What are DATIVE SUBJECTS? This paper argues that this term has been applied to two distinct constructions. In one, the surface subject is in the dative case. In the other, a dative-marked nominal that behaves like a subject in certain respects is not, we claim, a surface subject. Dative nominals of the second type have been analyzed in Relational Grammar as initial subject and final indirect object in the Inversion construction; for this reason we call them I-nominals. Neither adopting nor rejecting the initial-subject analysis of I-nominals, here we argue only that they are not surface subjects. To argue that I-NOMINALS are not surface subjects is not so traightforward, however. Where I-nominals fail to behave like surface subjects, it has been widely assumed that the subject behaviors in question are restricted to nominative subjects. This has made it possible to maintain that I-nominals are surface subjects despite their nonsubject behaviors, which are attributed to their dative case. To argue against this it is necessary to find a language where I-nominals' failure to behave like subjects cannot be attributed to their dative case. Here we argue that Russian is such a language. We show that Russian has a true dative-subject construction in which surface subjects are in the dative case. They behave like subjects in every respect. Russian also has I-nominals in the dative case which behave like subjects in only two respects; in other respects they fail to behave like subjects. This failure cannot be attributed to their dative case because true dative subjects, also in the dative case, behave like subjects in every respect. We conclude that dative subjects and I-nominals instantiate distinct constructions which must be recognized as such in syntactic typology and syntactic theory. The consequences of this result extend beyond Russian to the analysis of other languages. Where dative-marked nominals behave like subjects in certain respects, it is necessary to determine whether they are I-nominals or dative subjects, based on language-internal evidence. They cannot simply be assumed to be dative subjects, as has often been done. It takes more to be a dative subject than has generally been recognized: it takes language-internal evidence that the nominals in question are not I-nominals.
