It's still necessarily got a worst case complexity of O(n) - e.g. consider the case where the contents of nodes are perfectly disjoint. In that instance, you'd require O(n) space.
EDIT:
Sorry, wait, I was correct all along - as I was talking about the worst case complexity! Consider the case when our alphabet is {A,B}, and I want to store the strings "A" and "B", in that case, I have to use O(n) space, as I'd need at least two nodes for the two strings.
Here the input space is a set of dictionaries. A dictionary is a finite set of strings. A subset of the input space is a set of dictionaries.
Finite subsets of the input space are finite sets of dictionaries, thus they have bounded size.
That's just a specification of an arbitrary subset of the input space.