Verbs and Nouns: INFL and the Emergence of DP*
The Acquisition of Agreement Features
Joseph
Galasso
2001
(1999)
California
State University, Northridge
1.
Functional Parameterization
The study of the acquisition of IP and of the Determiner Phrase
(=DP) can help determine whether or not Functional Parameterization
has taken place in the child's syntax. Under the current Minimalist
Program, formal syntax provides a mechanism for Checking
morphological features within specific functional (local) domains
triggering movement operations either at post-Spell-out
(covert) LF, or pre-Spell-out (overt) PF-levels of representation.
For instance, if we assume that abstract Nom(inative) Case assignment
is checked under a Spec-Head Agr(eement) relation within IP, Gen(itive)
Case is checked via a Spec-Head relation within DP, [i]
and Acc(usative) either under a Verb-internal (Head-Comp) relation
or via Default, then, a central prediction might be made concerning
any possible absence of the functional categories IP and DP in
early child clause structure: viz., only instances of Accusative
Case assignment (via default) should be notable at prefunctional
stages of language development. The following sections examine
the acquisition of INFL(ection) along with the role that the DP
system plays in the early development of English. In §1.1, I begin
by examining the idea that a correlation exists between D and
I. §1.2 presents the relevant Data as follows: DPs (§1.2.1), Case
(§1.2.2) and Tense (§1.2.3) (Preview) Regarding
the early emergence of DPs found in the VP-stage (=DP>VP),
we come to the conclusion (following Radford: 1990) that they
represent for the child a miscategorized lexical category:
i.e., they function in the same manner as their more primitive
NP-counterparts. For instance, owing to this miscategorization,
Case features that are typically associated with Poss(essive)
DPs are postponed until the fully-fledged, well formed DP (=DP>IP)
has been acquired--here, being triggered by the productive usage
of the Possessive element 's. Regarding INFL, we conclude
that a Prefunctional-stage exists (=Stage-1)--viz., a stage
where all formal feature specifications are totally absent--before
the onset of an (optional) Underspecified-Functional-stage
(=Stage-2).
The material in this paper roughly
corresponds to my Chapter 4 INFL (see Galasso Ph.D. Essex
1999). For all cited references, see Galasso The Acquisition
of Functional Categories (2003 IULC Press).
1.1.1 Correlation Between D and I
Recent arguments have been put forward suggesting there to
be a Feature Correlation between the Head features of D(P)s
and the Spec features of INFL (cf. Felix: 1990, Hoekstra et.
al: 1996, 1996a,b,). The basic premise behind the correlations
results in the following conditions in (1) (Hoekstra et al:
1996). (Nb. An alternative and less constrained version of the
condition has emerged stating that non-finite clauses may in
fact opt for either specified or non-specified subjects (Hyams:
1997)).
(1)
a. When a D(P) subject is underspecified--for
Definiteness/Agreement -- then I will also be underspecified
for T(ense)/Agr(eement).
b. When a D(P) subject is specified--for Definiteness/Agreement--then
I will likewise be specified for T(ense)/Agr(eement).
The reasoning behind the posited correlation has to do with
conventional notions that claim that Subject Pronouns/(DPs),
when within a Spec-Head agreement configuration within IP, are
to be assigned Nominative Abstract Case. This appropriate checking
domain then allows for the checking of formal Agr(eement) phi-features
of D (e.g. phi (f)=
person/number/gender) to occur between D and I. The conditions
in (1) would therefore account for the fact that e.g., the 3sg
copula Is requires a 3sg Subject She. The correlation
holds owing to the stipulation that all uninterpretable features
(in this case, being the 3sg feature) must be checked-off. However,
the correlation does not hold indiscriminately--that is, the
correlation only holds where it would otherwise result in an
uninterpretable feature remaining unchecked (thus resulting
in a crashed derivation). For example, consider a D(P) 3sg Subject
which lacks a Gender feature--with the 3sg number property along
with all other phi-features remaining present. In this case,
there is nothing in the correlation that would predict IP to
be underspecified for 3sg number Agreement as a result of the
DP's lack of a gender property--the remaining uninterpretable
features may proceed to be checked and erased (only gender fails
to manifest itself). Furthermore, one might wish to extend the
above observation regarding an indiscriminate correlation, and
predict that a D(P), in principle, could appear within a VP
(a non checking domain) projection. In such cases, intrinsic
feature(s) of D(P)--presumably those more semantically oriented
(e.g., Definiteness)--would be permitted to go unchecked possibly
due to the default setting of D itself. In extending this potential
default DP to Subject position (say of underspecified IP), we
forgo all meaningful operations to the correlation. It is precisely
this observation which begs the question of a correlation in
general.
The above correlations predict e.g., that Subjects of Finite
clauses should be overtly marked for Agreement/Definiteness:
i.e., Finite verbs--I specified for T/Agr--should trigger Plural
Nominals (e.g., if the verb is plural) or Nominals with overt
D as in (2a,b) below (Radford:1997ms (simplified by ignoring
Spec-VP)):
(2)
(Nb. Henceforth, conflated trees
should be read as follows: the arrowed-letter (e.g. a>
) indicates the exact feature involved, while the primed-letter
counterpart (e.g. a'.) represents the token example expressing
that feature)
In contrast to (2), Subjects of non-finite clauses should have
bare nominals without their counterpart Determiner and Number
specifications as in (3 a,b) below:
(3)
The absence of the determiner with a singular count noun (cf.
3a) might be analyzed in two different manners: (i) it
may either indicate the failure to mark number; or rather,
(ii) it might indicate the failure to mark definiteness.
