Norbert Hornstein gives a very good overview of the perceived successes of the Minimalist Program here. (WARNING: this post won’t make a whole lot of sense unless you have a background in Generative Syntax.) One of the complaints you hear a lot about Minimalism – especially, for some reason, from the people who lament the move away from GB1 – is that it isn’t clear to them why things should work this way. Hornstein does an especially good job of addressing that question. By showing how a particularly simple version of the central structure-building operation (Merge) implies a lot of the "mysterious" properties of Minimalism, he shows that none of these theoretical commitments are arbitrary. The icing on the cake, of course, is that the centrality of something like Merge to syntax is not really controversial. Any theory of syntax needs a way to assemble parts into wholes. The beauty of Minimalim is that if you do the Cartesian exercise of stripping everything away from that operation that isn’t strictly contained in its concept, you really do end up seeing things more clearly.
I had two bones to pick, however.
Bone the first:
Before pressing on, a comment: unifying movement and phrase building is an MP innovation. Earlier theories of grammar (and early minimalist theories) treated phrasal dependencies and movement dependencies as the products of entirely different kinds of rules (e.g. phrase structure rules vs transformations/Merge vs Copy+Merge).
Only if by "earlier theories of grammar" you mean GB specifically, and its "Extended Standard Theory" predecessor. HPSG and Dependency Grammar both "unified movement and phrase building" long before there was Minimalism. (Chomsky would never admit it publicly, but I’ve always had a suspicion that HPSG strongly influenced the Minimalist Program.) Now, it’s true that HPSG partisans would not put it that way. Their marketing pitch is that they’ve actually eliminated movement: HPSG is meant to be an entirely representational theory, not a transformational one at all. So, Hornstein can get out of this on the technicality that what HPSG/GPSG is modeling with slash features isn’t movement at all. But that would be a misleading position to take, and in any case it’s at odds with the parenthetical admission in the first paragraph that "…when pressed, I have been known to complain that H/GPSG, LFG, RG, are all ‘dialects’ of GB." "Movement/displacement" is just one (metaphorical) way of understanding long-distance dependencies. HPSG uses a different metaphor, but it’s capturing the same phenomenon. And, crucially, it does it in the "Minimalist" way – by unifying it with the structure-building operation.
OK, but what about the fact that HPSG is "representational" and not "transformational?" That’s a better toehold to use in separating the two approaches – because again, HPSG likes to model grammatical objects as though there were no process involved in forming them. But that’s an aesthetic choice, not a substantive one. Any realistic look at HPSG can see that it has combinational rules, and that the way in which things are combined has consequences. So, there again, is a difference without a distinction. However "representational" HPSG is, it still builds structures.
And, crucially, it forms structures in a way that unifies long-distance dependencies and phrase building. And it (that is, its GPSG predecessor) was doing this ove a decade before Minimalism jumped on the bandwagon.
Bone the second:
In explaining movement through "internal merge" – that is, in explaining how what we used to think about as movement is actually "re-merge" of an item that has previously been merged, Hornstein writes the following:
Thus, whatever is a constituent in the input appears as a constituent with the same properties in the output. This implies (i) that all I-merge is to a c-commanding position, (ii) that lowering rules cannot exist, and (iii) that derivations are strictly cyclic.
Really, it only implies two of the three – (ii) and (iii). (And actually, there’s a way to get (ii) to work with these assumptions, but that’s a long digression.) (i) doesn’t follow in the spirit its intended, since you could in theory merge a part of one substructure into another substructure, and then merge these two substructures together into a unified structure. This is sometimes called "interarboreal movement" (or, as Hornstein’s own doctoral student prefers to style it using a term Hornstein probably coined: sideward movement). If you never get around to re-merging the shared item at the root of the newly-formed superstructure, you would have movement without c-command. Most sideward movement accounts rely on a stipulation that any interarboreal merge must be later harmonized by re-merging the shared item at the root of the superstructure at some point before the derivation ends, and they’ve had a terrible time deriving this stipulation from any foundational principles. And yet, sideward movement effects exist, and whether you model them with sideward movement or some other mechanism, they’re something that doesn’t fall out of simple merge. To take the standard example:
Which papers did John file [which papers] [without reading [which papers]]
"file [which papers]" and "without reading [which papers]" are both standalone phrases that have to be formed on the way to forming "Which papers did John file without reading?" Which is to say, they have to share a phrase before the structure in which the shared phrase ("which papers") c-commands both of its "traces" is formed. Whatever mechanism you use for that – whether it’s lowering or reconstruction or sideward movement – it’s an operation that doesn’t fall out of simple merge. Which means, in turn, that the requirement that a moved item c-command its trace probably doesn’t either.
Now, the response here, and it’s a fair one, would be that Hornstein was careful to say "that all I-merge is to a c-commanding position." Technically, I suppose that’s true. But that seems to be on the way to claiming that all movement is to a c-commanding position, since it was implied earlier that all movement is I-merge, and that bit is not true.