This page concentrates on syntactic tests, involving possible strings of words or subtrees. The handout of February 9 (downloadable from here) provides many morphological tests like 'Vs inflect for tense and aspect'. These are much more straightforward to apply and understand, because they just make generalizations about the kinds of words we find in the lexicon.

	Fido is a crazy dog.	N: dog	A: crazy
	That is a strange idea.	N: idea	A: strange
*	The sweater seems strange red.	A: red	A: strange	cf. The sweater seems strangely red.
*	The party ended strange quickly.	Adv: quickly	A: strange	cf. The party ended strangely quickly.

Hypothesis If X is an N, then X cannot have an NP right sister. That is, structures like the following are ungrammatical:

*	Fido is a dog the terrier.	N: dog	NP: the terrier
*	Ed eats rice Japan.	N: rice	NP: Japan
	Ed eats rice from Japan.	N: rice	PP: from Japan
	Ed eats rice that comes from Japan.	N: rice	clause: that comes from Japan

Comment: This test doesn't uniquely identify Ns, because other parts of speech also fail to appear in this configuration, most notably As:

*	Ali is happy the news.	A: happy	NP: the news	cf. Ali is happy about the news.
*	Ed is tired the snow.	N: rice	NP: the snow	cf. Ed is tired of the snow.

	There was time enough to discuss the topic.	N: time
	There was enough time to discuss the topic.
	He was happy enough with the idea.	A: happy
*	He was enough happy with the idea.
	Ali runs enough to compete in a marathon.	V: run
*	Ali enough runs to compete in a marathon.

So this seems to be a strong test. However, during class we ran up against the following apparently minimal pair:

(R)	She had read enough to pass the exam.	V: read
	She had enough read to pass the exam.

We did not figure out how the example works.

Comment: This is the hypothesis that Geoffrey K. Pullum operates under in 'The nature of prepositions', the reading for Assignment 4. Pullum reports judgments like the following:

	The child is sitting right under the coffee table.	P: under
*	The child is right upset.	A: upset
*	The child moved right quickly.	Adv: quickly
*	Ed right cried.	V: cried

However, we learned in class that some people accept examples like The child moved right quickly, and it is unclear whether we can take advantage of the fact that right seems to mean very in such examples, since this might in fact be just like the meaning of right in The child is sitting right under the coffee table.

And we noted also that some examples of right-modified Ns and As seem to be grammatical (these examples were brought to my attention by Geoff Pullum):

	I hung the picture right even with the window.	A: even
	Charlie is playing right outside.	N: outside (right?)

Thus, we were led to explore the following alternative:

This seems to be correct for a huge range of Ps. However, I suspect that we won't ever find grammatical right modification involving the prepositions of or throughout (for example):

*	I am tired right of the picture.	P: of
*	Charlie fell asleep right throughout the day.	P: throughout

Thus, it is doubtful that the right test can help us identify Ps.

Conclusion: It looks like we will have to rely on the other tests we have for prepositions, which are provided on the handout dated February 9 (downloadable from here). Do they hold up?

	We might see a movie.	V: see
*	We might moviegoers.	N(P): moviegoers
*	The baby might sick.	A(P): sick
*	The baby might on the couch.	P: on the couch

This seems to be working well. But, as is often the case, adverbs potentially cause us headaches:

We might quickly run some tests.

Adv: quickly

But three considerations suggest that we shouldn't view this as overly damaging.

First, the adverb is always deletable in these examples, whereas the V that follows it is not unless the context supports ellipsis:

	We might run some tests.
*	We might quickly.

Second, the adverb does not allow us to get away with non-Vs after it:

The baby might possibly sick.

Third, some adverb can appear before the modal or way after it, suggesting that it is fundamentally different from the Vs that follow modals:

	We often should run tests.
	Often, we should run tests.
	We should run tests often.

Thus, I conclude that we should test only after removing all optional adverbial elements:

Comment: This test actually tests what kind of phrase can appear as the complement to these verbs. The claim is that all and only the APs are allowed in this position. So we are really claiming that the only permitted value of X in the following is A:

We can think of this as a test for what is and is not an A because the A projects to the phrasal level --- this sort of projection is a topic we will soon address in detail.

Comment: We must stress 'and' in the example. It does not suffice to say that X is an A if and only if X can appear as the complement to one of the three verbs listed. X must be able to appear with all three.

Comment: We must be careful that we consider only the meaning of appear on which it means roughly 'looks like'. The meaning on which it means 'come into existence' is different. One piece of evidence for this distinction is that the 'come into existence' reading doesn't allow any complement at all: The train appeared (on the horizon) vs. The train appeared the passengers (on the horizon).

	The child became sick.	A(P): sick
	The child seems sick.
	The child appears sick.
	The child became the sick member of the family.	NP: the sick member of the family
*	The child seems the sick member of the family.
*	The child appears the sick member of the family.
*	The child became sleep(s)/sleeping.	Vs: sleep, sleeps, sleeping
*	The child seems sleep(s)/sleeping.
*	The child appears sleep(s)/sleeping.
*	The child became weepily.	Adv(P): weepily
*	The child seems weepily.
*	The child appears weepily.
*	The child became on the swings	PP: on the swings
*	The child seems on the swings.
*	The child appears on the swings.

Difficulty: But the generalization has a problem with PPs. What we find is that idiomatic PPs can appear as the complement to these verbs. For example:

	The child became under the weather.	Idiomatic PP: under the weather: sick
	The child seems under the weather.
	The child appears under the weather.
?	The linguist became on the ball.	Idiomatic PP: on the ball: attentive, doing well
	The linguist seems on the ball.
	The linguist appears on the ball.

Thus, we must revise the generalization somehow. Here are two options:

The first revision appears to be way too strong; we saw that on the swings doesn't pass this test. Moreover, it seems that if a PP does appear as the complement to one of these verbs, it must have its idiomatic sense. This is telling us something important. We should favor Revised hypothesis 2.

Determiners form a relatively small class of elements that, in an intuitive sense, complete our Ns by turning them into phrasal constituents (NPs).

Comment: We must restrict this with 'singular' because many plural Ns do not require a determiner element:

	Pigs fly
	Sheep frolic
	Buffalo buffalo buffalo!

Comment: We must restrict this with 'count' because mass nouns typically do not require, or do not allow, a determiner element:

	Rice is tasty.
	Phonology is fun.
	Totalitarianism scares me.

Comment: We must restrict this to looking at subjects (left daughter NPs of the root) because English has many (quasi)idiomatic phrases that permit bare singular count Ns.

	Columbus sailed from sea to shining sea.
*	(Shining) Sea fascinate me.

Of all the categories in English, adverbs are the freest in their distribution. Sometimes it seems that they can appear anywhere in a clause.

The following structures (though of course negotiable) serve to illustrate how the above generalization might work out structurally:

The above hypothesis also rightly predicts that we can't find adverbs between the verb and its complement, or inside an NP (to name two prominent examples).

*	Ed won obviously the race.
*	Ed won the obviously race.

The above hypothesis does not claim that adverbs must appear adjoined to the root node. Any such claim would be mistaken; it seems certain that they can also appear as VP modifiers, for example. What the above does claim is that we can identify adverbs by checking to see whether or not they have the freedom suggested by the hypothesis.