The present paper addresses the linking problem, i.e. the possible mappings between thematic roles and grammatical structure. It is part of a larger project with the object to show that Merging Tense with Verbs (including little v), henceforth the MTV-hypothesis, derives the backbone of the clause, as in (1):

(1) [_CP ..C [_TP ..T [_vP ...v [_TP ..T [_VP V..]]]]]

In my talk I will show how the MTV-hypothesis enables me to account for an apparent paradox in relation to the linking problem. Taking Baker's UTAH to say that there is a one-to-one relation across languages between the position where a DP is externally merged and its thematic role, we do not expect to find variation in the ways different languages express the argument frame of different types of verbs. However, as is well known, such a variation exists, consider e.g. the presence or absence, even in closely related languages, of impersonal passives, di-transitive verbs, null subjects, oblique subjects, etc. It is the main purpose of this paper to show how such a variation is handled in a minimalist grammar in which the MTV-hypothesis is implemented.

I will focus on two areas of cross-linguistic variation, the cross-Germanic variation with respect to existential sentences, and the Romance-Germanic variation with respect to null-subjects. English is most restrictive with the existential construction, allowing it only with some types of monadic verbs, lacking impersonal passives (*There was danced on the ship for five hours) and transitive expletives (*There ate a man the shark). The mainland Scandinavian languages have both impersonal passives and existentials with a greater range of monadic verbs than English, but lack transitive expletives; these are only found in German and Icelandic. The MTV-hypothesis predicts that impersonal passives are available only if the language has access to an expletive with the features [f , ut ] that can be merged within VP. In mainland Scandinavian this function is found with the various expletives (det, der), whereas English there can receive a f -feature only if it is inherited from a DP. In German and Icelandic, agreement has the same function as the expletive in the other languages. The whole range of variation follows from these lexical differences.

A comparative problem with respect to subject-verb agreement is that some languages with agreement (eg., Italian) are null-subject languages, others (German, Icelandic), are not. I will give these languages the same analysis, arguing that the agreement ending has the properties of an argument. Merged in Spec-vP, it is in an Agree-relation with the higher T. As an affix, it must attach to a word; this forces V-to-T. I will argue that the agreement affix is pronominal in the true null-subject languages, but anaphoric in the non-null subject languages. Binding Principles A and B thus force the agreement ending to be locally bound in German and Icelandic, but unbound (in its binding domain) in Italian. One piece of support for this is that a pre-verbal "subject" in null-subject languages is in an A-bar position to the left of T, whereas it is in an A-position in German and Icelandic. Hence a visible subject DP will bind the agreement affix in the latter type of languages, but not in the former type. As I will try to show, the presence of oblique subjects in Icelandic and its absence in German, although these languages have very similar case and agreement systems, will also follow from a small difference in the status of the agreement affixes.