I am out of my depths but I think it’s something like this.
I don’t know specific examples of flat VP structure grammars but I think dependency grammars might be considered “flatter” than X-bar theory. I don’t really know if there are people who still advocate dependency grammars.
I think X-bar theory has theoretical appeal partially with recourse to the classic idea of Occam’s razor, that it has theoretical simplicity and can account for more data with simpler structures, because in X-bar theory, all syntactic constituents have the same structure: a head and a complement (I think). There are just different types (NP, VP, etc.), but of the same template. I assume Chomsky claimed there was some cognitive-biological reason for this simple structure being at the foundation of language, but I don’t know.
Hierarchical syntactic treatments of verbs, where for example the VP is something like a verb and its complement, the complement respectively being a noun phrase, attempt to give a more general, cross-linguistic account of the syntactic nature of verbs, not restricted to the particularities of any one language, as I think a dependency grammar would - i.e., stuff as basic as SVO order, or subject-auxiliary inversion in English questions, or wh-fronting in English wh-questions - by making those “movement” rules on a higher level of the linguistic system (some of them phonological movement - they move position but still have a relationship, syntactically, to their heads or complements) - you can try to explain a widely present or universal structure that all human language has, and explain away differences in actually observed grammatical structures by movement on those higher levels. If you treat verb phrases more as “what you see is what you get” - that the surface level order is the type of grammar rules defining or dictating the language - you have a bunch of different rules for every language in the world, so more limited generalizable explainability (I think).
That’s the best of my current knowledge.
This looks like a great resource (partially) on this, I’d appreciate anyone filling me in on what I don’t know much about dependency grammars and their modern usage.
While this is not pedagogically helpful, I believe this excerpt from Chomsky’s Minimalist Program is discussing this idea: of exploring rules with higher levels of generality, but there sometimes being problems in doing so:
As for UG, the earliest versions assumed that it provided a format for
rule systems and an evaluation metric that assigned a “value” to each
generative procedure of the proper format. The crucial empirical
condition on UG, then, is that the system provide only a few
high-valued I-languages consistent with the kinds of data available to
the child, perhaps only one. If UG is feasible in this sense, the
fundamental problem (2b) can be addressed (Chomsky 1965). This
approach recorded many achievements, but faced a fundamental and
recurrent problem: the tension between descriptive and explanatory
adequacy. To achieve descriptive adequacy, it seemed necessary to
enrich the format of permissible systems, but in doing so we lose the
property of feasibility, so that problem (2b) is still unresolved. The
conflict arises as soon as we move from the intuitive hints and
examples of traditional grammar to explicit generative procedures. It
was quickly recognized that the problem is inherent in the kinds of
rule systems that were being considered. The most plausible approach
to it is to try to “factor out” overarching principles that govern
rule application generally, assigning them to UG; the actual rules of
grammar can then be given in the simplest form, with these principles
ensuring that they will operate in such a way as to yield the observed
phenomena in their full complexity (Chomsky 1964, Ross 1967). The
limit that might be reached is that rules are eliminated entirely, the
“apparent rules” being deduced from general principles of UG, in the
sense that the interaction of the principles would yield the phe-
nomena that the rules had been constructed to describe. To the extent
that this result can be achieved, the rules postulated for particular
languages will be shown to be epiphenomena.
(Chapter 1, p. 21)
This says in simpler language:
The original theories suggested that there was a universal grammar (UG) - a set of rules that could be applied to all languages.
The goal was to have a few, high-quality interpretations of language that could explain the data a child is exposed to when learning a language.
The problem was to fully describe a language (“descriptive adequacy”), we needed to add more rules, but this made grammars more complicated and even less “feasible”, or likely to be actual models of nature and the world (maybe) - “explanatory adequacy”.
The solution was to try and identify overarching principles that govern all languages.
These principles would be part of the UG, and the specific rules of each language would be simplified, with these principles ensuring they work in a way that explains the complexity of language phenomena.
The ultimate goal would be to eliminate the need for specific rules for each language entirely.
Instead, the interaction of the UG principles would explain the phenomena that the rules were initially created to describe.
This would mean that the specific rules for each language are just side effects of these universal principles.