Finding Causes
Consider the following:
(1) Sally hates Mary.
a. How likely is this because Sally is the kind of person who hates people?
b. How likely is this because Mary is the kind of person whom people hate?
Sally hates Mary doesn't obviously supply the relevant information, but starting with work by Roger Brown and Debora Fish in 1983, numerous studies have found that people nonetheless rate (a) as more likely than (b). In contrast, people find Sally frightens Mary more indicative of Sally than of Mary (the equivalent of rating (b) higher than (a)). Sentences like Sally likes Mary are called “object-biased,” and sentences like Sally frightens Mary are called “subject-biased.” There are many of sentences of both types.
Brown and Fish, along with many of the researchers who followed them, explain this in terms of an inference from knowledge about how the world works:
(2) Sally hates Mary.
a. Who is most likely responsible? Sally or Mary?
(The photo on the right came up on Flickr when I searched for pictures about "causes". It turns out Flickr is not a good place to look for pictures about "hating," "frightening," or "causes". But I liked this picture.)
Understanding Pronouns
Now consider:
(3) Sally hates Mary because she...
(4) Sally frightens Mary because she...
Most people think that "she" refers to Mary in (3) but Sally in (4). This is a bias -- not absolute -- but it is robust and easy to replicate. Again, there are many verbs which are "object-biased" like hates and many which are "subject-biased" like frightens. Just as in the causal attribution effect above, this pronoun effect seems to be a systematic effect of (at least) the verb used. This fact was first discovered by Catherine Garvey and Alfonso Caramazza in the mid-70s and has been studied extensively first.
The typical explanation of the pronoun effect is that the word "because" implies that you are about to get an explanation of what just happened. Explanations usually refer to causes. So you expect the clause starting with she to refer to the cause of first part of the sentence. Therefore, people must think that Mary caused Sally hates Mary but Sally caused Sally frightens Mary.
Causes and Pronouns
Both effects are called "implicit causality," and researchers have generally assumed that the causal attribution effect and the pronoun effect are basically one and the same. An even stronger version of this claim would be that the pronoun effect relies on the causal attribution effect. People resolve the meaning of the pronouns in (3) and (4) based on who they think the cause of the first part of the sentence is. The causal attribution task in (1) and (2) is supposed to measure exactly that: who people think the cause is.
Although people have been doing this research for around three decades, nobody seems to have actually checked whether this is true -- that is, are verbs that are subject-biased in terms of causal attribution also subject-biased in terms of pronoun interpretation?
I recently ran a series of three studies on Amazon Mechanical Turk to answer this question. The answer is "no."
This figure shows the relationship between causal attribution biases (positive numbers mean the verb is subject-biased, negative means its object-biased) and pronoun biases (100 = completely subject-biased, 0 = completely object-biased). Though there is a trend line in the right direction, it's essentially artifactual. I tested four different types of verbs (the details of the verb classes take longer to explain than they are interesting), and it happens that none of them were subject-biased in terms of pronoun interpretation but object-biased in terms of causal attribution (good thing, since otherwise I would have had nowhere to put the legend). There probably are some such verbs; I just only tested a few types.
I ran three different experiments using somewhat different methods, and all gave similar results (that's Experiment 2 above).
More evidence
A number of previous studies showed that causal attribution is affected by who the subject and object are. For instance, people are more object-biased in interpreting The employee hated the boss than for The boss hated the employee. That is, they seem to think that whether the boss is more likely to be the cause whether the boss is the one hating or hated. This makes some sense: bosses are in a better position to effect employees than vice versa.
I was able to find this effect in my causal attribution experiments, but there was no effect on pronoun resolution. That is, people thought "he" referred to the employee in (5) and the boss in (6) at pretty much the same rate.
(5) The boss hated the employee because he...
(6) The employee hated the boss because he...
Conclusion
This strongly suggests that these two effects are two different effects, due to different underlying mechanisms. I think this will come as a surprise to most people who have studied these effects in the past. It also is a surprise in terms of what we know about language processing. There is lots of evidence that people use any and all relevant information when they are interpreting language. Why aren't people using the conceptualization of the world as revealed by the causal attribution task when interpreting pronouns? And what are people doing when they interpret pronouns in these contexts?
I do have the beginnings of an answer to the latter question, but since the data in this experiment doesn't speak it, that will have to wait for a future post.
---------
Brown, R., & Fish, D. (1983). The psychological causality implicit in language Cognition, 14 (3), 237-273 DOI: 10.1016/0010-0277(83)90006-9
Garvey, C., & Caramzza, A. (1974). Implicit causality in verbs Linguistic Inquiry, 5, 459-464
Picture: Cobalt123.
Consider the following:
(1) Sally hates Mary.
a. How likely is this because Sally is the kind of person who hates people?
b. How likely is this because Mary is the kind of person whom people hate?
Sally hates Mary doesn't obviously supply the relevant information, but starting with work by Roger Brown and Debora Fish in 1983, numerous studies have found that people nonetheless rate (a) as more likely than (b). In contrast, people find Sally frightens Mary more indicative of Sally than of Mary (the equivalent of rating (b) higher than (a)). Sentences like Sally likes Mary are called “object-biased,” and sentences like Sally frightens Mary are called “subject-biased.” There are many of sentences of both types.