One interesting thing to note here, and something I wish
to emphasize, is the notion that a DP containing an overt Det
(specified for definiteness as in *3c) might be potentially
analyzed as the Specifier (in-situ) of a Lexical VP (a
somewhat demoted analysis given that the child's acquisition
of DP is commonly viewed as marking the Functional-Stage of
language development). In this sense, theoretical correlations
which place the acquisition of D(P) Subjects with that of IP,
need not necessarily apply (compromising 1b above). Radford
(op.cit) invokes such a principle by suggesting that a Determiner,
when seemingly associated with a vacuous/non-specified IP (e.g.
-T/-Agr), might simply take-on a default Objective Case
status.[ii]
In addition, if checking were involved, the DP could possibly
be assigned Objective/Accusative Case via a Structural relation
with the Verb (VP-internally). The possibility of a Default
Determiner, however, complicates issues surrounding the classification
of DP: typically speaking, DPs are considered to be a functional
category (cf. Fukui 1986; Abney 1987)--it is in this vein that
a correlation with IP is naturally intuitive. However, the above
default analysis of Subject DPs rather undercuts the issue of
classification and leaves open the question of whether or not
a (straightforward) correlation necessarily holds between D
and I. In other words, all aspects of the correlation may hinge
entirely upon whether or not D(P) is in Spec of IP (=DP>IP)
or Spec of VP (=DP>VP)--so complicating matters.
On Empirical grounds, the relatively early emergence
of D (as seen in my data and in various data in the literature)
as opposed to the protracted emergence of the full-fledged IP,
can be accounted for in quite independent manners--an observation
strongly favoring the opinion that no necessary-correlation
exists.
On Theoretical grounds, there traditionally exists a handful
of differentiating characteristics found amongst Functional and
Lexical elements (see Abney 1987: 64f). Abney, among
others, has suggested that there might be some reason to speculate
on a dual status for DP's function: i.e., in addition to maintaining
their traditional Functional-categorial role--a role which may
indeed correlate D to I, DPs might also play a semantic Lexical-categorial
role--a role which typically pertains to categories lower down
from IP in the structural tree. A number of arguments might go
as follows:
(i) Generally speaking, since Ds are typically associated
with Nouns or NPs (i.e., they tend to form maximal projections
of substantive elements), their referential properties might
likewise be substantive/semantically motivated.
(ii) When an overt DP appears within Functional projections,
(e.g., DP>IP), they take-on those more formal aspects (i.e.,
features of agreement) typically associated with Functional
categories.
(iii) However, when an overt DP appears solely within Lexical
projections, (e.g., DP>VP), they might simply take on those
relative substantive properties having to do with the Lexical-Thematic
Verb and VP.[iii]
Regarding this possible distributional asymmetry of D, Chomsky's
claim here would suggest that under the more formal D-feature
account (DP>IP), substantive references of Ds must continue
to be supplied nevertheless, in some other manner in the semantics
(presummably at (or even above) LF). This amounts to saying
that after any checking-off and deletion of formal D-features,
the reference of D itself must remain visible at LF for
reasons having to do with its substantive/semantic nature
(Chomsky 1995: 279). This leaves opening the question of whether
or not [+Def] may nonetheless be active in a (DP>VP). (See
below for an expanded treatment of this regarding [-/+Interpretable]
D-features.)
The above arguments are tantamount to readdressing outstanding
issues regarding the analysis of D. Firstly, if we assume the
DP-analysis (cf. Abney), it remains unclear whether or not a
determiner (e.g., The) should be analyzed as the Head
(D) or Specifier of a DP. Secondly, an NP analysis for D still
remains an option. An example of a similar dilemma is illustrated
in Radford (1990: 68ff) who claims that early possessors--like
determiners--are in Spec-NP (e.g., Mommy car, Dolly
hat, etc.) and not in Spec-DP. This analysis gives him a
readily available account for the lack of Case (genitive 's)
for such examples--i.e., the Case Filter was seen as being inoperative
due to the lack of the case-marking functional category
D.
Attempts to redefine the nature of DP via its maximal projection
and not by its inherent properties have been recently reported
in Language Acquisition literature. In fact, two independent
bodies of investigation--out of a small handful cited in the
previous chapters regarding literature review--claim such an
interpretation for DP on both empirical and theoretical grounds.
Theoretically speaking, Meisel (1994, 1995), while writing on
Language Mixing, states that only those functional categories
that universally host verbal elements (viz., IP and CP), as
opposed to functional categories that host strictly nominal
ones (viz., DP) obey Functional Constraints on Mixing.
Meisel goes on to give empirical evidence that DPs (within VP
environments) have nothing whatsoever to do with language specific
awareness. In other words, language mixing/code-switching that
incorporates such DPs violated all known functional constraints
on mixing. (See Chapter 6 on Code-Switching for a full discussion).
Hence, DPs might not function in such a strict manner typically
associated with Functional Categories. Paradis et al. (1996)
likewise have claimed recently that the Definite DP system emerges
relatively early in their data, coupled with the complete lack
of any other type of functional projection. They give ample
evidence to suggest that DPs and IPs are acquired independently
of each other--e.g., a Det could be omitted in a finite utterance
and a nonfinite utterance could contain a Det (op.cit: 25).
In sum, Abney's (1986, 1987 op.cit.) important observation
stating that functional categories generally tend to have affixal
natures--i.e., tend to be bound morphemes which are attached
to other categories (mostly lexical)--might be reexamined in
the light of recent reports of the distribution of DP. In addition,
Chomsky (1995: 349) makes clear the notion that among the functional
categories (T, C, D, and Agr), it is only Agr which can bost
of being free of interpretable features: T, C, and D having
interpretable features which provide instructions at interface
levels. In this broadest sense, Agreement--encompassing all
notions of Case--is the formal category par excellence.
