As we saw in chapter 1, phrase structure diagrams are used to represent several different kinds of information about the structure of a sentence: (i) word order; (ii) constituent boundaries; and (iii) the category of each word and constituent in the sentence. It is important to remember that these tree diagrams are just pictures which are supposed to represent certain linguistic properties of real sentences. Before we can draw the pictures, we first have to understand the linguistic reality which they are intended to represent.

For example, we stated that a Constituent within a tree structure corresponds to “all and only the material which is dominated by a single node.” But this is primarily a statement about how we draw trees. In order to apply this definition, we first need to be able to find constituents in real language data, i.e., to determine which groups of words function as a “unit” in a particular sentence. Our decisions about how to draw the tree are based on observable facts about the language.

This is a specific instance of a general point which seems obvious, but is nevertheless worth emphasizing: our analysis of the grammatical structure of a language must be based on linguistic evidence. In this chapter we will discuss various kinds of evidence which can help us answer basic analytical questions about constituent structure.