Brown and Fish, along with many of the researchers who followed them, explain this in terms of an inference from knowledge about how the world works:
Consider the two verbs flatter and slander… Just about everyone (most or all persons) can be flattered or slandered. There is no special prerequisite. It is always possible to be the object of slander or flattery … By sharp contrast, however, not everyone, by any means, not even most or, perhaps, many are disposed to flatter or to slander… [Thus] to know that one party to an interaction is disposed to flatter is to have some basis for predicting flattery whereas to know only that one party can be flattered is to know little more than that that party is human. (Brown and Fish 1983, p. 265)Similar results are found by using other ways of asking about who is at fault:
(2) Sally hates Mary.
a. Who is most likely responsible? Sally or Mary?
(The photo on the right came up on Flickr when I searched for pictures about "causes". It turns out Flickr is not a good place to look for pictures about "hating," "frightening," or "causes". But I liked this picture.)
Understanding Pronouns
Now consider:
(3) Sally hates Mary because she...
(4) Sally frightens Mary because she...
Most people think that "she" refers to Mary in (3) but Sally in (4). This is a bias -- not absolute -- but it is robust and easy to replicate. Again, there are many verbs which are "object-biased" like hates and many which are "subject-biased" like frightens. Just as in the causal attribution effect above, this pronoun effect seems to be a systematic effect of (at least) the verb used. This fact was first discovered by Catherine Garvey and Alfonso Caramazza in the mid-70s and has been studied extensively first.
The typical explanation of the pronoun effect is that the word "because" implies that you are about to get an explanation of what just happened. Explanations usually refer to causes. So you expect the clause starting with she to refer to the cause of first part of the sentence. Therefore, people must think that Mary caused Sally hates Mary but Sally caused Sally frightens Mary.
Causes and Pronouns
Both effects are called "implicit causality," and researchers have generally assumed that the causal attribution effect and the pronoun effect are basically one and the same. An even stronger version of this claim would be that the pronoun effect relies on the causal attribution effect. People resolve the meaning of the pronouns in (3) and (4) based on who they think the cause of the first part of the sentence is. The causal attribution task in (1) and (2) is supposed to measure exactly that: who people think the cause is.
Although people have been doing this research for around three decades, nobody seems to have actually checked whether this is true -- that is, are verbs that are subject-biased in terms of causal attribution also subject-biased in terms of pronoun interpretation?
I recently ran a series of three studies on Amazon Mechanical Turk to answer this question. The answer is "no."
This figure shows the relationship between causal attribution biases (positive numbers mean the verb is subject-biased, negative means its object-biased) and pronoun biases (100 = completely subject-biased, 0 = completely object-biased). Though there is a trend line in the right direction, it's essentially artifactual. I tested four different types of verbs (the details of the verb classes take longer to explain than they are interesting), and it happens that none of them were subject-biased in terms of pronoun interpretation but object-biased in terms of causal attribution (good thing, since otherwise I would have had nowhere to put the legend). There probably are some such verbs; I just only tested a few types.
I ran three different experiments using somewhat different methods, and all gave similar results (that's Experiment 2 above).
More evidence
A number of previous studies showed that causal attribution is affected by who the subject and object are. For instance, people are more object-biased in interpreting The employee hated the boss than for The boss hated the employee. That is, they seem to think that whether the boss is more likely to be the cause whether the boss is the one hating or hated. This makes some sense: bosses are in a better position to effect employees than vice versa.
I was able to find this effect in my causal attribution experiments, but there was no effect on pronoun resolution. That is, people thought "he" referred to the employee in (5) and the boss in (6) at pretty much the same rate.
(5) The boss hated the employee because he...
(6) The employee hated the boss because he...
Conclusion
This strongly suggests that these two effects are two different effects, due to different underlying mechanisms. I think this will come as a surprise to most people who have studied these effects in the past. It also is a surprise in terms of what we know about language processing. There is lots of evidence that people use any and all relevant information when they are interpreting language. Why aren't people using the conceptualization of the world as revealed by the causal attribution task when interpreting pronouns? And what are people doing when they interpret pronouns in these contexts?
I do have the beginnings of an answer to the latter question, but since the data in this experiment doesn't speak it, that will have to wait for a future post.
---------
Brown, R., & Fish, D. (1983). The psychological causality implicit in language Cognition, 14 (3), 237-273 DOI: 10.1016/0010-0277(83)90006-9
Garvey, C., & Caramzza, A. (1974). Implicit causality in verbs Linguistic Inquiry, 5, 459-464
Picture: Cobalt123.
10 comments:
Hi
Thank you for an interesting post.
It seems that this effect is linked to cognitive biases. Our brains are have various mechanisms for dealing with information gaps in the world around us - and these heuristic systems lead to predictable errors in judgement.
There are fundamental qualitative differences between the sentence 'Sally hates Mary' and the sentence 'The boss hated the employee'.