Specifically speaking, English DPs (CPs as well) tend not to
fall systematically into the affixal distinction (owing
to their quasi-substantive make-up)[iv]
and, as a result, might well be classified as having a dual
status: (i) a lexical-category status (i.e., having
an objective/default value) when projected from a Spec of VP
(=DP>VP); (ii) a functional-category status
when projected from a Spec of IP (=DP>IP). More concretely,
the above observation regarding the grammatical properties of
DPs--i.e., the grammatical features which play a role in the
syntax of a Spec-Head Agreement relation between the Subject
and the Verb--might be further expanded into notions of an asymmetry
found between phi-features and Case (Chomsky 1995: 278).
Suppose Case differs from phi-features in that Case is always
[-Interpretable]/ [+Formal]--hence, in need of checking.
Suppose (grammatical) phi-features are indeed [+Formal], though
with an added stipulation that they may also be [+/-Interpretable].
The most likely candidate for a possible (D) [+Interpretable]
phi-feature would seem to be Definiteness: this is based
on its more semantic-pragmatic referential properties. In principle,
this may leave the remaining (D) phi-features (e.g., gender/--and
to a lesser degree--person/number) as possibly deriving the
features: [+Formal/-Interpretable]--hence, in need of checking.
(Nb. caveat: At the moment there seems to be no straightforward
consensus on which phi-features constitute as an interpretable
feature--just as there equally seems to be no consensus on which
formal features have semantic properties. Nonetheless, we shall
consider here Definiteness as foremost in pertaining
to such intrinsic semantic properties. This judgment is based
on data which tend to show that children universally acquire
[+]Definitness--the "concrete-volition"
here-and-now--way ahead of [-]Definiteness--the "abstract" there-and-then.
Furthermore, the latter phi-features, unlike +Def, may very
well consist of feature-properties that have to be checked on
the corresponding functional Verb.) By further pursuing this
notion that definiteness is [+Interpretable], we can begin to
reconcile our ideas that some Determiners--those carrying only
the [+Interpretable] definiteness feature and no other phi-feature--may
be reduced to having an objective and/or default status:
the idea being that such Ds might maintain some sort of inherent
case given by a thematic V is also viable.
1.2 The Data
1.2.1 DPs
Empirical support for the above analyses of Definite Subject/Object
DPs is widely borne out in my own data.[v]
The Determiner system for Definiteness is reported to
emerge in the very earliest Files (starting with file 2: 1;10).
The fact that they emerge on the scene way ahead of any unambiguous
Finite INFL(ection) suggests there to be no correlation between
the emergence/acquisition of Subject D(P)s and IP (See Data:
Stage-2 below for arguments against any possible correlation
between Object DPs and INFL). The findings suggest that children
come to realize that the case-feature of D can be optionally
applied in syntax--when it lacks Case, it is spelled-out as
a Default Case form (in either Subject or Object position)--with
no other phi-feature being specified. Consider some of the earliest
token Ds found in my corpus (while keeping in mind that, at
the same stage, Ds are frequently optionally omitted):
Regarding the feature specification of The in (4), it
seems to be the case that only the definiteness features has
been acquired--for instance, number and case don't
appear in the early Files (2-7) as stated in the in-note
above. (The first marking of plural "s" doesn't
emerge until well into the later Files).
Firstly, consider the feature Case. The most natural
way to determine if Case had been properly assigned to The
(Spec of DP) would be to examine if the Head V(erb) is correctly
spelled-out for its Spec features. For example, consider the
following sentence: The boys are reading the books. The
Spec-features of the Head V(erb) are requires a Nominative
specifier for its subject: e.g., ARE [Spec=Nom]. In this
sense, it is clear that the DP The boys must carry Nominative
case--if it were to carry Objective case, the derivation would
crash: e.g., *Them are reading the books. Hence, it remains
a feature of the Head (V) to determine if the case requirements
of a Spec (D) are being met. The token examples in (4) cannot
maintain whether or not case is specified in the above sense:
all forms of verbs taken here are non-specified in all the crucial
areas (i.e. the Spec features of the Head Verbs in question
do not contain the relevant feature specification.
Secondly, consider the features number and person.
They too rely on the Head V(erb) to determine the features--again,
examples in (4) don't suffice. Taking the same sentence, the
DP The boys must also carry [3Pnom] since the verb are
indicates P(lurality): e.g., *The boys is... The 3person/plurality
can be easily demonstrated via the binding of an anaphoric reflexive:
The boys read themselves /*ourselves/yourselves to sleep.
Although the above analyses seem innocent enough, they don't
naturally follow from out of the intrinsic make-up of DP
itself: DPs can variably consist of either Nom or Obj case
(cf. The boys read vs. I read to the boys)
singular or plural (cf. The boys vs The boy).
However, interestingly, regarding 3person, there seems to be
no other variable option for DPs (cf.*The boy am/are)
DPs are intrinsically 3person. In consideration of the lack
of number and case features, The would seem to be underspecified
both for phi-features and case--with the sole exception of the
feature [+Definiteness]. Such a sweeping underspecification
of phi-features and case brings us to the question of how to
analyze such structures. Suppose we are correct in stating that
only the [+Def] feature appears on the early Ds in question.
We then could run the two possible stories alluded to earlier,
suggesting the following (taken from (1) §1.1, restated here
as 1-prime):
(1')
(i) If The is underspecified for case
and number, but specified for Def-- then DP is in Spec-VP,
hence (DP>VP). (ii) If The is fully specified
for case, number and Def, etc.-- then DP is in Spec-IP,
hence (DP>IP).
This amounts to saying that The always appears as a
Specifier of DP in spite of all other underspecifications--dispensing
with the alternative notion that The may remain in Spec
of NP. The notion that Definiteness is well established relatively
early in the child's development increasingly favors the notion
that interpretable features of items come on-line at the onset
of the early two-word/multi-word stage.
