Can you explain a little more about exactly what you're trying to figure out?

Do you understand feature-checking in general? Do you just want to know differences between, for example, GB and MP?

Suggested reading: maybe "Distributed Morphology and the Syntax-Morphology Interface" by Embick & Noyer in The Oxford handbook of linguistic interfaces (2007). It isn't strictly about feature-checking (in fact, it never uses that term) but it is about Distributed Morphology and Minimalist syntax, with discussion of how features are implemented there, in a way that wasn't true in earlier theories. I just mention this article because it happens to be relevant to my research. There are probably better articles out there. You could, of course, start with a textbook on the MP.


Or do you want to know what the idea of feature-checking is in general? It isn't, conceptually, all that different in the MP from other uses.

Short answer: feature-checking involves a computational mechanism that checks whether two features match. So I might have a predicate looking for an object with a particular feature, such as [±finite]. The verb "want" in English expects a non-finite complement, so it would accept the phrase "to eat" but not "that I eat". Feature-checking would allow the generation of "want to eat" (because everything matches) but reject "want that I eat" (because the features don't match). The exact implementation varies by theory, but the idea is the same. We might say, for example, features can only be checked locally in some particular domain.


(Of course, remembering the MP is just "how can we make our theory simpler?", rather than any specific theory in itself. Each author has a slightly different approach, though there are some commonalities. Be aware of that, because it makes everything confusing to keep up with the latest "advances in the MP".)

p.s. would a grammar then be necessary if we all spoke telepathically?

Sure. If you have two robots that can communicate wirelessly and send information directly to each other, there must be some format to that information. That format can be considered a grammar. There would likely be different properties of that grammar, but you'd still need structure.
However, you might mean do we need "linguistic grammar" and perhaps not. In that case, we might just need thought, which has its own structure. (Although some would say thought and language are the same at some level, but that's controversial!)

To start with, I guess, could you please explain the interrelations between a grammar, its PF representations, and its LF representations? It seems to me that LF representations are the "ends" of all language. We use grammar and PF "means" to reach that end. So LF representations depend on PF representations which depend on a grammar. Is that basically right?

What you're referring to has been developed over a long time with slightly varying terminology and smaller or larger changes along with changes in theory. At least one name for it is the "T-model". Just searching for that I found this paper which seems useful:
http://cogprints.org/5432/
It's presenting something else as an argument, but the background information on the T-model is useful. It's a bit older, though, so it's not exactly what is current in the MP now. But the MP is built on all of that, so unfortunately to get what's going on in the MP you need to understand the history. (Again, the MP is an attempt to streamline the older theories, so you have to start with them.)

Personally I've discussed and learned more about that level of syntax from the perspective of [Formal] Semantics (dealing with LF). So that's another area to explore for information.

I do think that your average textbook introducing the MP will explain at least the basics of this model though.

Note that one major complication is whether there's order to it or not. Are there merely endpoints, or is the lexicon "first" then PF and LF "last"? Taking "generative" literally, it seems to be directional. But often that's just a metaphor. Read specific authors carefully to see.

Also, I'm not exactly sure what the principle of full interpretation or what converging at LF or PF means.

http://www.glottopedia.org/index.php/Full_Interpretation
As I understand it, anything that ends up in PF must be represented in the pronunciation, and anything at LF must be included in the meaning. PF is the interface with pronunciation (or articulation more generally) and LF is the interface with semantics.

Also, could you please give one example of a "crash"? I appreciate your help!

Crashes occur when the derivation can't continue or when something goes wrong. There's nothing wrong with lining up words in a certain way (or in MP terms "merging them") until you find that features don't check or there's a violation of some principle. An example depends on having a specific theory in mind, but the idea is straightforward. Let's say that an element requires an accusative complement (so the verb "eat" wants some object in the accusative). If its (intended) object is in the nominative, the derivation will crash. One way to explain that would be to say that features can't be checked. Another would be to make some sort of principle that objects must be in the accusative.



As I said, unfortunately to really understand the MP you have to understand the history. Almost nothing of what you've asked about is unique in the MP, and in fact it operates exactly the same way (but in a somewhat different context) in earlier approaches like GB.

Sure. Why wouldn't it be? But usually it's the head rather than the full phrase, to be precise.

That depends on a number of factors we haven't worked out in detail.

But why not? "The" isn't marked as plural in English, but there's no reason it couldn't carry that feature. In Spanish, for example, "los" (the.MASC.PL) would be marked and could be "responsible for" feature checking for verbal agreement.

DPs in general would have some trouble because it seems like the features are coming from the head noun. I imagine the general way to deal with this is to assume that the D inherits some features from the N.

It's all part of the system. So if you assume hierarchical structure where D is a head above N, then there must be some way to get that information out. Another would be long-distance agreement. But given that sometimes Ds agree to Ns, why not?

I don't have a particular source in mind, but this sort of thing (and/or similar things) would be discussed in an intro textbook to the MP.

Just the first thing that came up on a search, this is about T and C sharing/inheriting features:
http://muse.jhu.edu/journals/lin/summary/v038/38.3richards.html
https://www.google.com/search?q=dp+feature+inheritence+minimalism


(The difficulty with the MP is that nothing is an easy paper to read to get started-- it's all based on lots of previous literature. And there isn't one right or even standard answer. Further, almost nothing is based on direct evidence; it's all built up on lots of assumptions, many from the previous literature, and then, within those, the argument ends up being in some sense straightforward, but to see that you have to be working within the MP rather than just looking at a sentence and asking how the D agrees with the V, for example. [That's what I mean by "It's all part of the system."-- making a number of assumptions, what I said does follow. But not everyone would make those assumptions, and so forth.] For those reasons, I can't give an exact citation for these answers, and also because I don't claim to be an expert on the MP, just that I'm familiar with some of the work in it.)