10/30/2009

Differing interpretations of conditionals

This is not really a philosophy post; instead, I pretend to be a (bad) linguist.

The textbook I'm using for my critical thinking class this term (Feldman's Reason and Argument) claims that the ordinary English sentence 'If Joe is a professional basketball player, then he is tall' is true. This surprised me somewhat, since there are professional basketball players who are not tall (though of course there are relatively few).

I wanted to know if I was strange in this regard, so I took a quick survey of my students. I asked them whether they thought the sentence 'If today is a February day, then the high temperature today is under 40 F in Geneva, NY.' (Highs in Geneva in February are around 30 or so.) 12 of 24 thought this was true. We then had a little discussion about how one group was requiring conditionals to be exceptionless, whereas the other group was allowing a few exceptions.

Thinking about this later, I realized that the majority of people who said it was true were female, and the majority who thought it false were male. I didn't tally votes by sex of respondent, so I don't know how pronounced the difference was, and my sample size is extremely small, so the difference was almost certainly not statistically significant. But it does seem like it might be something worth investigating.

We could perhaps generalize this by asking: when there are multiple ways for a hearer to interpret a speaker's utterance, only some of which are true, are female hearers more likely to attribute the true interpretation than male hearers? In Gricean terms, are female hearers more likely to assume the speaker is following the maxim of quality (Contribute only what you know to be true; Do not say false things; Do not say things for which you lack evidence) than male hearers?

Perhaps this has already been dealt with in the pragmatics literature. But a quick google search did not reveal an answer to this specific question (though I did find interesting research on gender differences with respect to other Gricean maxims).

11 comments:

Jonathan Livengood said...

Greg,

As you may or may not know, Justin Sytsma and I have a website for experimental philosophy research. This looks like the sort of thing that we could test pretty quickly. Would you be interested in actually writing up such a study with us?

Whether we do that or not, though, I'm not convinced that it's Grice's maxim of quality that is at play. As I understand that maxim, applying it means treating utterances as *believed to be true* by their utterers and perhaps treating utterances as true when we have no reason to doubt them. But in both cases you mention, we have reason to doubt the claims if they are understood as exceptionless.

I wonder if some people are treating the conditional as rules licensing ampliative inferences. So, conditionals would then be read something like: if x, infer that probably y.

Greg Frost-Arnold said...

Hi Jonathan --

Thanks very much for the offer! I definitely might be interested in pursuing this, if I can find the time.

I'm not sure I completely understand your remarks in the second paragraph. I agree that the maxim of quality involves the hearer "treating utterances as believed to be true" by the utterer. But it's hard for a hearer to accept that someone who lives in upstate NY does in fact believe that every single February day has a high under 40 F.

My hypothesis is that (in my tiny sample) male hearers are more willing to attribute obviously untrue beliefs to speakers than female hearers are. Perhaps that hypothesis is distinct from Gricean issues; but I'm not 100% sure why it would be.

Jonathan Livengood said...

I absolutely agree that it is hard (impossible?) to accept that someone who lives in upstate NY actually believes that every single February day has a high under 40 F. Given that, it seems unlikely that those students who endorsed the conditional did so on the basis of Grice's principle of quality. That was my point.

But maybe that was too hasty. (Let me just think out loud for a bit.) Do you think you can get a principle of charity from the maxim of quality?

I don't see how to get from 'X accepts the maxim of quality, and X says that p' to 'p is true'. I can see how to get to 'X believes that p is true' or even to 'X has good evidence for (reason to believe that) p'. Both of those may be true and p still be false.

The point so far, then, is that your hypothesis isn't *obviously* Gricean. It would be if your claim were that males are more likely to think that a speaker is lying. But a speaker may very well utter falsehoods without violating the maxim of quality (i.e. without lying).

But half of your students, who ought to know that p is false, endorsed the conditional anyway. Why?

Here are two possible explanations. (1) Half of your students have a different conception of the conditional--they do not treat conditionals as exceptionless. (2) Half of your students applied a principle of charity by which they interpreted the conditional so that it was more likely to count as true, even though they share your conception of the conditional as exceptionless.

We have identified (roughly) two interpretations of the conditional: (p) if today is a February day, then the high temp today is under 40 F in Geneva, NY. One interpretation (as a material conditional) makes p false. The other interpretation (as a statement that the consequent is likely given the antecedent) makes p true (or at least more likely to be true).

Given an utterance of p, we should pick the interpretation most likely to make it come out true, by the principle of charity. How might this have something to do with Grice? Well, maybe the maxim of quality entails the principle of charity. If we assume that everyone does his or her best to say only what is true, then we should expect what a speaker says to be true. If it is apparently not true, then we should ask how someone might come to believe that it *is* true. That might force us to interpret utterances charitably. (I for one would be more likely to make a guess about the evidence the utterer might have and would point to defeaters for the conditional. But then, I think of conditionals as exceptionless.)

Greg Frost-Arnold said...

Hi Jonathan --

Thanks for the thoughtful reply. You're probably right that decoupling this stuff from the whole Gricean apparatus would probably make things clearer..I just wanted to make sure I understand something you said:

"Here are two possible explanations. (1) Half of your students have a different conception of the conditional--they do not treat conditionals as exceptionless. (2) Half of your students applied a principle of charity by which they interpreted the conditional so that it was more likely to count as true, even though they share your conception of the conditional as exceptionless."