Tense and Agreement Inflections on Verbs/Aux emerge around
File 8 (2;4). The total count of [+Def] DPs numbers well into
100 tokens (starting with files 2-7).
(No Functional Categories) Though one might think
that the No Functional Categories (=Structural Deficit) approach
would analyze examples (4i-p) as simple VP projections (since
no functional categories are visible), examples (4a-h) could,
however, be potentially analyzed as IPs owing to the presence
of D(P). (Though recall that the very few early nominative constructions
here (cf. *e) were considered as semi-formulaic in nature:
e.g., "I want+N" (see note 6)). Based on an
extreme and highly oversimplified view of the correlation, one
could build a case, appealing to the Featural Deficit
approach, suggesting that all clauses containing a D(P) must
be considered as potential IPs--notwithstanding some feature
non-specification of I--since functional categories (here being
D) have come on the scene.
Such views, I believe, rely on a confusion that all types of
Ds--regardless of their featural specification--must involve
movement to a Spec-Head configuration within a Functional Category--be
it I or Agrs, Agro.[viii]
Chomsky (1995: 262), however, is clear on this point regarding
the biuniqueness of the phi-features of DP: regarding movement,
although it is the case that all formal features of D(P) must
involve movement into a functional domain (for checking purposes),
such movements do not necessarily apply 'across-the-board' with
respect to those ordinal categorial-features of D. In this sense,
the ordinal phi-features of D resemble phi-features of N and
don't require checking. Though Chomsky asserts that a sort of
pied-piping applies activating the entire DP to move
along with one or a number of its formal phi-features (presumably
case is the leading motivation for movement), such pied-piping
of the category works only when a formal feature of the category
is obliged to move in the first place. In other words, nothing
should force a default or inherent-cased DP to
move, since none of its formal features are present requiring
checking. This, in fact, is the case of a DP with only its [+Def]
features specified (cf. DP>VP). In light of this, the same
debate arises in how to accommodate (i.e., keep with principles
of Economy, etc.) a seemingly functional category (DP) stripped
of its formal features. Demonstrating potential pitfalls, consider
how such inert-DP constructions (ex. 4a-h) might be analyzed
via the two hypotheses under consideration:
(4')
It is clear that a parsimonious VP-projection suffices in handling
the lexical material in (4'a-e). Specifically, the VP analysis
does not require the added (superfluous) stipulation that all
clauses are obliged to project from a IP regardless of whether
or not their Spec/Head has any lexical content. The above VP
analysis of a DP +def,-f/-case] could be taken as having a Default
Objective status (cf.§1.2)--paralleling the following structure
where the DP carries the default/Objective case:
(6)
a. [VP Daddy [V'[V seeing] DP the boy]]
a' [VP Daddy [V'[V seeing] DP him/her/us/*He/*Her/*We]]
It is clear from (4 & 5) that (1b) cited earlier cannot
stringently apply--this is taken from the above observations
in (6), which demonstrate that specification of both Subject
and/or Object DPs for Definiteness may endure even within 'underspecified-for-finite'
environments. However, it must be clearly stated that in accordance
with Structural Uniformity, once an IP-projection is
acquired, we must take it for granted that IPs are henceforth
always projecting--notwithstanding underspecifications pertaining
to certain INFL-related features. This amounts to saying that
the child should never exhibit a lexical VP-stage simultaneously
with a functional IP-stage (in what had once been described
in the literature as a sort of bigrammaticality): once
IPs project, the "VP-stage" is left behind to dwindle.
(But see Radford 1990: 290 for questions raised along these
lines.)
However, a similar fundamental issue remains outstanding here.
Specifically, how can we reconcile our VP approach here in light
of recent remarks (Chomsky 1995: Ch.4) which suggest that all
clauses are to be (minimally) assigned an abstract T(ense) via
a T-operator (also see §1.2.2 below). This view was raised with
respect to the notion that non-finite verbs in adult syntax
are tense-bound--i.e., c-commanded by a T constituent.
While The Feature Deficit proposal escapes such problems
by claiming that the entire structure of the clause is available
in theory--including a T-operator, The Structural Deficit
proposal needs a further argument in its defense since all lexical
VP-clauses certainly would lack an overt T-operator--otherwise,
how would we account for the obvious empirical lack of tense
in VP-clauses? In defense of the Structural Deficit view, one
could arguably assert that Functional Categories are
introduced into derivations either by a PF merger operation
or by an LF merger operation (p.c. Radford). In other words,
at the pre-functional/thematic VP-stage of language acquisition,
the T-operator (or T(ense) P(hrase)) could be present and functioning
at LF, but not at PF. (Wexler's OI-stage would then result due
to variable TP alternations at LF and/or PF. Moreover, it is
not implausible that the two halves of an IP may be split via
truncation--with AGRs
being projected into PF and TP into LF. (Also see §1.2.2. for
a treatment of a split INFL)).
(Null DPs) While the above section examines the
early emergence of DP, it remains incumbent on us to account
also for the apparent lack of DPs in required contexts. It was
noted in (5) (Table 1.1) that e.g., (file 7) had 78% correct
overt Determiners--this obviously translates into the fact that
22% were also omitted. Moreover, the fact that Null/Non-specified
DPs occasionally do co-occur within the overall (unambiguous)
IP phrase (e.g. with Nom Case, Finite Verbs (cf. 7d-f), I think,
lends additional support to the notion that no general correlation
exists between the Head features of Subject-D(P) and Spec of
INFL. In accounting for Null DPs, the present model being
envisioned here would maintain that D, similarly (and independent)
to I, can undergo (autonomous) underspecification of its D features.
