Skip to main content.
July 27th, 2003

Ambiguity Confusions

I was doing some contextualist bashing over on Crooked Timber when I noticed that I hadn’t realised before. The explanation for this is probably completely obvious, and it’s just revealing my own ignorance that I don’t know the explanation, but I couldn’t figure out why (1) is three-way ambiguous, when by rights it should be four-way ambiguous.

(1) Bill said that he loves his mother. So did Tom.


The three readings I think the second sentence has are

(1a) Tom said that Tom loves Tom’s mother.
(1b) Tom said that Tom loves Bill’s mother.
(1c) Tom said that Bill loves Bill’s mother.

I don’t think the second sentence can mean

(1d) Tom said that Bill loves Tom’s mother.

Now why should this be? I thought the explanation for the original ambiguity was that we can break down the first sentence in (1) the following ways under lambda abstraction. (This will all be in words because I don’t really know how to use the right symbols in HTML.)

(1a’) Bill self-ascribed the property of being a person who loves his mother.
(1b’) Bill self-ascribed the property of being a person who loves Bill’s mother.
(1c’) Bill ascribed to Bill the property of loving Bill’s mother.

But we could equally analyse the first sentence in (1) as (1d’).

(1d’) Bill self-ascribed the property of having a mother loved by Bill.

And if Tom also did that, then ‘So did Tom’ would mean (1d). But that’s blocked somehow? What’s going on?

One explanation is that (1d) is a disambiguation of ‘So did Tom’, but it’s never the salient disambiguation because of pragmatic reasons. But what would the pragmatic reasons be?

Another explanation is that the differences are due to differences in the binding constraints on ‘he’ and ‘his’. I’m not really sure what those constraints could be, since they couldn’t be absolute constraints. When (1) means (1d’), ‘he’ means just what it does in the (1c’) interpretation, and ‘his’ means just what it does in the (1a’) interpretation. (This sentence is rough, but I hope it’s clear enough.)

To complicate matters, it seems (2) is four-way ambiguous.

(2) Bill said that his mother loves him. So did Tom.

I think that with a bit of effor we can get all four readings for this sentence.

(2a) Tom said that Tom’s mother loves Tom.
(2b) Tom said that Tom’s mother loves Bill.
(2c) Tom said that Bill’s mother loves Bill.
(2d) Tom said that Bill’s mother loves Tom.

To get the final interpretation, consider the following dialogue.

Who said that Bill’s mother loves them? Bill said that his mother loves him. So did Tom.

And to get the (2b) reading, simply stress ‘his’ in the first sentence of (2). At first I was unsure whether either (2b) or (2d) were available readings, but it seems that both are.

So maybe it’s just a pragmatic blocking on (1d), because it’s hard to see how we could have a semantic or syntactic reason for blocking (1d) as a disambiguation, while keeping all four here. Or maybe I’m just not alert enough to find a context where the (1d) reading is allowed, and if so there isn’t a problem at all. Very confusing.

Posted by Brian Weatherson in Uncategorized

16 Comments »

This entry was posted on Sunday, July 27th, 2003 at 12:58 pm and is filed under Uncategorized. You can follow any responses to this entry through the comments RSS 2.0 feed. Both comments and pings are currently closed.

16 Responses to “Ambiguity Confusions”

  1. Kent Bach says:

    Here’s a context where the (1d) reading is allowed:

    Q: In that discussion of guys loving other guys’ mothers, who said that he loves Tom’s mother?

    (1) Bill said that he loves his mother. So did Tom.

    I.e., (1d) Tom said that Bill loves Tom’s mother.

    I hope that’s not too big of a stretch, and that the set-up in Q is not too big of a cheat either. For in this context a speaker, in framing the first sentence of (1), would more naturally use ‘her’ rather than ‘his mother’: ‘Bill said that he loves her’.

  2. Michael Kremer says:

    Kent Bach’s example doesn’t really address Brian’s perplexity. Further it is not clear to me that it works at all.

    First, Brian’s perplexity was this: in a context in which “Bill said that he loves his mother” means that Bill said that Bill loves Bill’s mother, why can’t “So did Tom” mean that Tom said that Bill loves Tom’s mother? To this perplexity Bach’s example is irrelevant since Bach is imagining a context in which “Bill said that he loves his mother means that Bill said that Bill loves Tom’s mother.

    Second, I am not myself convinced that even in the imagined context “Bill said that he loves his mother” can mean that Bill said that Bill loves Tom’s mother.

    Consider the following dialogue:

    A: In that discussion of guys loving other guys mothers, who said that he loves Tom’s mother?
    B: Bill said that he loved his mother. So did Tom.
    A: That’s not the point. Who said that he loved Tom’s mother? (emphasis on “Tom”)
    B: I meant that Bill and Tom both said that Bill loved Tom’s mother.
    A: Then why didn’t you say so? What you said was that Bill said that he loved his mother. What does that have to do with who said that they loved Tom’s mother? (emphasis on “Tom” again)

    A here treats B’s first sentence as saying that Bill said that Bill loves Bill’s mother, and so as a non-sequitur. And, even in this context, that’s how it sounds to me, too. I don’t really see it as meaning, even in this context, that Bill said that Bill loves Tom’s mother. In other words, I think A is right in the above.

  3. Brian Weatherson says:

    Michael’s right that the question I meant to ask was whether ‘So did Tom’ could have that meaning while the first sentence means “Bill said that Bill loves Bill’s mother.” But to be fair to Kent, I didn’t say that that was the question I was interested in, and I really should have made that clear.

