The structure of CPAs the complementiser is a word and the IP is a phrase, we immediately see a potential
head–complement relationship between them and if we apply X-bar principles to this
situation the structure we expect is:
(9) CP
C'
C IP
If this is so, we expect the complementiser would demonstrate certain head-like
properties and it is fairly easy to show that it does. For example, compare the
following sentences:
(10) a I wonder [if Charles likes chocolate]
b I think [that Charles likes chocolate]
The embedded clause in (10a) is interrogative as it can act as the complement of the
verb wonder and this subcategorises for interrogative complements. The clause in
(10b) is declarative as it can act as the complement of the verb think which
subcategorises for declarative clauses. But the only difference between the two clauses
lies with the complementisers. The IP in both cases is identical. This would suggest
that the interrogative/declarative nature of the clause is fixed by the complementiser
and not by anything inside the IP. In other words, it is the complementiser that
provides the force of the sentence. As it is heads that provide the properties of the
structures they head, this demonstrates that complementisers do have a head-like status
within the clause.
Another claim made in (9) is that the IP is the complement of the complementiser.
Can this be substantiated? One obvious relevant observation is that the IP follows the
complementiser and as we know complements follow heads in English. Thus the claim
that the complementiser is the head and the IP its complement accounts for this fact
about English word order without resorting to stipulation. Furthermore, we have seen
that functional heads such as determiners and inflections have a very limited range of
possible complements: determiners always have NP complements and inflections
always have v/VP complements. The element which follows a complementiser is
always an IP and so this fits the pattern very well. Finally, note that different
complementisers introduce different IPs. If and that both introduce finite IPs, while for
introduces a non-finite IP. As heads select for their complements, this is again an
indication that the complementisers are heads selecting for different types of IP
complements. This is very similar to plural determiners selecting for plural NP
complements and singular determiners selecting for singular ones.
In what follows, we will be assuming the structure in (9) as essentially correct,
though we will see that some extension will be needed for elements that appear
between the complementiser and the IP. We will start by discussing facts that do not
concern these elements however, and so for the time being (9) provides us with an
adequate model of the top part of the English clause.