Why the negation movement rule that negation in the embedded clause may move to the higher clause cannot be applied to the Korean bi-clausal structure where the negative concord item (i.e., NCI) undergoes the so-called exceptional case marking (i.e., ECM) is because the ECMed element occupies some position in the matrix clause. When the so-called CP anaphora kulehkey ‘so’ replaces the embedded clause of the ECM construction, the ECMed NPI cannot be incorporated in it. These observations are in accordance with Lee's (2006) argument that the ECMed element is in a non-thematic argument position of the matrix clause (i.e., Spec of the matrix vP). Merchant's (2004) and Park's (2013) suggestions that the invisible negative head in fragment constructions can only appear in the matrix clause, but not in the embedded clause reveal the secret that there arises a grammatical contrast between fragment answers to an ECMed wh-question and a matrix question containing a wh-question in the embedded clause. The NCI fragment as a response to the ECMed wh-question is slightly degraded but acceptable, whereas the NCI fragment as a response to a matrix question containing other wh-questions in the embedded clause is considerably degraded and unacceptable.

• Doo-Won Lee(Korea National University of Transportation)