We started considering the possibility that subjects did not start out in the specifier of TP position when looking at Irish Gaelic, whose basic word order is VSO. The behavior of main verbs, auxiliaries, and the placement of Tense in Irish seems to resemble what we saw in English and French, but if the subject starts in the spec-TP position, we would still predict SVO basic word order in Irish.

We thus altered our analysis to have the DP that normally becomes the subject to be introduced by a verbal projection, located below (in terms of linear precedence, to the right of) any other aspectual auxiliaries and Tense. Since the external arguments that normally become subjects are in complementary distribution with the be verb that is an expression of passive Voice, we analyzed this verb head that introduces this argument as active Voice. We can now say that, in Irish, the DP that would normally become the subject starts in this position and remains there, giving us VSO basic word order once the main verb is raised to V, while in English and French, that DP is raised to the specifier of TP position, giving us the SVO basic word order.

This analysis has several other desirable properties:

• Our previous approach, where an argument of the verb was introduced in the spec-TP position that may actually be far from the verb, never had a good explanation. If the subject is an argument of the verb, we would expect that it would enter the structure close to the verb, since it has a close semantic relationship to the verb (similar to the kind of semantic relationship that complements have with the verb).
• This gives us a natural analysis of passive sentences, where the external argument that would otherwise occur as the subject does not occur: passivization doesn't remove an argument of the verb, but instead the passive morpheme competes for the position where that argument would otherwise be added. We can either attach the overt passive morpheme to the main verb, or we can attach the thing that introduces the external argument to the main verb, but not both.
• Although the textbook doesn't go into detail on this topic, this approach to Voice heads gives us a very natural analysis of the complex Voice systems in Austronesian languages like Tagalog, where we seem to have overt markers of both the passive and the active Voice head (and other options of Voice heads we might use in that location).

Likewise, although the textbook doesn't bring this point out, the analysis of subjects entering the tree in a Voice-phrase-internal position is consistent with our approach that meaning-related issues determine where a constituent enters our tree--such as the requirement that complements (a semantically close relationship) always enter our tree structures immediately adjacent to the head that selects them. The raising of subjects to be "near" the Tense morpheme is then consistent with the observation that the actual marking of Tense is somehow sensitive to grammatical properties of the subject: present tense, as for example in English, is actually expressed by an agreement morpheme that is sensitive to the person and number characteristics of the subject, even though logically Tense has no direct relationship to these properties of the subject.