Possessor extraction in child English: A Minimalist account

Elena Gavruseva & Rosalind Thornton
University of Iowa

This paper will argue that 3 to 6 year-old children acquiring English natively and as a second language have access to a parametric setting that makes the following structures possible in their grammar:

(1) a. Who do you think 's fish is in the cradle?
(cf. Whose fish do you think is in the cradle?)
b. Who do you think 's Spiderman saved cat?
(cf. Whose cat do you think Spiderman saved?)
c. Who do you think 's sunglasses Pocahontas tried on?
(cf. Whose sunglasses do you think Pocahontas tried on?)

The patterns in (1) are possessor-questions which were produced by children in an elicited production task set up as a 'guessing game' with the puppet. 12 English-speaking children and 2 Russian-speaking children participated in the experiment. A total of 154 possessor questions were produced by L1 children and 28 such questions were produced by L2 children.
The structures in (1) are interesting in that they show that some children ask possessor questions not by moving the entire whose-phrase ('whose fish', 'whose cat', or 'whose sunglasses') to the matrix COMP as is required in adult English, but rather by fronting only the constituent 'who' (i.e. the wh-possessor). We will argue that the patterns in (1) should be taken as evidence that extraction of wh-possessors is a grammatical option available to children via Universal Grammar.
Our argument rests on the crosslinguistic work exploring the syntax of possessor questions in a variety of languages: Tzotzil (Aissen 1996), Mohawk (Baker 1996), Chamorro (Chung 1991), the Germanic and Slavic languages (Corver 1990), Hungarian (Szabolcsi 1983/4, 1994). The existing work suggests that there is a three-way division between languages with overt wh-movement, that is, possessor questions can appear in one of the three forms:

(a) split (as in Chamorro), which means that a wh-possessor is obligatorily extracted;

(b) optionally split (as in Hungarian, Tzotzil, or Russian), which means that either a wh-possessor is extracted or the entire whose-phrase is pied-piped;

(c) only pied-piped (as in the Germanic languages), which means that the entire whose-phrase must appear in the matrix CP.

We assume that this three-way split between the surveyed languages reflects the parametric values of the possessor extraction parameter. Thus, it is logically possible that children's acquisition of whose-questions will be consistent with each of these values. This is exactly what we found in the L1 and L2 data. Of 12 native English-speaking children, four kids split all possessor questions, six kids produced both split and pied-piped questions, and two pied-piped exclusively. Two Russian kids acquiring English as L2 produced only split whose-questions, despite the fact that their native language allows both a split and a non-split option. Thus, both sets of data support the analysis of the patterns in (1) as reflecting a UG-compatible grammatical option. This conclusion is contrary to Snyder et al. (1997, 1998) who argue that left-branch violations in child English (and split whose-questions count as such) have no linguistic motivation.
The split possessor questions in child L1 and L2 English have several characteristics that raise a host of interesting theoretical and developmental questions. First, native English-speaking children utilized a wh-possessor extraction option only when asked to produce long-distance questions. This finding can be interpreted in at least three of the following ways:

(a) it reflects a more advanced developmental stage, where possessor extraction is restricted to long-distance contexts (the stage at which children split matrix whose-questions remains to be empirically determined);

(b) it reflects another parametric setting which permits possessor extraction only in long-distance contexts (this is consistent with Baker's (1996) observation that possessors in Mohawk are allowed to extract in long-distance but not in matrix questions);

(c) it reflects the fact that the presence of the medial COMP in a syntactic configuration is somehow crucial for possessor extraction to be allowable.

We will argue that the interpretation in [c] is theoretically plausible in the feature-checking theory of movement (Chomsky 1995). We will suggest that wh-possessors can be extracted in long-distance questions because the medial C0 provides an alternative checking domain for the Case/D-feature of the wh-possessor 'who' in child English. We argue that the C0 in child English can be optionally assigned a [-Interpretable] strong D feature as it is drawn from the lexicon into the numeration (Chomsky 1995). The strong D feature on the medial C0 needs to be checked/eliminated before Spell-Out. This checking operation is performed by the wh-possessor 'who' which carries Case/D-feature. The overt 's in split questions in (1) is then a morphological spell-out of the strong D-feature in COMP, as shown schematically in (2):

(2) [CP Who do you think [CP t' ['s=strong D [TP Spiderman [vP saved [VP [DP t cat]]]]]]].

The structure in (2) shows that 'who' proceeds to move to the matrix CP to check its [+wh] feature. This analysis accords with recent syntactic analyses of successive-cyclic wh-movement that propose the presence of strong D feature on the medial COMP (Fanselow & Mahajan 1996). The data on split possessor questions provide additional empirical evidence for such proposals as well as suggest that wh-possessor extraction in child English proceeds via successive-cyclic movement.
Thus, we propose that there are two possible ways for the wh-possessors to check their Case/D-features in child English: (a) in the checking domain of the medial C0 and/or D0. The proposal is supported by the distribution of the morpheme 's in split whose-questions. The data show that 's is consistently spelled out as a clitic on the matrix verb think. The questions where 's is spelled-out on the lower verb are unattested in the child data ("*Who do you think Spiderman saved 's cat?").
Finally, we will consider a number of developmental questions concerning (a) the default setting of the possessor extraction parameter; (b) how children unlearn the checking of D-feature in the domain of C0.

< Back to List of Abstracts | Back to PLC23 Home Page >

About the PLC23 Committee
Previously held Penn Linguistics Colloquium: PLC22 (1998), PLC21 (1997)

Penn Department of Linguistics
University of Pennsylvania