Aug 202018
 

A large portion of my interests lately has centered around the proposition that syntactic theory has been overprivileging semantic argumentation. I’ve talked a lot, and in various forums, about why I think this is bad for syntax. Here, I’d like to say a bit about why I think this is bad for semantics.

Let us stipulate that a central goal of any semantic theory is to be compositional. The slogan here is often something along the lines of, “How the meaning of a complex expression is related to the meanings of its parts.” But I think this particular formulation actually obscures a crucial point. For anyone who is not a flat-Earthist about syntax (à la “Sentences are just flat sequences of words!”), a compositional semantics is not about relating the meaning of a complex expression (directly) to the meanings of its atoms. Instead, it’s about relating the meaning of every constituent to the meanings of its immediate daughters.

It follows from this that a compositional semantic theory is only as good as the syntactic theory it is hooked up to. A syntax developed completely in the service of semantics – i.e., using argumentation from meaning as its principle guiding light – risks rendering the entire project of compositional semantics circular. If you have a syntactic structure crafted to capture (mainly) generalizations about meaning, then achieving compositionality in the semantic analysis is really no achievement at all.

Let’s clarify: computing the meaning of an utterance from the meanings of its atoms in one way or another is already no small feat. But such a computation can lay claim to being a compositional model of natural language semantics only to the extent that it composes those atoms in the same manner as the syntax (of natural language) does. Therefore, once we base our theory of the latter – how the atoms compose syntactically – on meaning, we compromise the goal of truly building a compositional natural language semantics.

Fortunately, there is a fairly straightforward way to break out of this circle: base your syntactic theory (primarily) on string-acceptability, not on semantic interpretation. If your semantics is hooked up to a syntactic theory of this kind, then compositionality becomes a meaningful achievement.

Subscribe
Notify of
guest

62 Comments
Inline Feedbacks
View all comments
Gillian Ramchand
Gillian Ramchand
1 year ago

I am completely on the other side in this vendetta against semantics, and I have been meaning to reply to you for a long time in fact. Here´s a quick first go, just to see the fur fly. First of all, in other contexts I often do find myself arguing against formal semanticists and siding with the syntacticians in giving a primacy to the nature of (specific to linguistic systems) syntactic representations, and what evidence… Read more »

Gillian Ramchand
Gillian Ramchand
Reply to  Omer
1 year ago

I quote you: It’s really not hard to show that there’s recursion and hierarchy just based on distributional facts! Of course, some of these touch on meaning – constituency diagnostics require some rudimentary notion of “… and the meaning stayed the same!” Precisely. Syntacticians always underestimate this part. Try to mount convincing evidence for actual hierarchy without using once of these implicit diagnostics. Now, I don’t know what you mean by interpretation or semantics for that… Read more »

Gillian Ramchand
Gillian Ramchand
Reply to  Omer
1 year ago

Thanks for the example. Kratzer is indeed one of the few actual semanticists who are engaged with the facts and results of syntactic theory, and is interested in how syntactic representations put together the kinds of meaning ingredients and dependencies that it does. But forgive me, it sounds to me that what you are describing is a success story rather than something to bemoan: 1. Semanticist makes a proposal because she needs a workable solution… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

The bulk of the Kratzer 2006 manuscript was dedicated to arguing in detail that minimal pronouns did obey the locality constraints usually associated with Agree, even if this appeared not to be so on the face of it. (“The paper challenges the widely accepted belief that the relation between a bound variable pronoun and its antecedent is not necessarily submitted to locality constraints.”) The arguments might be wrong (and really, I don’t think anyone has… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

Again, that’s ignoring the German data, where minimal pronouns had a significantly more restricted distribution. If one takes the German data as Kratzer presents it (I’m not in any position to cast doubt on it), then there’s certainly some prima facie evidence for syntactic locality constraints being involved. You’re also ignoring the fact that Krazter (IIRC) did not think that minimal pronouns were involved in all examples where the phi-features of a bound pronoun appear… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

The only thing I’m advocating for here is not comparing Kratzer to a flat Earther. I don’t know what the correct approach to the relevant pheomena is. The German data are in the Kratzer 2006 manuscript (http://semantics.uchicago.edu/workshops/spl/anaphora/kratzer06.pdf) and probably the subsequent published version too (I haven’t checked). The data show intervention effects of a sort, and are intended to demonstrate that the apparent long-distance dependencies result from a a composition of local Agree operations. I… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

I’m not especially partial to Kratzer’s minimal pronoun analysis. I just think it deserves to be defended against claims of Flat-Earthism. After all, I would be willing to defend even some really bad papers against that charge! The data I’m referring to are the data at center stage in the ms I already linked to. See e.g. (5)-(8) or (22)-(23). I didn’t say that all intervention effects are necessarily syntactic. I don’t have strong views… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