Hence, as a consequence, erroneous Default/Objective Determiners
(cf. §1.3.2 ex.14' for genitives) as well as a total omission
of D itself may ensue. Consider the following token examples
that show Null DPs:
(7)
Null DPs |
a. I want [DP ø car] (file 7) |
d. I am [DP ø pig] (file 19) |
b. I want [DP ø plane] (file 8) |
e. You are [DP ø red man] (file 25) |
c. I want to close [DP ø door] (file 24) |
f. She is [DP ø baby] (file 25) |
(7')
The above analyses suggest that Def(initeness) is non-specified
(ødef) (=having no specific definiteness properties (Radford:
class lectures)). This was taken as a natural extension of Hyams'
account that sought to formalize the feature specification of
D. The argument follows from the observation that the Definiteness
Paradigm involves a Ternary setting: (i) [+def]
(The boy), (ii) [-def] (A boy), (iii)
[ødef] (boy). (Longobardi (1994) has argued that
predicative nominals--in the sense of (iii) [ødef]--can be NPs,
contra the general DP-analysis as espoused above.) It is important
to note here however that no consequential implications emerge
out of analyzing such missing Ds as Null DPs (rather than simply
analyzing them as bare Nouns/NPs). This follows from previous
analyses (cf. §1.2) which suggest that no general correlation
holds between IP and DP--that is, DPs could be equally analyzed
as a functional category (whenever associated with IP) or, alternatively,
as a lexical category (whenever associated with VP). In this
broader sense, nothing is theoretically lost by analyzing them
as Null DPs, and, child-to-adult Continuity is achievable at
no expense.
A number of complications arise from the above cited non-correlation.
For instance, as in the adult syntax, the child's syntax can
generate those specific features that fall under the D node:
e.g., definiteness/case/number/ person/gender. However, unlike
the adult syntax, in child syntax one or more of the feature
specifications can remain underspecified. The lexical entry
specifies the features which items can carry--for instance,
suppose The= [+Definite]. This amounts to saying that
whenever the lexical entry The is inserted under the
D node, only the definite feature will be present: i.e.,
the +Def feature will manifest irrespective of number (plural
or singular), gender (masculine or feminine), or case (subjective
or objective). (I leave open here the issue of whether or not
The/A carry inherent third person or is personless. Similarly,
although A= [-Def, singular] its distribution of features
would remain similar to The as stated above.) In the
above sense, The strictly means "presence of
definiteness", but may tell you nothing of the specification
of other possible features--[+Def], being the sole feature of
the child's lexical entry, will not identify with those features
that remain underspecified within the D node.
The above notion of feature specification should lead to a
number of predictions.
First, the absence of the child's Determiner (The)
should always mean the underspecification of Def. Conversely,
its presence should always indicate Def.
Second, regarding the specification of number--as it is related
to D--count nouns may carry number while still being underspecified
for Def. For instance, when a child omits a determiner and says
I want car (cf. 7a), car may only be marked for
number. (The same distributions of feature specification would
likewise hold true for the possessive "s" under
the D node for genitive constructions (see below)).
1.2.2 Case
(Distribution of Nominative/Accusative Case)
INFL/Case reports taken from my Data-base suggest there to
be Two Developmental Stages of Case assignment: (i)
An early, predominantly No-Inflection/Caseless-Stage--where
mostly nominal elements are used,*
and (ii) An Optional-Inflection/Case-Stage--where
Case assignment is seen as fluctuating between two usages: a.
Nominative vs. b. Accusative (default).
(Stage-1) The main characteristic of Stage-One
is the evident lack of any overt Case assigning Agreement (or
Tense). Specifically speaking, all of the utterances found in
files 1-7 use either (i) Caseless Nominals, [N+N]
Genitives (cf. ex. b,c),[ix]
or demonstrate some sort of Semi-formulaic Nominative
construction (see note 6); while (ii) Tense Inflections
(e.g., 3per/prs. +s, ed) are left omitted (see §1.2.3
for Table). Consider the following token examples taken from
files 1-7:
(8)
(8')
* A total number of 9 Case (stereotypic) constructs
were reported among Stage-1 files: 1-7 (1;9-2;3), all exclusively
showing "correct" Nominative Case contra the relatively
late emergence of Accusatives found in file: 11 (2;5) (see note
6). (Acc Case typically is the first case realized by the English
speaking child.)
In examining the data, it is indeed very difficult
to determine whether or not the child has developed any sort
of Case System at all in these early files. It is however interesting
to note that the data differ somewhat from Radford's and Vainikka's
findings (op.cit.), in that (with my data) no Genitives or Accusatives
seemingly appear at stage-one (correctly or incorrectly). (Recalling
that Radford reports the first emergence of Case to be predominantly
Objective/Accusative with some sporadic incorrect usage of Genitives--I
might liken such a stage to my stage-2 (optional infinitive
stage) as described below.) The semi-formulaic strings of Nominative
["I want" +"Noun"] likewise
remain inadequate for the determination of Case here. These
frame-utterances seem to be unanalyzable strings.
One very interesting correlation which seems to verify the formulaic
make-up of such strings is the observation that no Accusative/Genitive
Case errors are seen to emerge in this first stage (other than
the two reported above with asterisks for File 3). The observation
that the stage is predominantly "error-free" increases
the plausibility that, indeed, no real case system has yet been
assimilated (the fact that abundant case errors emerge with
File 8, I think, eventually signals the real onset of a formal
case system for the child). Similarly, the emergence of what
otherwise seems to be evidence for an IP warrants some scrutiny.
All instances of the early Copula Is (cf. 8g--numbering
four in all) make-up formulaic Copula+Subj (VS) orderings and
entirely lack any productivity. (see note 7).
