs have modal meaning. Cf. 3%of
finite forms.
English: 13% of bare forms have modal meaning
Cf, 12% of finite forms..
H&H’s hypothesis
Number is an inflectional property both of
the nominal and the verbal system.
though it arises in the nominal system.
Missing determiners and RIs are both a
symptom of “underspecified” Number.
Spec-head agreement
communicates number
(under)specification
to the verb.
H&H (1998) BUCLD
Looked at Niek (CHILDES, Dutch).
They found that with “finite DPs”, the verb
was pretty much always finite too.
They found that with “nonfinite DPs”, the
verb was somewhat more likely to be
nonfinite than with a finite DP, but still
overwhelmingly favored finite DPs.
Only null subjects didn’t overwhelmingly
favor finite V. (NS 45% nonfinite).
H&H (1998) BUCLD
All things being equal, we might have
expected a 1:1 correlation between finite
DP subjects and finite V, if it were a matter
of Spec-head agreement. We don’t have
that. We have a one-directional relation.
If DP is finite, V is finite.
If V is nonfinite, DP is nonfinite.
H&H (1998) BUCLD
In a sense, one setting “cares” about its partner in
the Spec-Head relationship, and the other setting
doesn’t.
Finite V seems not to care whether the subject is finite or
not.
Nonfinite V does seem to care, and requires a nonfinite
subject.
More specifically, there is a “default”, and the
“default” does not need to be licensed (and nondefaults do).
This goes along with an assumption that either the syntax
doesn’t make person distinctions if the morphology
doesn’t, or that this part of “checking” is really about
morphology.
H&H (1998) BUCLD
In Dutch, 3sg is default.
1sg verb licensed only by a 1sg subject.
3sg verb licensed by any old subject.
In English, 3sg is not the default. It’s the
one marked form.
3sg verb licensed only by a 3sg subject.
bare verb licensed by any old subject.
Thus
The doggie bark.
He bark
Doggie sit here.
*Doggie barks.
*Het hondje hier zitten.
*He hier zitten.
Hondje hier zitten.
Hondje zit hier.
cf. Schütze & Wexler
“…the English bare form is ambiguous
between an infinitive … and a finite form…”
(H&H98:101)
Although stated in different terminology,
and addressing a slightly different arena of
facts, the basic concept is the same as that
in S&W96.
[+T+A] -> finite (-s)
[+T-A], [-T+A], [-T-A] -> “nonfinite” (stem)
but +A ones will have +A properties (e.g. NOM),
even if just a stem form. Same for +T.
English bare form ≠ infinitive
S&W and H&H agree that the English bare
form isn’t strictly speaking (necessarily)
the true infinitive.
H&H and interpretation
Claim: RIs are interpreted as [-realized], the
contribution of the infinitival morpheme itself.
Languages with an infinitival morpheme and
actual RIs should show 100% modal ([-realized])
interpretation with RIs.
English, with a Ø infinitival morpheme, obscures
the correlation; in practice, we expect only some
(the actually infinitive) bare forms to be modal.
epistemic vs. deontic
John must leave.
Deontic: About the way the world isn’t now but needs to
be.
John must know French.
Epistemic: About our beliefs about the world.
Seems to be a correlation between “eventivity” and
modality type, in the adult language.
Modality and kids
In other circles of research, people have
proposed that kids basically “don’t have”
epistemic uses of modality (John must be
a genius) before about 3 years old—for
whatever reason.
If that’s true, there’s only deontic modality
(John must go to class).
If deontic modality only goes with eventive
predicates, we’re done. Kids RIs are
modal, necessarily deontic, hence
necessarily with eventive verbs.
English must be different
English bare forms are not (necessarily)
infinitives, not necessarily modal, hence
not necessarily deontic, eventive.
Hence, the EC and MRE appear not to
hold of English, but for reasons we can
now understand.
A pause to regroup
English bare form is unmarked, only -s is unambiguously
+T+A.
Do is a reflex of +T (and/or +A), and as expected, almost
never in negative sentences was there a post-negation
inflected verb (she doesn’t go vs. *she not goes).
The actual infinitive morpheme in English is Ø, so we
can’t differentiate bare forms between infinitives and
other bare forms.
The infinitive morpheme seems to carry modal
meaning—in languages where you can see it you can
tell. Effectively RI only with eventives.
A pause to regroup
H&H propose that the languages which show
OIs are those which rely (only) on number in
their inflectional system. Those that don’t
(Japanese [tense only], Italian [person]) seem to
be immune. Hence, person is the special,
possibly omitted thing for kids.
This isn’t really distinctly at odds with ATOM.
Wexler suggests that the problem is with doublemovement of the subject, but movement of the
subject might itself be driven by person features
in recent versions of the syntactic thy.
A pause to regroup
H&H observed a correlation between specified
(“finite”) subjects and verbal form.
Specifically,”finite” subjects seem to “cause”
finite verbs. Not obvious why this would be
under ATOM directly, but it might be something
like what H&H suggest—there is feature sharing
between the subject and the AgrP. It might be
interesting to see if “finite” subjects necessarily
always show the reflex of AgrP and not
necessarily of TP.
Legendre et al. (2000)
Wexler: During OI stage, kids sometimes omit T,
and sometimes omit Agr. Based on a choice of
which to violate, the requirement to have T, to have
Agr, to have only one.
(cf. “Kids in a pickle” slide)
Legendre et al.: Looking at development (of
French), it appears that the choice of what to omit
is systematic; we propose a system to account for
(predict) the proportion of the time kids omit T, Agr,
both, neither, in progressive stages of development.
Optimality Theory
Legendre et al. (2000) is set in the
Optimality Theory framework (often seen
in phonology, less often seen applied to
syntax).
