On the topic of constituency tests, there are a few standard tests out there, and in this age of the ubiquitous internet, I’m sure some of you will head out into the world to see what other people have said about them.

This may be the first point, certainly not the last, where the internet might provide some information that is not entirely consistent with what you’ve gotten in class. There are a couple of reasons for this. A big source of difference comes from the fact that we’re formalizing a scientific model of syntactic knowledge, and there are different views about both the content of the model and the right way to formalize it. LX522 is not aimed to give you a comparative overview of syntactic theories, because we have enough to do to just get a handle on one. The specific one we’re working with is a generative syntax of the “minimalist” flavor, in its essentials the system outlined in chapter 4 of The Minimalist Program book (Noam Chomsky, MIT Press, 1995). This differs in certain ways from, e.g., the interpretation of minimalist syntax in Andrew Radford’s textbook Minimalist Syntax, and from advanced papers in the field, where the model, terminology, and notational conventions continue to evolve.

So while I don’t mean to discourage anyone from seeking other points of view on syntax, or just finding other helpful notes out there on the internet, do keep in mind as you look that for this course the textbook, the handouts, and the class discussions are the definitive source for how we’re approaching things. And on homework, an answer that might be right under a different approach to syntax doesn’t necessarily qualify as right on the homework.

Back on point, consider the Wikipedia entry “Constituent (linguistics)“, which outlines the idea of constituents and has some short discussions on constituency tests. There, you’ll find a few tests other than the ones we talked about in class. It’s not specifically useful for the homework, since I have prescribed the tests you are to use, but if you’re at a party trying to determine if something is a constituent and have understood the idea of how those other tests work, then these make good additional resources. For example, the coordination test, or the passivization test.

You’ll also notice that in their discussion of the replacement test, it says that substitution normally involves replacement by a pronoun in place of a phrase or clause. This isn’t quite how we talked about it class, though it is basically the same idea. And the test is not always reliable, because you can manipulate things in such a way as to made a non-constituent replaceable, even with a pronoun (for an example, see below). First of all, the class of things that count as legitimate replacements that we discussed in class (and that are discussed in the textbook) is broader than just pronouns. This has advantages and disadvantages. By limiting valid replacements to pronouns, you lose the ability to test verb phrases, for one thing. Although “eat the sandwich” is a constituent in “Pat will eat the sandwich” there is no pronoun you can put after “Pat will” that will lead to a grammatical sentence. On the other hand, opening it up to any word increases the risk of false positives: “has class on” is not a constituent in “Pat has class on Tuesdays” although you can replace it with “despises” (“Pat despises Tuesdays”) and get a grammatical sentence. The risk of a false positive is reduced by trying as much as possible to keep the meaning the same and by using single-word replacements, but there is still an element of subjectivity in the decision of whether to trust the test you construct.

Here is an example of a false positive using the replacement test: Is “the ice cream has” a constituent in “The ice cream has melted”? Well, you can use “it” as a replacement and get “It melted”. That’s grammatical, and even feels like it might have the same meaning, but it’s a false positive: “The ice cream has” is not a constituent. Try a different test, like clefting: “*It is the ice cream has that melted.” No good. The fragment test? Well, that could be considered to give a false positive too (“What melted?” answer: “The ice cream has”) if you aren’t very careful about observing that the tense/aspect doesn’t match between the question and answer. Notice too that the false positive with the replacement test came about by using a pronoun, so even under the more constrained version of the test outlined on the Wikipedia page, you can still get false positives.