Following, then, the notions previously laid out in Chapter
1 regarding e.g. Principles of Economy of Representation
(Chomsky 1989), Minimal Lexical Projection (Grimshaw
1993a,b), etc.; the No Functional Category/Structural Deficit
Approach (cf. Radford, Vainikka, Rizzi) seems to be the most
plausible alternative in describing such an initial Caseless
stage. Otherwise, we would be forced into making a number of
unwanted assumptions: most notably--given a full-fledged Case
system is attributed to IP--the unwanted assumption would be
that all phrases (IP, and possibly CP) would equally project
regardless of whether or not their Heads are lexically (or vacuously)
filled. It is at this juncture, and for the above stated reasons
of Economy, that we adopt the No Functional Categories
Hypothesis for describing our very earliest Stage-1 (files
2-7) of Case development--eventually resorting to the Feature
Deficit/Optional-Infinitive approach in order to account for
the later files which make-up our Stage-2 (files 8+).
The benefits of acknowledging such an early Stage-1 are two-fold
and will become clearer as we move on--as it holds important
consequences for how we will later deal with the questions of
Bilingual Language Separation (Chapter 6). In
acknowledging such a stage, firstly, we can uphold the position
that only lexical categories (e.g., Ns, Vs) project--with the
added stipulation that DP>VPs may ensue as a lexical projection.
Secondly, a Lexical/Prefunctional Stage-1 holds consequences
for word orderings and parameterizations: viz., since no functional
categories are present, movement operations are disallowed--thus,
any variant word ordering must be derived via base-generation.
Let us be clear on one point, however, before moving on to
Stage-2. It does not necessarily follow that in endorsing such
a stage-1, we, in the process, falsify the Underspecification
Hypothesis (represented here as stage-2). It is my understanding
that the Feature-Underspecification Stage (i.e., Optional-Infinitive
stage, cf. Wexler) is meant to capture the notion of optionality
of feature projection. In order for an OI-stage to exist, by
definition, the particular feature has to have at least emerged
at some point within the general clausal development of the
child. In other words, where the acquisition of a certain
feature has taken place but perhaps where the mastery
of the feature projection/realization has not yet been achieved
by the child. (In the above sense, the one-word stage--with
no syntax or manifestation of features of which to speak--most
certainly could not be described as an Optional or Underspecified
Stage in any sense of the term being used here). The Structural
Deficit Hypothesis, however, establishes that an early (multi-word)
stage indeed exists--prior to Underspecification--where Underprojections
of Features/Categories manifest. I believe my stage-1 represents
such a stage. Furthermore, clear evidence has been given to
suggest the validity of a Subject VP-internal Stage-1:
evidence taken from the early placement of Negation demonstrated
that the Subject had to be in Spec-VP. If we were correct in
assuming the Optional/Underspecification stage throughout, we
would need to account for such Neg initial constructions. This
observed stage-1 jibes with Radford's Structure-Building
account that describes a bottom-up method of language
acquisition--
[C]hildren's initial clauses are VPs; later
they form extended projections of VP into IP (resulting in IP>VP
structures); still later they form a further extended projection
of VP into CP (resulting in CP>IP>VP structures. When
extended projections are first formed, they are optional: hence,
children in the early IP stage alternate between IP>VP and
VP (Radford: class lectures, '97)
(Stage-2) The frequency counts of the Optional
Infinitive Stage (=OI) that consists of files 12-17 is presented
below in (9) Table 1.3.[x]
(9) Table 1.3
The widely reported Subject/Object asymmetry,
as evident in the literature, is likewise borne out in my data.
As the far right column points out, there are no reported instances
of Nominative Case being wrongly assigned in Accusative contexts.
However, the converse is evident. In all but one of the Accusative
Case error examples, the verb is either ambiguously marked for
finiteness (e.g., Me work) or is overtly nonfinite.[xi]
Some token examples of (9) are given here in (10), and are analyzed
accordingly in (10'):
(10)
Nominative Case [-T/+Agr] |
Default Accusative Case [-T/-Agr] |
a. He cut the tree (=pres.) (file 21: 3;0) |
e. Me kick (file 13: 2;6) |
b. I play a water (=past) (file 23: 3,3) |
f. Eat me (VS) (file 17: 2;8) |
c. He get a bat (file 24: 3;4) |
g. Me get it (file 21: 3;0) |
d. He do it (file 25: 3;6) |
h. *[What] him doing? (3;6) |
|
i. Him gone (file 25: 3;6) |
—————— |
j, Him eat (file 25: 3;6) |
(10')
In (10'a-d), IP is both well Structured
and Utilized due to the correct usage of abstract Nominative
Case--an assignment that typically is done via movement into
a Spec-Head relation with INFL (but see Pierce (ibid.) for an
alternative account). The Subject is seen as moving out of Spec-VP
in order to check-off its Strong Agr-Case features (presumably
its strong D-feature in accordance to EPP). It must be said
that the Nominatives/Accusatives reported here at stage-2 are
fully productive--i.e., the Nom paradigm I/You/He-She
is complete along with its optional agreements--unlike what
was seen in stage-1 where only a sampling of the stereotypic
"I+want+N" along with no Accusative Case markings
were reported. Moreover, unlike stage-1, the correct distributions
of Nominative and Accusative Subjects in stage-2 surely suggest
that the Case system has emerged. We can now tentatively conclude,
by these observations alone, that stage-1 is without Case. The
overriding questions (i) how the child delearns
this rote-learned Nominative and (ii) how she eventually
obtains the proper Case-driven (Nominative) grammar can only
hoped to be answered by first fuller understanding of the intrinsic
modules of the brain which underpin these distinctions laid
out between the two stages: (viz., thematic-lexicalism vs. functionalism
for stage 1, 2 respectively). (Interestingly, the same question
equally applies to adult grammars--e.g, how do simple-form-meaning
constructs differ with respect to their formal categorial
counterparts--recall, even adult grammars often make productive
use of primitive constructs such as Idiomatic expressions
(Kick-the-bucket), "Tourist" foreign
vocabulary rote-learnings (based purely on phonological-pragmatics)
(Arrivederci, It.) etc. Let it suffice to say that though
we have no clear picture (to date) of how the child initially
processes her language (be it by semantic bootstrapping or by
other cognitive means), we can, I think, nonetheless say that
the child (by stage-2) has now seemingly turned-on her
formal grammar much in the same manner as the adult.
