Consider the following synonymous sentences:
1.1) That the girl plays on the street might be dangerous. 1.2) For the girl to play on the street might be dangerous.
In 1.1) there is a matrix clause "X might be dangerous", which has a finite embedded clause as a subject, namely "that the girl plays on the street". In 1.2) the matrix clause is the same as in 1.1.), but the subject is a non-finite clause. Let's take a closer look at the embedded sentences:
1.1.1) That the girl plays on the street 1.2.1) For the girl to play on the street
The subject sentences have their own subjects - in both cases it is the girl. As of their surface form, these NPs are indistinguishable. But notice what happens once we replace the full NPs with pronouns:
2.1) That she plays on the street 2.2) For her to play on the street
In 2.1.1) the subject pronoun bears NOMINATIVE Case. In 2.2.1), by contrast, the subject pronoun bears ACCUSATIVE Case. In GB it is assumed that, even though there is no surface manifestation of Case on the subject-NPs in the examples 1.1.1) and 1.2.1), these NPs bear abstract case which is, of course, the same as the ones borne by the correponding pronouns. If two case forms have the same morphological realization as in these examples, we talk of case syncretism. The identical occurances of the girl in 1.1.1) and 1.2.1) "hide" their distinct Cases. By contrast, in the examples where the subjects are pronouns, the Case-difference can be read off the pronouns morphological form (she vs. her). The underlying form of the sentences 1.1.1) and 1.2.1) is thus:
1.1.1) That [the girl]NOM plays on the street 1.2.1) For [the girl]ACC to play on the street