1) If two words merge into one phonologically, they've already done so syntactically.
Notice the following progression:
- I am going to go home.
- I'm going to go home.
- I'm gonna go home.
- Ommina go home.
It pretty well settles right there. Why don't "ommina" and "go" fuse together in the same way? Because "go" is still a highly variable element. (I could just as easily say "Ommina wait," or "Ommina stay here.") If the next word in a sentence is highly predictable, it just makes sense to fuse it together with the word before it. That way you only have to remember one word with one meaning. That meaning might be a little more complex than either of the two original words, but that just means the new word is more specific.
As you can see, a single word ("ommina") can do the job of four ("I am going to"). Still, there's no empirical justification for splitting such words up into allomorphic forms of the words we imagine to be there down deep:
- ah = I
- m = am
- un = going
- uh = to
It's still possible to split these words up phonetically, but give it a few generations, and pieces will seem to fall out of the equation. That's when a traditional analysis will fall apart. When I was younger, I had a bit of a Texan accent, so my pronunciation of "ommina" was just a little shorter, something like "ommunh" (with the same nasalized vowel you hear in "huh"). There's really nothing you can point to in this version of the word that you could call an allomorph of "to" (except maybe a "null morpheme", but that's just cheating). If you largely ignore such modern simplifications in your analysis of a language, you'll always be fighting the current; language changes as you study it.
OK, so it seems like a pretty simple argument right? Forms such as "ommina" that look like one word really are one word. Plenty of languages have "conjugations" and "inflected forms" that used to be sequences of words, and we don't try to split those up. Just look at the Spanish verb system. Speakers didn't decide on those conjugations because they were logical; they used existing words to communicate what they needed to and wound up sticking later generations with the resulting inflectional mess.
You might be wondering why I'm even arguing this. Who's disagreeing with me? How about Noam Chomsky and his UG minions for starters? If I say "ommina" is one word, they simply can't split the sentence up into a subject and a predicate, which makes the whole thing not a sentence the way they define it. The most popular theory these days would say "ah" is a noun phrase serving as the subject and "munna go home" is a verb phrase serving as the predicate. This makes an extremely artificial division in order to force the data to fit the theory.
2) A word can't merge into a syntactic unit with another non-adjacent word until it's already done so with all the words in between.
This is really just an extension of my first assertion. Two adjacent words will necessarily hold together into more of a phonological unit than two non-adjacent words. This is pretty intuitive, I think.
I have a two-and-a-half-year-old daughter who's constantly making me really proud with the kinds of sentences she's able to produce. (As an exercise in futility, try counting the number of words in the preceding sentence.) Yesterday, I was washing dishes, and she wanted to play with me, so she somewhat indignantly asked, "What are you washing them for?" Depending on what sort of grammarian you are, you might say her sentence was ungrammatical because of the preposition with no object at the end of the sentence. That's absolute nonsense, and modern syntacticians have nothing to do with such assertions. However, they will insist that the object of "for" in that sentence was "what". It makes sense in a way. You don't have an object after "for" because you've already said "what", but you still don't have an object after "for", do you?
I would say that all this "what" does is signal that something will be missing in the string of words to follow and that that missing something is the focus of the question.
Think about the following sentences:
- What do you think the problem is?
- What do you think I should do?
- What do you think they're eating for lunch?
- What do you think tastes the best?
- What do you think he paid that cop so much money to get his mother out of?
Do the different instances of "what" above really feel like they're functioning in different ways? As you're saying "What do you think...", you know you want to get the listener's take on something, but you might not know what until you get a little further in the sentence. This bring me to my third idea.
3) Sentences are formulated in roughly the same order they're spoken.
This just makes sense. Otherwise, you'd have to think of your whole sentence before you even open your mouth. This may not seem too controversial, but it does highlight the ridiculousness of Transformational Grammar. How could I be moving the "what" from its natural position at the end of the sentence if I haven't even thought of the end of the sentence yet? And in Minimalism, you actually tend to build English sentences starting with the final word and ending with the first, so that's no improvement.
So keeping true to these three cardinal rules, how can I explain the heirarchical nature of the sentence and all its intricate connections? I believe in constructions (signs with variable parts), so that helps. Also, I've been toying with two ideas today: supply and demand. These ideas aren't too far off from the posited features and feature-checking of Chomsky's Minimalist Program, but without all the baggage of Universal Grammar. I'll post more when I get this worked out a little better!