Just to make sure I understand: Hypothesis 1 is that the hearers think the correct interpretation of THE CONDITIONAL (whatever that is) allows some exceptions.
Hypothesis 2 is that the hearers think the correct interpretation of THE CONDITIONAL is exceptionless, but are willing to attribute to the speaker a different interpretation of the conditional in which the conditional allows exceptions.

Is there a way to distinguish between those two experimentally? I would've thought asking 'Is P true?' as opposed to 'Does the speaker believe P?' would distinguish them, but I could be wrong.

Jonathan Livengood said...

The short answer: I don't know. I thought a little bit about experiments for this problem, but I'm worried that nothing really simple is going to work.

If you're interested in actually running an experiment or two, we should involve Justin (and maybe Edouard as well) and discuss some possibilities.

Justin Sytsma said...

Did your text offer a reason for thinking that "if Joe is a professional basketball player, then he is tall" is true? I have to say that it doesn't strike me that way. It would be interesting to test out a few sentences like that on a larger sample and see if the perceived gender difference holds up. I would probably then ask them to explain on a second page. Assuming that a non-trivial percentage answer that the sentences are true, the explanations might cast light on which hypothesis is correct.

Off the top of my head, one way you might be able to get at Hypothesis 1 versus Hypothesis 2 is to give a story in which you vary information about the person expressing the sentence, then ask if the sentence is true. If Hypothesis 1 is correct you wouldn't expect information about the speaker to matter, while it should for Hypothesis 2. I'm not sure what the best way to set up the story is, but you might describe the speaker as being very earnest and saying the sentence very insistently, or sarcastic and saying it flippantly. You would probably want to run a few control sentences as well (something uncontroversially true/false to make sure that people’s like/dislike for the person isn’t generating a difference).

Greg Frost-Arnold said...

I'm still thinking about this, in particular about how to distinguish between Hyp 1 and Hyp 2 experimentally.

How do we tell whether someone really holds the view that 'If... then...' statements allow for exceptions? Here are a couple of ways that might do it.

Suppose we give subjects the sentence:
(*) If someone is in the NBA, then he is at least 6'2".

1. Ask: In which of the following scenarios is (*) FALSE?

a. Exactly 1 NBA player is under 6'2".
b. 1% of NBA players are under 6'2".
c. 50% of NBA players ... .
d. 0 NBA players ... .

2. Ask: does (*) mean the same as any of the following?
a. All NBA players are over 6'2"
b. Most NBA ... .
c. Some NBA ... .
d. None of the above
[Perhaps it would be better to ask: 'Which of the following is closest in meaning to (*)?', instead of 'means the same', since I'm not sure the latter is theoretically neutral.]

Unrelatedly, I've been thinking about another interpretation of the data, distinct from Hyp 1 and Hyp 2. On hyp 1, hearers think the conditional allows exceptions; on Hyp 2, hearers think the conditional is exceptionless. Here's an alternative: 'If..., then...' statements are semantically underdetermined. They don't "really" have either interpretation simpliciter. The meaning of the conditional is open between the exceptionless and exception-allowing meanings, and it is up to the hearer to decide between them given the context of the utterance. (Just as speakers resolve ambiguities.)

Trey Delap said...
This comment has been removed by the author.
Trey Delap said...

"If today is a February day, then the high temperature today is under 40 F in Geneva, NY."

I don't think this is a good example to test or teach critical thinking. What is February? What time of day is it, evening, midnight, early morning? It requires to much conditional evidence. What is the usual range of temperatures in Geneva NY in February? I have no idea what the parameters are so I cannot evaluate this claim at all. How valid is this evidence? How long has it been tracked? Even if trending is suggestive of consistency how do we account for variations? Are we talking about the place now called Geneva NY today or in prehistoric times?

What happened to "Practical Reasoning in Natural Language" by Stephen Naylor Thomas? That is an outstanding introductory text. I highly recommend it.

Justin Sytsma said...

Hey Greg,

I like your idea for a test between Hyp 1 and Hyp 2. I think it might be better to breakdown the first question into a between subjects design, telling a given participant either (a), (b), (c), or (d) and then asking them whether the statement is true. You might then vary how the information is presented, putting the statement into a given person's mouth, saying that he correctly believes (a), (b), (c), or (d), and asking whether what the person said is true. This could then be followed up on a second page by asking them whether the statement means the same as ... or perhaps to be a bit less leading, just asking them to paraphrase the statement.

Anonymous said...

Wilmot Sweeney at your service-

Perhaps men are more likely to
assume a "necessarily", as in
"the temperature is (necessarily)
below 40 degrees"; where women are more likely to assume, to insert, a "probably".
If so, this could be in accord with the stereotype of women as likely to interpret in a way that allows them to assent more easily to another's point of view and avoid aggressive analysis that leads to possible contradiction of a group member--causing social friction-- and so women assume the weaker assertion by the statement.
Whereas, the stereotypical male is less concerned with such and more with interpreting to give the easiest occasion for proferring his own judgement---the interpretation that is most likely to make him correct and therefore powerful--and so assumes the statement makes the stronger
assertion, more easily denied.
I have found stereotypes at certain times and in certain contexts, accurate; is this one of those instances? I can only speculate.
Gender differences, a fascinating area---