Pied-piping in wh-questions:What do children say about it?

Errors/non-target responses characterizing sub-extraction of a wh-phrase from complex DPs in child speech are found in first language acquisition studies (van Kampen 1997 among others) and have provided the basis for arguing the complexity of question formation involving pied-piping.
In this dissertation, data were drawn from 81 children, aged 3;0-6;0, participating in two experiments, with one eliciting a D-linked question in complex phrases such as inda milo ‘which apple’ in Cypriot Greek. The results validated previous literature on sub-extraction phenomena and have provided the first observation for such cases in the specific variety. Errors were characterized by movement of the operator and stranding of the noun in which+NP structures, such as ‘which apple’. Another error involved movement of the operator and pied-piping of a noun, but stranding of the second noun in wh+NP+NP structures, such as ti xroma tsenda (lit., ‘which color bag’).
Results from the production experiment show that children show high percentages of omission of the NP in D-linked questions (up to 50%) in all age groups. Their responses involve stranding of the NP (7%-17%), which does not seem to fade out even in the oldest age group. These errors appear across ages when children produce a wh-question with the wh-phrase ti ‘which’. In a set of responses, where inda ‘which’ is used, errors are found only in the youngest group and do not appear with the successful use of inda ‘which’.
A comprehension task was later administered to a subset of the children that participated n the production experiment and some of the data collected are used to compare the acquisition of D-linked questions between production and comprehension. Children provided more than 60% successful responses in the comprehension experiment showing a steady development by age. Lower percentages are found in object D-linked questions, suggesting greater difficulty in the comprehension of object D-linked questions in comparison with subject D-linked questions (Goodluck 2005 and subsequent work). Subject D-linked questions initially appear to be acquired at the age of 4, whereas object D-linked questions appear at the age of 6.
With focus on sub-extractions, the Immediate Move Hypothesis is proposed to account for these errors in D-linked questions and other environments of similar type. It predicts the ‘optionality’ in pied-piping, expands the syntactic term ‘shortest’ in the Minimal Link Condition and provides a theory of movement in children based on the smallest possible element satisfying the maximum number of requirements in syntax.
The types of errors produced by children involve a logical explanation under which fundamental notions of Minimalism, such as Economy, are expressed through different structures defining these errors as innately-motivated patterns.