It all depends on the details, as far as I can see. Take, say, “*John, I know a man who likes t”. It’s not especially hard to think of reasons why an Agree dependency between “John” and its trace shouldn’t be able to go via “who”. Especially if the comparable dependencies Kratzer is talking about are possible in part because of certain syncretisms in the morphological paradigms of verbs/pronouns. Should Kratzer have discussed these issues… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

Relative pronouns do enter into dependencies with things outside the relative clause and with things inside the relative clause, to all appearances. I mean, in “the man who John likes”, “who” presumably agrees with something inside the RC, and will also show overt agreement with “man”, in some languages at least. Now, obviously, the analysis of RCs is hugely contentious, and maybe “man” is not really outside the RC, or maybe one of the relevant… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

Whether or not that would be possible would depend on the exact formulation of the CNPC (or whatever other conditions derive it). I’d add that the mechanism by which agreement is established between “who” and “man” in “the man who likes John” may well be a special mechanism tied to relative clauses, and Kratzer would be perfectly free to exploit that mechanism in tandem with Agree.

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

If your definition of the CNPC is that restrictive, then I do not think we have very strong evidence for the existence of the CNPC, so I don’t see the issue. Again, the assumptions Kratzer requires are very minimal. A relative pronoun can agree with the (presumptively external) head of a relative clause. It can also agree with something inside the relative clause. It’s not difficult to find apparent confirmation of both possibilities. What’s then… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

>llusions created by deletion-under-identity,

Yes, sure, maybe. That’s one of many possible ways of dealing with the issue. But whether it’s Agree-ment proper or some other “agreement” operation, Kratzer can still compose it with other operations to form longer step-by-step dependencies.

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

I do not understand your penultimate comment.

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

As far as I can see, it allows one to relate some things inside an RC to some things outside an RC sometimes. The existence of such a mechanism (whether or not it is Agree proper) seems to be required by fairly basic facts about RCs.

Alex Drummond
Alex Drummond
Reply to  Alex Drummond
1 year ago

One tidbit that I didn’t get to fact check yesterdsy. It’s worth noting that Ross’s original formulation of the Complex NP Constraint does not block a relative pronoun agreeing with the external head of an RC, even if one ignores the fact that this is agreement rather than movement: “No element contained in a sentence dominated by a noun phrase with a lexical head noun may be moved out of that noun phrase by a… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

No, I wasn’t suggesting such a theory, that’s why I said “even if one ignores the fact that this is agreement rather than movement”. The point is that Ross only bans movement out of the S within NP if that movement exits the NP. So a dependency between the head of the NP and something within the S would be fine. I am increasingly sympathetic to Kratzer’s not mentioning islands, since the more I think… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

>I thought that this kind of “mediated side-stepping” of the CNPC is a horse we beat good and dead in our earlier comments, no? Not at all. It’s easy to see why you couldn’t side step a Ross-like CNPC using the same mechanism in other cases. Take the example of topicalization in violation of the CNPC. You just need to assume that there’s no intermediate landing site within the NP. It follows then that the… Read more »

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

That is really not the case, Omer, I’ve given a rough formulation of the CNPC (Ross’s, but without the restriction to movement dependencies) which is consistent with the dependencies Kratzer assumes, and which doesn’t obviously allow any unwanted dependencies to sneak in. If you think this is not so, let’s look at a specific example of overgeneration.

Alex Drummond
Alex Drummond
Reply to  Omer
1 year ago

Do you really mean to say that it’s question-begging to assume the absence of an intermediate landing site within NP? I don’t understand which question is supposed to be begged there. Of course, it does not make much difference if there is an intermediate landing site at the edge of NP. If so, you just need the “Complex XP constraint”, where XP is a nominal projection below NP that contains the head of the relative… Read more »

Gillian Ramchand
Gillian Ramchand
Reply to  Omer
1 year ago

Phew!

Marc van Oostendorp
Reply to  Gillian Ramchand
1 year ago

A small point on the necessity of using ‘rudimentary semantics’ in syntactic argumentation. I agree with Omer that this does not imply that semantics and syntax are more intimately connected. It would be almost impossible to do phonology without a similar kind of ‘rudimentary semantics’, as the minimal pairs test is based on it: we know only know that two phonemes are different or that a feature is contrastive because we do a minimal pairs… Read more »

Avery andrews
Avery andrews
1 year ago

Well maybe but I recall Ken Hale remarking that in Warlpiri you can find an interpretation for almost any string. So looking for the best explanation of the form-meaning mapping still look best to me. Concepts such as part of speech and phrase type etc are also traditionally based to a large extent on substitution arguments, which I find do not make much sense if you don’t take semantic composition into account.

