We argue for the ideas of McCloskey 1996, who claims that the EPP and nominative Case are features checked by two distinct heads T and Agr respectively, which crucially can be separately active or inactive. Similar claims are found in Carnie (1995) and Harley (1995). In this paper we demonstrate that if the clausal architecture argued for by McCloskey is correct, we are forced to rework the standard account of the distribution of PRO. In particular, we draw the following three conclusions:
i) Case assignment may not be dependent upon or linked to Tense.
ii) Since the distribution of PRO in languages like English is linked to tense, the conditioning factor governing its distribution can not be Case.
iii) The conditioning factor governing the distribution of PRO in infinitivals is the EPP.
If the above conclusions are correct, we make two strong predictions, listed in II:
i) Languages demonstrating no EPP effects will permit overt nominals in the subject position in infinitivals.
ii) PRO is case marked in the same manner as any overt NP.
We show that the first prediction is true of Irish, where overt nominals are always possible in the subject position of infinitivals. The second conclusion is borne out by evidence from case agreement facts in Icelandic (SiggurDson 1991).