Forming Yes-No Questions
When we form a question that includes a modal auxiliary, the modal moves from the T-head position to the C-head position:
Could you could read this for me?
When a sentence contains no auxiliary, only a lexical verb, it appears that the lexical verb cannot move out of its V-head position. So we’ve proposed that the auxiliary do enters at the T-head position and then moves up to the C-head position:
Did you did see her tattoo?
*Saw you saw her tattoo?
One piece of evidence that do enters the sentence at the T-head position and moves to C-head, rather than just entering at the C-head position, is that it bears the tense feature of the sentence: if the tense is [+past], then we observe the form did, but if the tense feature is [-past], we observe the form do or does. Another piece of evidence is that whatever verb follows the inserted do is in its bare form, not its [-past] or [+past] forms:
Did she talk to Darren?
*Did she talks to Darren?
*Did she talked to Darren?
Does she speak Italian?
*Does she speaks Italian?
*Does she spoke Italian?
Both of these observations suggest that the inserted do gets its tense morphology from the tense feature (either [+past] or [-past]) in the T-head.
English also uses do-support to form negated sentences, which follow the same pattern: sentences with modals don’t need do, but sentences with lexical verbs and no auxiliaries do need do:
I could not believe that rumour.
*I did not could believe that rumour.
*She speaks not Italian.
She does not speak Italian.
If we accept that not is in a fixed position between T-head and its VP-complement, then the distribution of do makes sense. Just like in questions, the evidence suggests that lexical verbs cannot move out of their V-head position up to the T-head position.
This pattern of how do behaves in questions and negative sentences gives us a clue about how the other non-modal auxiliaries, have and be, behave.
Notice that the verb be can always move up to C-head in questions, both when it’s a genuine auxiliary:
Are you are going to the concert?
Was she was joking about that?
And when it’s the only verb in the sentence:
Are you are serious?
Is this is the place?
Likewise, be appears before not both when it’s an auxiliary and when it’s the only verb:
You are not are going to the concert.
She was not was joking about that.
You are not are serious.
This is not is the place.
But have seems to have two different patterns of behaviour. When it is a genuine auxiliary, it behaves like be. It can move up to C-head in questions:
Have they have moved to Texas already?
Had she had already heard the news?
And appears before not in negated sentences:
They have not have moved to Texas already.
She had not had already heard the news.
But when have is the only verb in the sentence, it behaves like a lexical verb. It can’t move up to C-head and can’t appear before not.
*Has she has five sisters?
*Have you have a headache?
*She has not has five sisters.
*You have not have a headache.
Instead, when have is behaving like a lexical verb, it needs do-support.
Does she have five sisters?
Do you have a headache?
She does not have five sisters.
You do not have a headache.
The Deep Structures
From all of this evidence, we can conclude that MERGE treats these three kinds of heads differently:
Modals are generated in T-head, from where they can move to C-head if necessary to support a [+Q] feature.
Be is generated in V-head, but moves up to T-head (to the left of not) and from there up to C-head if necessary
Have is generated in V-head, and can move up to T-head and from there up to C- head only if it is an auxiliary (that is, only if it has a VP complement). But if it is the only verb in the sentence (and has no VP complement), then it behaves like a lexical verb.
Lexical verbs in English are generated in V-head and cannot move to T-head or C-head.