“Grammar is a system of ranked and
violable constraints”
Optimality Theory
In our analysis, one constraint is Parse-T,
which says that tense must be realized in
a clause. A structure without tense (where
TP has been omitted, say) will violate this
constraint.
Another constraint is *F (“Don’t have a
functional category”). A structure with TP
will violate this constraint.
Optimality Theory
Parse-T and *F are in conflict—it is
impossible to satisfy both at the same
time.
When constraints conflict, the choice
made (on a language-particular basis) of
which constraint is considered to be “more
important” (more highly ranked)
determines which constraint is satisfied
and which must be violated.
Optimality Theory
So if *F >> Parse-T, TP will be omitted.
and if Parse-T >> *F, TP will be included.
Optimality Theory
Grammar involves constraints on the
representations (e.g., SS, LF, PF, or
perhaps a combined representation).
The constraints exist in all languages.
Where languages differ is in how important
each constraint is with respect to each
other constraint.
Optimality Theory: big picture
Universal Grammar is the constraints
that languages must obey.
Languages differ only in how those
constraints are ranked relative to one
another. (So, “parameter” = “ranking”)
The kid’s job is to re-rank constraints
until they match the order which
generated the input that s/he hears.
Legendre et al. (2000)
Proposes a system to predict the
proportions of the time kids choose the
different options among:
Omit TP
Omit AgrSP
Omit both TP and AgrSP
Include both TP and AgrSP (violating UCC)
French v. English
English: T+Agr is pronounced like
/s/ if we have features [3, sg, present]
/ed/ if we have the feature [past]
/Ø/ otherwise
French: T+Agr is pronounced like:
danser
a dansé
je danse
j’ai dansé
NRF
(3sg) past
1sg (present)
1sg past
The idea
Kids are subject to conflicting constraints:
Parse-T
Parse-Agr
*F
*F2
Include a projection for tense
Include a project for agreement
Don’t complicate your tree with
functional projections
Don’t complicate your tree so
much as to have two functional
projections.
The idea
Sometimes Parse-T beats out *F, and then
there’s a TP. Or Parse-Agr beats out *F,
and then there’s an AgrP. Or both Parse-T
and Parse-Agr beat out *F2, and so there’s
both a TP and an AgrP.
But what does sometimes mean?
Floating constraints
The innovation in Legendre et al. (2000)
that gets us off the ground is the idea that
as kids re-rank constraints, the position of
the constraint in the hierarchy can get
somewhat fuzzy, such that two positions
can overlap.
*F
Parse-T
Floating constraints
*F
Parse-T
When the kid evaluates a form in the
constraint system, the position of ParseT is fixed somewhere in the range—and
winds up sometimes outranking, and
sometimes outranked by, *F.
Floating constraints
*F
Parse-T
(Under certain assumptions) this
predicts that we would see TP in the
structure 50% of the time, and see
structures without TP the other 50% of
the time.
French kid data
Looked at 3 French kids from CHILDES
Broke development into stages based on a modified
MLU-type measure based on how long most of their
utterances were (2 words, more than 2 words) and
how many of the utterances contain verbs.
Looked at tense and agreement in each of the three
stages represented in the data.
French kid data
Kids start out using 3sg agreement and
present tense for practically everything
(correct or not).
We took this to be a “default”
(No agreement? Pronounce it as 3sg. No
tense? pronounce it as present. Neither?
Pronounce it as an infinitive.).
French kid data
This means if a kid uses 3sg or present
tense, we can’t tell if they are really using
3sg (they might be) or if they are not using
agreement at all and just pronouncing the
default.
So, we looked at non-present tense forms
and non-3sg forms only to avoid the
question of the defaults.
French kids data
We found that tense and agreement
develop differently—specifically, in the first
stage we looked at, kids were using tense
fine, but then in the next stage, they got
worse as the agreement improved.
Middle stage: looks like
competition between T
and Agr for a single node.
A detail about counting
We counted non-3sg and non-present verbs.
In order to see how close kids’ utterances were to adult’s
utterances, we need to know how often adults use non3sg and non-present, and then see how close the kids
are to matching that level.
So, adults use non-present tense around 31% of the
time—so when a kid uses 31% non-present tense, we
take that to be “100% success”
In the last stage we looked at, kids were basically right at
the “100% success” level for both tense and agreement.
Proportion of non-present
and non-3sg verbs
40%
35%
30%
25%
non-present
non-3sg
adult non-pres
adult non-3sg
20%
15%
10%
5%
0%
3b
4b
4c
Proportion of non-finite root
forms
35%
30%
25%
20%
NRFs
15%
10%
5%
0%
3b
4b
4c
A model to predict the
percentages
Stage 3b (first stage)
no agreement
about 1/3 NRFs, 2/3 tensed forms
*F2
ParseT
ParseA
*F
A model to predict the
percentages
Stage 4b (second stage)
non-3sg agreement and non-present tense
each about 15% (=about 40% agreeing,
50% tensed)
about 20% NRFs
*F2
ParseT
ParseA
*F
A model to predict the
percentages
Stage 4c (third stage)
everything appears to have tense and
agreement (adult-like levels)
*F2
ParseT
ParseA
*F
Predicted vs. observed—
tense
40%
35%
30%
25%
non-present
predicted non-pres
20%
15%
10%
5%
0%
3b
4b
4c
Predicted vs. observed—agr’t
40%
35%
30%
25%
non-3sg
predicted non-3sg
20%
15%
10%
5%
0%
3b
4b
4c
Predicted vs. observed—
NRFs
35%
30%
25%
20%
NRFs
predicted NRFs
15%
10%
5%
0%
3b
4b
4c