First, let me get the usual caveats out of the way: MP is a program, not a theory. It tells you what kinds of questions to ask about syntax, and guides you in comparing the answers from competing theories. And so on. But in practice, what you're learning is the theoretical framework that everyone in MP uses. You may later learn why they think MP leads to that framework, but not until later.
I'm learning about minimalism at the moment. I'm not sure if I understand the move operation.
I think I understand that in English the move operation takes place when you want to formulate a question. E.g. She has gone. --> Has she gone?
That's how 1960s transformational grammar worked: you started with a "deep structure", and applied "transformations" for things like questioning, passive, etc. until you got to "surface structure". For simple sentences, the deep structure is the same as the surface structure, so you don't need to move anything.
But MP doesn't start with a deep tree and transform it into a surface tree; it starts with an empty tree, and builds things up node by node out of the Merge and Move operations. Simple sentences get built the same way as questions, passives, etc.
Could a tree be built entirely with Merge, with no Move? Technically, yes—but if the sentence has any words which require "feature checking", no, and all sentences do.
"Feature checking" is how, for example, a noun gets its Case—how it goes from "he-or-him (missing case, so I can't speak it)", to "he (nominative case, so I can)". If you can't speak that "he-or-him", you can't speak the sentence. And in MP, feature checking only happens when you move a node to a position that can check it, so that "he-or-him" is going to have to move.
Something to do with causative verbs.
Causative verbs are part of the argument for why, even without any of the theoretical assumptions from MP, we need VP shells and abstract verb movement. If you want to explain why an earlier theory like GB also always had movement even when it's not obvious, causatives are an important example. But in MP, there's always movement, so the whole bit about causative verbs is just an extra confusion.
Let's take the simple sentence "Mary left", and let's simplify the hell out of the theory, because it's still pretty complicated.
You don't start with a deep structure tree, you start with a bag of words and an empty tree. At each point, the only things you can do are:
- Pull a word out of the bag (which makes a trivial tree out of it).
- Merge two trees together, making one of the head.
- Move a node from one part of the tree to another. You're only allowed to do this to check features.
- If there are no words left in the bag and no overt features left to check, "spell out" the tree to phonology.
- If there are no covert features left either, "interpret" the tree to semantics.
- Put [Mary +needscase], [leave +needstense +hasnomcase], [+past +hastense] in a bag. (This is a nonstandard notation that I invented on the fly. I'm using a much simpler set of features than real MP, and I think using the standard notation could be misleading here. But the idea should be clear: we have, e.g., a word with the meaning "Mary" and the phonology /meiri/ and the feature "this word needs to get case before it can be pronounced".)
- Make [+past +hastense] the root of the tree: [+past]T.
- Merge that tree with [leave +needstense +hasnomcase], with the T as head: [T' [+past hastense]T [leave +needstense +hasnomcase]V].
- Move the "leave" from the complement to the head. The +hastense and +needstense cancel out, so we end up with just [T' [leave+past +hasnomcase]T 0V].
- Merge the "Mary" with the T', leaving the T' as the head, so now you have [TP [Mary +needscase]N [T' [leave+past +hasnomcase]T 0V]].
- We can't move [Mary] to [leave+past], because "Mary" and "leave" are both actual words, so they have incompatible features; you'd end up with some monstrosity like "gMaryo". Fortunately, a T can case-mark the specifier in a TP. Which means we screwed up; we should have merged Mary into the VP specifier before moving anything! So, let's backtrack and do that.1 Now, after moving the subject from the VP spec to the TP spec, we've got [TP [Mary +nom]N [T' [leave+past]T [VP 0N 0V]].
- There are no more overt features that need to be checked—all those +needstense and +hasnomcase and so on have cancelled out. So, we can pass the tree to spell-out. English morphophonology turns [Mary+nom] into /meiri/, [leave+past] into /left/.
- There are no covert features that need to be checked either. So, we can immediately pass it to interpret too, and get something like PAST(LEAVE(MARY)).
Notice that I had to insert [leave] as a complement of the T, then move it from V to T, and I had to insert [Mary] as a specifier of the V, then move it to the specifier of the T. That's the absolute minimum amount of movement in the simplest possible MP theory. A realistic MP theory will have even more structure and more movement (and Case is more complicated…), but this is enough to show why there's always movement, never just merging.2
1. This is where actual Minimalism comes in: It tells us that any properly minimalist theory wouldn't have done the last few steps and backtracked; it would instead always stick the subject in the VP-spec position before moving the V.
2. This is why they often teach people GB first: Maybe it was conceptually more complicated, but the practice was a lot simpler.