In response to a question I got over email, let me just give a couple of examples of evaluating Binding Theory satisfaction and violations.

For Binding Theory, the thing to keep in mind is that binding only goes one direction, from higher in the tree to lower in the tree. For the most part, things that are further left in an English sentence are higher in the tree than things that are further right, although it is more complicated when there is a larger constituent on the left (like John’s mother).

Specifically, a node A binds a node B when A c-commands B and they share an index.

So, in something like John_i told Mary that he_i saw himself_i we have three things that share the index i. There are no larger constituents to worry about here. The binding relations are:

• John_i binds he_i.
• John_i binds himself_i.
• he_i binds himself_i.

So, everything binds to the right. (Though, keep in mind, binding really depends on c-command, not just being to the right. See the second example below).

Then, to evaluate Binding Theory, you look at each one in turn.

• John_i is an R-expression, and is not bound by anything.
• Principle C says that R-expressions cannot be bound.
• So, Principle C is satisfied (for John_i).

• he_i is a pronoun, and is bound by John_i.
• Principle B says that pronouns cannot be bound within their binding domain.
• The binding domain for he_i is the clause containing it.
• The clause containing it is (that) he saw himself, and John is not inside.
• So, Principle B is satisfied (for he_i).

• himself_i is an anaphor, and is bound by John_i and he_i.
• Principle A says that anaphors must be bound within their binding domain.
• The binding domain for himself_i is the clause containing it.
• The clause containing it is (that) he saw himself, and he_i (but not John_i) is inside.
• So, Principle A is satisfied (for himself_i) by virtue of being bound by he_i.

A more complicated example would be one like: His_i mother thinks John_i is a genius.

• Here, nothing binds anything else.
• His_i is inside the constituent His_i mother and so it does not c-command John_i.
• John_i is further down the tree and so clearly does not c-command his_i.

• For his_i, Principle B says it must not be bound within its binding domain.
• The whole sentence is the binding domain.
• But nothing binds his_i at all, so it is safe.
• Principle B is satisfied (for his_i).

• For John_i, Principle C says it must not be bound.
• Nothing binds John_i, so it too is safe.
• Principle C is satisfied (for John_i).

And, of course, these results are what we want, since both sentences are grammatical.

One last example, this one is not good: *John_i saw him_i.

• John_i binds him_i.
• John_i must not be bound according to Principle C.
• And it is not, so Principle C is satisfied (for John_i).

• him_i must not be bound within its binding domain, according to Principle B.
• But it is bound, by John_i, and John_i is within the binding domain.
• So, him_i violates Principle B.
• And, indeed, the sentence is ungrammatical.

Just something like a clarification: You can only Adjoin to something that doesn’t have any Merges left it needs to do. Effectively, this means that you can only Adjoin to XPs. You don’t have to necessarily wait until the very end of the derivation to Adjoin — you just have to do all of the Merges first for the projections of any given head. That is, you can Merge the V and its object, forming a VP, and if there are no more [uN] features on the V, you can then Adjoin something to the VP before continuing on to Merge the v and VP (to satisfy the Hierarchy of Projections).

What you can’t do is, say, Merge “to Mary” and “introduce” (which will check off one of two [uN] features it has), then Adjoin something to the result, and then Merge “John” to the result to check off the other [uN] feature. You have to check both of those [uN] first with Merge, before you Adjoin.

Another quick note based on a question I got. Although it is possible to say “John read” (and one can debate whether this implicitly means “John read something” or not), for the purposes of the problem on homework 4, assume that “read” is a regular transitive verb. With an Agent and a Theme. Like “see” or “kick”, etc.

Just to be clear about the expectations in homework 4: Don’t draw any T nodes. We hadn’t covered T at the point when homework 4 was assigned, and we aren’t quite finished talking about it. You’ll get a chance to draw plenty of T nodes in the future.

(I had been relatively explicit about that for “give” but I intended the same thing for “read.”)

In part II of homework #3, just to clarify, the sentence you are trying to construct does contain more than just the words called and for. There are five lexical items (five words). Three of those words are pronouns (which means I perhaps should have included something like a [PRN] feature on them as well). But the sentence contains three pronouns, the verb called and the preposition for.

I didn’t give you the pronunciation of the pronouns, but the features that I gave for them will determine the pronunciation of each unambiguously.

In the sentences (4a) and (4b) in Part 1 of homework #3, you’re asked to consider the theta-roles of the participants in a scolding.

As one of you has pointed out to me in an email, there is a sense in which when, say, John scolds Mary, John is doing something, but Mary is kind of feeling something as well. It’s weird to say John scolded the rock. There does seem to be some kind of requirement that the scoldee be a conscious being.

This is an unintentional complexity here, and I think if I ever use this homework problem again, I’ll try to pick a different verb. For the purposes of winding up where the homework is intended to lead you, it would be better to put aside thoughts of the conscious effect of scolding on the target of the scolding. You want to treat the scoldee as being a more generic object, more like the object of hit or saw.

We probably do not want to admit the possibility that two participants in an event can be assigned the same theta-role (even in the situation where we are hypothesizing that the verb assigns two theta-roles, but of the same type). If we did allow for that, then there is not so much of an issue. And, perhaps we do want to have some way to allow for this, abstractly, because if John is the Agent of John ate, then Mary and John are probably both Agents in Mary made John eat. However, there Mary is the Agent of make and John is the Agent of eat, presumably. As we will see relatively soon, the connection between theta-roles and structural position is going to be hypothesized to be quite tight, in a certain abstract sense, and so it would not be compatible with supposing that the same verb can assign two theta-roles of the same type to two different arguments.

In summary, treat the scoldee in (4) more like the story in (2) or the poetry in (1).