Of course, the typical debate ensues regarding
whether or not Default Accusative Case constructions without
Tense--[-Agr,-T] (cf. 10e-j)--should constitute as an IP (with
tense and agreement unspecified). It is apparent that at stage-1
they do not. However, conditions placed on Structural Uniformity
would lead one to suppose that once a functional category has
been acquired (at PF), they henceforth must structurally project
(at PF). Thus, by definition, the OI-stage must describe all
projections (minimally) as IPs notwithstanding feature underspecifications
having to do with IP itself. Moreover, a second issue remains
outstanding here regarding the positioning of such underspecified
Subjects at the OI-stage. In light of previous discussion concerning
the Principles of Economy as well as with tentative findings
presented earlier (recapped here) regarding Neg initial constructs,
we shall make do with the null-hypothesis and assume
that such Subjects may remain within Spec-VP whenever possible
(ex. 10 e-j).
(Neg Initial Recap: Insert) Let us
digress a moment to review data on Neg initial constructions
as presented earlier in--in recapping our findings, we may determine
the plausibility of the Subject to alternate between (in-situ)
Spec-VP and Spec-IP positions (at the OI/IP-stage). The findings,
as a whole, suggested that Neg initial constructions could Head
both lexical and functional projections: NegVP and NegIP respectively.
The distributions were seen as follows:
(11)
Files 21-24
Neg initial
VP Neg initial IP
a. No [VP me get
it] d. [IPi...No
[VPi I got
it]]
b. No[VP it broken]
e. [IPi...No
[VPi He sleeps]]
c. No [VP baby
do it] f. [IPi...No
[VPi it works]]
The token examples above suggest that even at the IP-stage,
early Nominative Subjects may remain in situ in Spec-VP.
Pierce (1989) cites a number of examples that suggest the same.
She concludes that, at the early IP-stage, children may fail
to raise the Subject from out of Spec-VP. She further suggests
that the structural Nom Case appearing on the subject has been
assigned under a government relation with IP (here indicated
by the trace). However, there are a number of problems with
this approach--two of which are most notable: first, the Discontinuity
between child-adult grammars; and second, the notion that Neg
escapes barrierhood status for Nom case marking (cf. Adjacency
Condition) (see Radford: 1994ms for further discussion on this
topic). However, such problems may become less costly if we
can also find further evidence to support Pierce's claim--presumably,
in examining whether or not the Aux(iliary)/M(odal) is restricted
from such NegIP constructions. In other words, if Subjects (in
Spec-VP) were to precede the Aux/M of such NegIP constructions,
(e.g., No I can go/No I do make, etc.) then a case could
be made (against Pierce) that Neg is, in fact, situated above
IP within Comp (i.e., taking an IP as a complement (cf. Stromswold:
ms1996, Laka: ibid)). However, this is not borne out in my data
(nor is it extensively borne out in data elsewhere):[xii]
all instances of Auxs/Ms are correctly positioned after the
Subject and before the Neg (e.g., I do no(t) know...etc.
(see (20) below for Aux-Do Insertions). The correct positioning
demonstrates overt Subject raising into Spec-IP and rejects
an alternative Neg-in-Comp analysis--favoring Pierce's account.
In maintaining an IP-Structure contra a full IP-Utilization--since
default case is assumed) for examples (10e-j)--we keep within
the spirit of the Underspecification model as mentioned above.
E.g. proponents of Structural Uniformity would rightly
argue if some clauses are IPs (at stage-2), then all
must be (at least) potential IP-structured projections (at stage-2).
However, owing to the dilemma faced here in assuming that the
(Default/Acc-Cased) Subject originates and remains within Spec-VP,
(since there is no empirical evidence of its movement for purposes
of checking), we are rather cornered with projecting a theoretical
and abstract IP (indicated by (ø)) for Spec-IP). One important
implication here, regarding an abstract IP, is that there may
be some motivating factors having to do with an IP-driven T(ense)
operator (Chomsky: 1995). Speaking directly to the uniformity
condition (as mentioned above), some support is gaining in favor
of positing that IPs must always project (at PF) once they are
acquired (not mastered) by the child. This follows from
the fact that unlike a weak [-Agr]--which basically reduces
to a non-Agr [øAgr], given Agr crucially and exclusively
depends on strong feature for its existence--T essentially remains
strong throughout a given derivation. (This distinction of permissible
strength variance between T and Agr may be due to the nature
of (EPP), as well as with the semantic nature of T itself.)
In other words, T elicits and obliges its Spec position [Spec-T]
to be filled by a strong D-feature. Chomsky (ibid: 282) notes
that while the (EPP) may be divorced from Case, all values of
T (weak or strong) induce the (EPP) in English--i.e., TP can
project infinitival as well as null-cased clauses.[xiii]
Arguably, once a Spec-T has been projected at PF by the child,
the category [T] must thereafter remain at PF. The fact that,
at the OI-stage, the child may treat T as underspecified only
refers to its feature strength--the functional category [TP]
however remains intact within the phrasal projection (e.g.,
TP>VP) with its feature labeled as [T-weak] (see §1.2.1.
for [T] at the VP-stage). Hence, I make use here with a trace
in Spec-IP for such cases only to capture this theoretical T-operation
while, to all intents and purposes, not committing myself on
any further "real" utilization (per se) for
the IP. Moreover, in support of the abstract IP analysis for
stage-2 here, it has been suggested in the literature that children
may not initially set feature strengths correctly at the OI-stage:
i.e., whereas, if in the target grammar the relevant feature
is strong, children may initially set the relevant feature as
weak or optionally strong: (alternatively, a default mechanism
may initially set all UG-P(arameterized) feature values as non-specified
or weak [ø/-UG-P] from the outset, awaiting further input).
