Specializations > Semantics and Pragmatics



I enrolled in a linguistics class this semester. One of the questions asked in the exam was:

1-) John likes dogs and hates cats.
2-) John bought a TV-dinner and ate it.
3-) John likes and Mary hates cats.

Which one of these are true coordination and why?

My thought was second one is true coordination. John bought a TV-dinner and than ate it. It seems that there is a coordination in this sentence.
Because the noun was linked by bought and ate verbs.

I am not a linguist. So any help would be appreciated.

Hi Osman and welcome,

This is a very complicated topic. There is a fundamental assumption in linguistics that "true coordination" exists and that therefore we can separate out the anomalous cases from the "true" cases. It is very difficult to actually figure out what that means because "coordination" is borrowed from traditional grammar (as is "subordination") and it isn't well defined. My research is in this area, and I've come to question the very basic idea that we even CAN define "coordination" in anything like the way that seems intuitive at this point. But I'll attempt to go with what I think your instructor was focusing on, because it is a widespread position.

You're asking about this regarding Semantics, so I will assume this is within the approach that takes "and" to be a logical operator & / AND. It's truth-conditional coordination.
p q p&q

So logical coordination occurs when both p and q must be true for p&q to be true. It also requires nothing else. That's the whole "meaning" of the conjunction.

So let's see:
1. John [p 'likes dogs' & q 'hates cats']
2. John [p 'bought a TV-dinner' & q 'ate it'].
3. [p 'John likes' & q 'Mary hates'] cats.

Easy one first: (3) is incoherent. It doesn't involve the coordination of constituents (obvious parts of the sentence) but clearly some kind of transformation in which "and" is used to shorten parts of the sentence. If we expand it to say "John likes cats and Mary hates cats", then this could be "true coordination", but that's not what it says. This is, simplifying a bit, a trick that English can do to make sentences shorter.
So (3) is non-constituent coordination, which causes complicated problems for compositional semantics. It's not (3). (Unless you expand it. I'm not sure why you can't. But... let's assume you can't.)

Then we have (1) and (2) which appear very similar, right? They are two verb phrases that share a subject. You might think that if one works then the other does too. But there's an important detail here (probably mentioned in your class at some point-- it's a popular topic).

Truth-conditional coordination is symmetrical: p&q = q&p. The order does not (cannot) matter.

1') John hates cats and likes dogs. SAME MEANING
2') John ate it and bought a TV-dinner. DIFFERENT MEANING
(Or, if you prefer: John ate a TV-dinner and bought it.)

The "and then" meaning you picked up on is more than truth-conditional &. It's ordered in the order of events. So p&q then isn't q&p, even though technically those much be equivalent. Therefore, the "&" analysis (truth-conditional operator) is not sufficient.

The standard analysis for the "and then" readings is that it is based on pragmatic implicature. It's just normal "and" like in (1), but it also has an implicature that the order is the way in which it happened. We can test this:
John ate a TV-dinner and also bought it, but not necessarily in that order.
But by default it feels like assymetric (ordered) temporal coordination, rather than just logical coordination. Therefore (2) is not "true coordination".

To be honest, these are very specific points that were probably discussed in class but don't seem to have much of a basis in the real world. Your instructor seems to want language to line up on the surface with logic, and that's just not how language actually works. But if it did, then (1) would be the right answer. This is a "because I said so" kind of answer in my opinion. Both (2) and (3) could also be analyzed as "true coordination" with the right assumptions (about the relationship between form and meaning) but not at the surface level.

Can you define Tallerman's constituency test? I'm not familiar with anything by that name.

There is a general constituency test using coordination, often called something like "the coordination [constituency] test".

That's quite simple: if you can coordinate X and Y, then X and Y are constituents. So (1), (2) and (3) all pass it. However, that test is well known to be unreliable and sometimes give false positives. (With constituency test it is always recommended that you use several and look for confirming results!)
So while (3) passes that test, it isn't actually constituents based on results from other tests.
(Again, make sure we're talking about the same test here!)

My favorite constituency test is the "book title" or "newspaper headline" test: would it make a good title?
"likes dogs" -- not amazing, but sure, it could work.
"John likes" / "Mary hates" -- really odd, keeping the same meaning as before (transitive).


[0] Message Index

Go to full version