The other day in one of the classes I took at university, we were given a task to draw a syntactic tree of a sentence that contains “take note” as a verb. Since I forgot the sentence, let’s say it is as follows:
He takes note of what she says.
It is quite challenging, using certain theories of syntax, in this case generative syntax, to correctly analyze the idiomatic “take note.” One might try to analyze take as V and note of what she says as a NP (or, in class, DP) complement. This analysis is of course flawed and should be ruled out, because note of what she says is not so obviously a constituent.
You can try substitution with a pro-form…
*He takes [ note of what she says ]i, and Dan takes iti too.
… Or fronting (so-called “topicalization”)
*Note of what she says, he takes.
… Or using it as an answer
What does he take?
*Note of what she says
And so on.
So instead I went with the analysis that take note behaves like a compound, and labelled it V, taking a PP complement, essentially shifting the responsibility to the lexicon. It wasn’t perfect, but it was the best I could think of at the moment with the framework required.
But a week later this (already somewhat ugly) analysis also turned out to be wrong; we were presented with data that showed take note cannot be together because it is possible to insert careful in between, hence take careful note.
The lecturer discussed the case and attributed the careful case to sociolinguistic variation, and we ended up having a non-unified analysis where note of what she says is analyzed as a constituent when there is careful, and my analysis applies when there isn’t. This is far from satisfying if not worse than my previous analysis. I also find that although more frequent on British National Corpus, take careful note also occurs in COCA.
My afterthought is that Construction Grammar is a perfect solution to cases like these. Duh, the theory itself was created to accommodate these “marginal” idiomatic cases with varying degrees of schematicity. In this case one only needs to propose a construction along the line of [ take (careful) note of NP ] and everything is solved.
Once one allows for a primitive unit to be complex when truly necessary, all the headache that comes with atomic syntax is gone.