    I don’t know how I’d interpret Kent’s sentence. I could certainly imagine the ‘his’ referring to Tom if it was accompanied by a demonstration of Tom. (I don’t think Michael means to deny this.) I suspect that anything that can be done with demonstrations can be done with antecedent salience, so I suspect there is some context where ‘his’ can pick out Tom, but I admit I can’t think of such a context off the top of my head.

  4. Matt Weiner says:

    Maybe there is a syntactic/semantic difference between (1) and (2).

    It seems to me that
    ?(3) His mother loves Bill
    may be OK.

    • He loves Bill’s mother,

    though, is right out.

    If this means that there’s a syntactic/semantic difference between (3) and (4), you might expect it to carry over to (1) and (2) in some way. On the other hand, I’m not at all sure that (3) is more acceptable than (4).

  5. Matt Weiner says:

    Intended interpretations of (3) and (4) are “Bill loves Bill’s mother,” of course.

  6. Michael Kremer says:

    Brian, you’re right about the demonstration of Tom. I’ll accept that (but hadn’t thought of it). That’s all for now.

  7. Brian Weatherson says:

    I’d like to think that something like Matt’s solution is right, because then this would be a (relatively) familiar puzzle about co-indexing. But I couldn’t figure out how the constraints might apply to allow all three versions in and block the fourth. I agree though that this looks to be the same phenomenon.

  8. Matt Weiner says:

    A linguistically savvy friend tells me that, in the current literature, (3) is OK and (4) is bad.

    The idea is (I think) that a pronoun can’t be bound by a noun that it c-commands.

    In (3) “His” does not c-command “Bill,” because “his” is part of a complex NP; so the binding is OK.

    In (4) “He” c-commands “Bill,” so coreference is bad.

    Sorry if this is what you meant by the relatively familiar puzzle, Brian—in linguistics in general “relatively familiar” doesn’t imply that I know the first thing about it. Anyway, I’ll leave it to the linguistically savvy among you to figure out whether C-commanding issues can take care of (1d).

  9. Brian Weatherson says:

    Relatively familiar is a pretty abused term here – it doesn’t mean that I know anything about it, just that I have a good idea about who to ask that will know the answer.

  10. Kent Bach says:

    Sorry for being off-line all day, but belated thanks to Michael for straightening me out, not once but twice. So let me try again, this time recognizing that the first sentence of (1) is to be taken with both ‘he’ and ‘his’ referring to Bill (I am not implying anything one way or another about ‘he’ and/or ‘his’ being anaphoric on ‘Bill’ or about ‘his’ being anaphoric on ‘he’).

    It seems that Q below, awkward though it may be, sets up a context where Brian’s (1d) is allowed:

    Q: Who said that Bill loves his respective mother?

    I.e., who said of his mother that Bill loves her, or, in the language of Brian’s (1d’), who self-ascribed the property of having a mother loved by Bill?

    (1) Bill said that he loves his mother. So did Tom.

    I.e., (1d) Tom said that Bill loves Tom’s mother.

    Still, this may seem a bit fishy and, indeed, I doubt that (1d’),

    (1d’) Bill self-ascribed the property of having a mother loved by Bill.

    gives a genuine reading of the first sentence of (1). That’s because ‘Bill loves’ is not a syntactic unit; it’s not a constituent of (1). So (1d’) is not a proper lambda-abstraction. But set-up question Q creates a conversational context in which one can get (1d) anyway.

  11. Matt Weiner says:

    Maybe we can prettify Kent’s example by changing Q to “Who said, ‘Bill loves my mother’?”

  12. Kent Bach says:

    Matt’s prettification seems fine if direct quotation doesn’t have to be strict.

    Robert May has reminded me that in their book, Indices and Identity (MIT Press 1994), Robert Fiengo and he discuss a virtually identical example in great detail. I checked and, indeed, in sec. 4.1 they present the example on p. 130, discuss the “predication” theory (Sag and Williams) on 132ff, take up various “condition” theories on 135ff, and in sec 4.2 (147ff) present their own theory. I should have consulted their book before opening my mouth, and urge you all to check out their intricate and illuminating discussion.

  13. Kent Bach says:

    Robert May sends the following link:

    http://kleene.ss.uci.edu/~rmay/Ellipsis.pdf

    to a recent piece he did on ellipsis for the Macmillan Encyclopedia of Cognitive Sciences. In it he briefly remarks on the relevant cases, discussing their implications for predication theory and their treatment in the F/M theory.

  14. Matt Weiner says:

    Hmmm. Do Brian’s 4 readings of (2) counterexample May’s claim that “What we actually observe in this case, as well as in those of increasing complexity, are only n + 1 readings; readings do not grow exponentially” (p. 1098, or p. 5 of the pdf)? Or is that restricted to (1) and its analogues?

    I can’t figure out whether the dependency theory yields the right prediction wrt (2). Well, I should be putting my computer in a box soon anyway, but I hope you all have hammered it out by the time I return.

  15. Robert May says:

    As Kent Bach mentioned, Bob Fiengo and I in our book Indices and Identity discuss the sort of cases under discussion in great detail. (We refer to it as the “many-pronouns puzzle.”) This includes, re: the previous post, examples like (2) in the initial post – see our example (53) on p. 156.

    While I appreciate the interest in these cases, I think that it would be worthwhile for people to have a look at the relevant material in the linguistics literature. While Fiengo and my book, especially ch. 4, is the most extensive and empirically detailed discussion of the specific cases, they date back to a paper by Osten Dahl from 1974.

    The paper of mine to which Kent linked is an overview article that I thought might be useful in orienting people to the development of the issues surrounding strict and sloppy identity. It is not meant to be comprehensive; for that, I once again refer people to Fiengo and my book.

  16. Matt Weiner says:

    Fair enough. I should point out that “I can’t figure out…” in this case is a statement about my cognitive limits, rather than a criticism of the theory or its presentation.