In short, there are implications here for how we can account
for observed Subject VP-internal structures seemingly at the
Optional-Underspecification IP-stage. Since a weak Agr has been
selected (cf. 10'e,f,g) no morphological movement can be attracted
to it. Firstly, Agr can only exist when it has strong features
attracting only overt movement--Nb. Agr can't attract covert
movement (cf. Chomsky op.cit: 351). Secondly, however, since
Tense is obliged to project regardless (at PF or LF), the Subject
may covertly raise at LF (within Spec-T--as seen at the
VP-stage) where T may or may not maintain a strong nominal D-feature.
Next
Page >>
[i] See Radford (1997: ms) for an alternative analysis suggesting
that Genitive Case (e.g., My subjects) are actually Analogical
Nominatives--i.e., the child taking the /m/ genitive stem prefix
and adjoining it to the nominative form /aI/ yielding /maI/ (my).
Radford (pc) thus believes that Early Child English Grammar really
only makes use of Nominative and Accusative Cases. (See note 9).
[ii] Of course, one could equally assume underspecification
(viz., IP) by claiming the following: (i) Det is specified
for Def/Case but not for Number; (ii) Det is specified
for Def/Number but not for Case.
[iii] This observation matches that of Wexler's (1996) who
similarly proposes that D-features of DPs are optionally [-Interpretable]
during the OI-stage. Wexler notes that such a D-feature may be
entirely non-syntactic--alluding to the notion that such Ds are
consistent with child DPs and resemble substantive properties
much like N (cf. Schütze 1997: 261).
[iv] CPs might be viewed as having semantic properties based
on the idea that their Specs can host a variety of semantic features
(e.g. Question operators/quantifiers, Scope operators, etc.).
[v] The reasons why I chose to concentrate on definite Ds
(The) here is incidental and doesn't affect the outcome
of the overall analysis. Of course, the same arguments could apply
of indefinite Ds (A)--e.g., if A is underspecified
for case, it may be in Spec-NP, of which I have ample Data.
[vi] Instances of early use (files 1-7) of Nominative Case
seem to suggest a semi-formulaic character: viz., they tend to
be all 1pers-sg constructions combining "I want+noun".
Only 9 examples of early nominatives were found between Files
1-7. They include: I want--the car/the water/the bottle/this/that/down.
Productive use of Nominatives begins at around File 8 where a
wider selection of verbs enter into the construction: e.g., I
know (file 8+), I throw, I cut, I eat, etc.,
etc. One idea is that the string I-want is based
on a Piaget type volition stage-I of cognition and is actually
representing a single lexical item at PF and at LF. Budwig (1990)
has come to consider the possibility that some verbs may select
a specific type of subject (e.g., NomSubj+V/AccSubj+V) depending
on the verb's lexical-thematic properties, etc. In this sense,
the verb want may only select a Nom subject and all checking
is done internally in Spec-VP
[vii] Similar to what we find in note 6, the early emergence
of copula Be complicates the issue of whether or not this
signals the emergence of IP. The few constructions found follow
a schematic routine ["Is+N"] and hence could
be considered as formulaic in nature: Is perhaps being
interpreted and used by the child as a element of locative focus
(e.g., "This place+N" ). Furthermore, we came to the
conclusion that such early copula constructions projected semi-formulaic
VS orderings: this was concluded on the bases that the Nouns in
such VN constructions were taken to be real Subjects in light
of two considerations: (i) no evidence was found for expletives
e.g., It/there (either in null or overt form); and (ii)
the fact that children typically ground their language around
concrete topic-comment themes, further suggests that the nouns
used in these VN constructions are indeed topic/subjects.
[viii] The CP>IP>VP framework being utilized here does
not represent the entire scheme--put forward by recent minimalist
accounts (cf Chomsky 1995: Ch4)--of all possible movement operations
motivated by purposes of checking, etc.
[ix] Such caseless possessive forms might suggest a simple
NP analysis where Daddy in (Daddy truck) is in
Spec-NP being that there is no case to check. This would be
consistent with our more general No Functional Category
analysis of stage-I (see Radford 1990 for full discourse leading
to this conclusion):
[x] A steady decline in the rate of case errors seems to
begin at around File 18.
[xi] The sole example of Acc with main verb is: 'Him cries'
(File 25: 3;6). Three other examples of Acc with Copula V surface
in File 25/diary: Him is big, Him is my friend, Him is
hiding.)
[xii]But see Stromswold (ms1996) for a contrary opinion.
[xiii] Schütze (1997: 203), following Wexler, presents Tense
as being associated with Subject (Agreement) by having the following
features: [+/-finite], [+/-past], ([1p/2p/3p]), [+/-plural], selects
V (-participle). In this sense, it is not clear what the absence
of T would mean in child grammar--Schütze assumes that children
may only omit the past features while keeping to [-finite].
Schütze's approach that aims to
associate T with Subject Agreement differs with what we wish to
propose here--mainly, that T (as manifesting in the 3sg "s")
is only to be associated with the features of T and not Agr. An
overriding advantage with keeping to a Disassociated T/Agr is
that, otherwise speaking, children would need to learn (at an
extremely early age) that T and Agrs are fused together
in English. If this were not learned early on, we might expect
to find simultaneous past tense-ed (T) and 3sg-s
(Agr) errors (* e.g. daddy walk-ed-s) to occur. Such errors never
occur even in the earliest of data.
|
|