Case marking variation and syncretism repairs in Icelandic ECM/raising

Prior literature claims that lexical case overrides structural case when it appears in ECM and raising constructions with verbs of saying and believing. This is, however, not (always) the case. Both for the preverbal position of the subordinate clause (ECM) and in the passive subject position of the matrix clause (subject-to-subject raising), there is considerable variation in whether the lexical case is preserved or absorbed into structural case. In grammaticality judgment tasks, lexical case preservation — especially in raising-to-subject constructions — received low acceptance rates, and corpus searches yielded very few examples of it. Similarly, the case absorption examples — where structural case overrides previously assigned lexical case — are also unexpected. However, it appears that speakers generally disfavor these constructions altogether when asked to choose between structural and lexical case. Neither option is judged as particularly acceptable, and neither variant is well attested in corpus data, especially when compared to examples in which no lexical-case-assigning verb is present. I suggest that the overall unacceptability of constructions involving a lexical-case-assigning verb in the infinitive arises from a conflict that occurs when a lexically case-marked DP moves into another case-marked position, where it simultaneously tries to preserve its lexical case while receiving structural nominative case. This is supported by judgments on syncretic forms: NOM/ACC syncretic forms in subject-to-subject raising with ACC-subject verbs receive significantly higher ratings than either unambiguous NOM or unambiguous ACC forms. This appears to be an instance of case syncretism repairing conflicting morphosyntactic requirements.