Extraction from subjects (under reconstruction)

Recent research on subject islands has shown that not all subjects give rise to equally robust islands effects. At least three distinct factors have been argued to induce islandhood: (i) Derived position (vs. thematic position), Stepanov (2007); (ii) external argument (vs. internal argument), Chomsky (2008); (iii) D-linking (Jiménez Fernández 2009). These different assessments agree on two cases: on the one hand, unmoved (and non-Dlinked) internal arguments are transparent for extraction; on the other hand, D-linked external arguments in a derived position are islands; however, all other cases of extraction from subjects constitute a grey area where acceptability judgements are unstable (cf. Jurka 2010). Taking this observation very seriously, we note that the first and second factor, which refer to easily detectable structural properties of the subject, are not compatible with such variability. Our starting hypothesis, then, is that D-linking (presuppositionality) is the crucial factor, and the other two are relevant to the extent that they contribute to determine presuppositionality.