Avery Andrews
Avery Andrews
Reply to  Omer
1 year ago

I did say ‘almost’. But I think the question is not what you can use as a method in some particular language, but what the generally applicable idea should be, and it seems to me that meaning-free substitution as often found introductory textbooks perhaps back even to Gleason 1957 (perhaps should reread that carefully) is not sufficient. For example, we can ‘show’ that ‘a thesis on Thursday’ is an NP by observing that you can… Read more »

Gillian Ramchand
Gillian Ramchand
1 year ago

I fear that Omer might call ME a flat earther because I believe very strongly that although (i) I am completely convinced that there are host of patterns and generalizations that we know now about syntax that we did not know before the chomskian turn, and did not even have the wherewithal to ask, I am also convinced (ii) that the precise formal implementation, symbolic devices and modular divisions we are employing now ( in… Read more »

Gillian Ramchand
Gillian Ramchand
Reply to  Omer
1 year ago

Fine. But I am still against the taking a back seat part ..

Callum Hackett
Callum Hackett
1 year ago

I’m very sympathetic towards a greater severance of syntax and semantics, though I come at it from a different angle and don’t believe that string acceptability will work as an alternative metric for grammatical theory, not least because I think that interpretability is a crucial, albeit violable, component of acceptability. I won’t get into further alternatives, but let me say a few things about the circularity. I actually think it ought to be uncontroversial to… Read more »

Peter Svenonius
1 year ago

This discussion is very interesting, but it seems a bit strange to me to rely so heavily on the CNPC, as you do in the back and forth with Alex Drummond. First of all, the CNPC is a description, not an analysis. Maybe if you replace the CNPC description with an analysis, it would become clearer whether movement and agreement should be expected to behave differently across relative clauses. Second, Norwegian and Swedish speakers blithely… Read more »

Peter Svenonius
Reply to  Omer
1 year ago

I think that at this level of debate, the theory-neutral, mid-level generalization is not going to be sufficient to settle the matter, and you have to pull out the theoretical guns, especially for things like to what extent movement and agreement are the same or different. I know the Kush et al. paper, and despite their experiments failing to detect the phenomenon, I still think that Norwegian and Swedish falsify the CNPC as stated by… Read more »

Ethan Poole
1 year ago

Just a small point in relation to the last paragraph of Omer’s original post: The “upwards” Agree (i.e. feature unification) that Kratzer 2009 uses is straight out of Pesetsky & Torrego, which is not mentioned in the previous comments as far as I can see. Perhaps people have taken Kratzer 2009 as a convincing argument for upwards Agree, but Kratzer didn’t come up with the idea. She took it straight out of the syntactic literature,… Read more »

Ethan Poole
Reply to  Omer
1 year ago

Yes, I was responding to your initial criticsm of it, the first point below when you first bring up Kratzer 2009: Kratzer’s 2009 Minimal Pronouns paper convinced a generation of syntax students (in some circles, anyway) that we “knew” that there was an AGREE operation that transmitted phi-feature values downward in the syntactic structure, did not respect basic locality restrictions, etc. etc. Alex already said what I would have said about locality in Kratzer’s proposal… Read more »

Ethan Poole
Reply to  Omer
1 year ago

I think that I’ve made my point, but I just wanted to make it clear that I do not think that just using something already proposed vindicates problems with a paper’s proposal. I’m just defending the methodology: showing how various syntactic proposals might give us a handle on a semantic problem. Of course, those syntactic proposals might be wrong, maybe they aren’t suited for the semantic problem in question, etc., but that’s just part of… Read more »

Itai Bassi
Itai Bassi
1 year ago

(moved to the top level, hope that’s okay with you; O.P.) Hi, and thanks for an interesting thread! just a few (long) comments on my pet topic, the minimal pronoun debate (without entering the broader issue that this post concerns): 1) Kratzer and Locality: I’m not a big advocate of the minimal pronoun approach in general (see below), but I think it’s unfair to accuse Kratzer 2006/9 of not caring about locality issues in her… Read more »

Itai Bassi
Itai Bassi
Reply to  Omer
1 year ago

Re: (1) – Yes, I agree that binder-bindee feature agreement does not obey usual structural constraints on ‘normal’ phi-agreement. Your basic point is that this should be taken as evidence against the idea that binder-bindee agreement is mediated by some syntactic relation (i.e. Kratzer’s rightmost arrow) – and I’m sympathetic. Actually you don’t need the RC construction to argue that – this conclusion is reached also for “Only I brushed my teeth”, as we discussed… Read more »

62
0
Would love your thoughts, please comment.x
()
x