I’m going to break from my planned schedule and drop into the topic that really interests me: Syntax. Before I begin, I just wanted a few notational issues out of the way.
Something marked * is ungrammatical.
Something with a ? is questionable, or acceptable only within some dialects.
Words marked within parentheses are optional.
The goal of syntax is to create a series of rules, which, together with the addition of our vocabulary will generate all grammatical sentences, while not overgenerating (generating ungrammatical sentences). This descriptively adequate grammar should have 3 qualities.
1) It will describe the linguistic data by general principles. For example, if I went through every spoken English sentence to this point, and classified them as either grammatical or ungrammatical, based on their acceptability, and then took that data and made a function (one that took as its inputs sentences, and as its outputs either ungrammatical or grammatical), it would not be a descriptively adequate grammar. Why? Because while it would be correct, it would have no principles. The second a new sentence was uttered, it would lose its description. Morever, it has no predictive power over unspoken sentences. A scientific theory of grammar, since it would have reasons for grammaticality or ungrammaticality, principles, would have predictions about sentences. One can then continuously tweak such a grammar to deal with predictions that run counter to the theory.
2) It will be economical. Many things will explain the data. But a theory which fits the data with fewer proposed elements is better. Why? Because elements that are individual are less predictive, as shown above.
3) It will be psychologically possible. Some models which are perfectly good at explaining the data might not fit with psychological truths. A good descriptive grammar should in some way stand for the neurological facts which occur. This does not mean that a speaker must have conscious knowledge of the proposed principles, only that the brain processes should reflect our principles. This third one will be harder for me to test, but where possible, I will try and provide such psychological evidence.
But why posit an innate grammar at all? The above psychological problems of language acquisition are highly relevant to a scientific grammar because of the poverty of stimulus, the exact reason for the necessity of innate grammar. Language acquisition is an inductive impossibility. If perfect data were presented to a language learner, the principles could be induced. But the data set available is imperfect. Firstly, the data set is incomplete, and many of the generalizations we make are made without evidence. Generally, ungrammatical sentences go unspoken. But since the data set is finite, many grammatical sentences go unsaid. So a sentence being unused does not suggest that it is ungrammatical. And yet we know which sentences are grammatical or ungrammatical, even should they be sentence types we have never heard. Secondly, speech is full of errors. So somebody learning a language will hear wrong sentences, both on account of errors and on account of the nature of conversational language: full of terms left out, gestures, ummm’s, and changes mid sentences. Despite the flawed, limited data set, we know which sentences are right, and which are wrong. This suggests an innate linguistic ability. However, this linguistic ability is not so precise as to cause the learning of only one language (the variety of language shows that quite clearly). Instead, we say that there a number of parameters that languages have set one way or another. But the innate ability to identify which way those parameters are set is what allows a child to determine what his language is like.
A Descriptive Grammar of English will be derived from Universal Grammar, and yet will have had parameters set. A more expansive enterprise may attempt to look past the parameters, to a pure Universal Grammar, but this account shall stick to English, for the most part.