In the first sentence, we know nothing about Sally and Mary. We don't know who they are and what relationship they have with each other. Lacking this contextual information we tend to fall prey to fundamental attribution error and allocate causality to some personality-based explanation. Lacking any other information, we conclude that it must be something to do with how one of the protagonists behaves rather than the situation in which they find themselves or the history of their relationship. The object-bias of 'hates' combined with fundamental attribution error encourages us to ascribe causation to Mary. So, when we add 'because she...' we are likely to complete the sentence 'because she is...'
In the second situation, we are given more contextual information about the protagonists. We know something about their relationship: one is the boss, the other is the employee. We know something about their situation and can make experience-based inferences about their history. This reduces the likelihood of fundamental attribution error. Rather than attributing the cause of the hate to personal characteristics of the boss or the employee, we are just as likely to attribute it to something that has happened between them in the past, or just the fraught nature of the boss-employee relationship. Now when we add the explanation 'because he...' we are less certain how it will continue. It could be 'because he is...' or it could be 'because he did...' or 'because he failed...', 'because he lied about...', etc.
Added to this, there's the the fact that one uses the present tense and the other the past. The use of the present tense implies an ongoing situation which is not changed by situational actions and so is more likely to be attributable to personal characteristics. The use of the past tense implies causation linked to a specific event that happened at a particular time.
(By the way, you may need to edit the first paragraph of the More Evidence section - it doesn't seem to make sense. And have you got object and subject mixed up or am I just confused?)
Without more detailed information about the data it's hard to know whether these concerns are relevant to your experiment or not. Apologies if I have got hold of the wrong end of the stick.
David
@DW Errors in the "More evidence" section are fixed. What an I say? Don't write blog posts at night. (I wouldn't normally, but this post also serves as additional debriefing for participants in the experiment, and it was overdue as was.) It should make more sense now.
As far as how people continue the sentence after "he". I find your intuitions there very interesting and probably relevant. For instance, I get different intuitions for
Sally liked Mary because she is a dax.
Sally liked Mary because she daxed.
with the second sentence being slightly less object-biased (the word "dax" here is just a made up word -- a noun in the first case and a verb in the second).
Overall, though, the difference between "like" and "frighten" shows up across many different ways of doing this experiment. I gave only one example. Another one is the method used above: write out the full sentence and then ask people to interpret the pronoun.
Way back when I did the 'Find the Dax' experiment, I remember thinking what I still think now, that sound symbolism will be a significant factor in the results. That is, when a sentence could be either subject-biased or object-biased, people will be swayed in part by what sort of image the look and sound of the word "dax" conjures up (e.g. it may sound alien, robotic, untrustworthy, etc). A proper study of pronoun resolution would require several invented words in order to cancel out this effect.
Do you discuss such things in your report on the experiment? I do hope so.
"A proper study of pronoun resolution would require several invented words in order to cancel out this effect."
You raise a good point. I agree, and I do control for such things. Well, I did in Pronoun Sleuth (the novel word is randomly assigned for every participant for every sentence).
It happens that I didn't do so in this particular study; instead, I controlled for the potential confound other ways (for instance, by running experiments with and without novel words -- which incidentally gave the same results).
I don't usually go through all the methodological details in posts like this on the theory that most people don't care. But I'm always happy to discuss details in the comments.
Are "hates" and "frightens" comparable verbs?
"Hates" is an active verb expressing a relation from subject to object.
"Frightens" is a passive verb expressing a relation from object to subject (or with those reversed).
"Mary hates Sally" is comparable to "Mary fears Sally" because "fears" is an active verb.
"Mary frightens Sally" is comparable to "Mary disgusts Sally" because "disgusts" is a passive verb.
Sorry I don't know technical linguistics terms for what I call active/passive.
We can explain uncertainty of users because passive language is known to be difficult.
@jvn0: You have some interesting points. I think you are focusing on the fact that subject of "hate" and the object of "frighten" have to be animate (people, animals, etc.), and animates are usually subjects. On the other hand, though, the object of "hate" and the subject of "frighten" are in some sense the cause of the emotion, and causes are usually subjects. So while there are good arguments that "frighten" is somehow backwards, there are just as good arguments that "hate" is the weird verb.
In terms of research, there was a movement in the 80s to argue that basically the language system can't figure out what the subject should be, so for some verbs people agree to make the animate the subject (e.g., hate) and for some the object (e.g., frighten).
In the 90s, other linguists pursued a theory close to what you advocate here. It didn't do well. It turns out that most of the data they thought supported the theory were simply misinterpreted. When I started my work a few years ago, I actually was pushing the opposite claim: that "frighten" is the normal kind of verb and "hate" is weird. But the data I've gotten in the last couple years suggest that something much, much more complicated is going on.
What I think is going on will require a number of posts (I said it was complicated!), so I won't go into it here.
In any case, I'm not sure anything in my data suggest that people were confused by the "frighten" verbs. Which part of the post were you referring to?
Hi. Stumbled over your blog and have been reading about pronouns. I have to ask, have you considered signed languages?
In ASL for example, the example sentences are not ambiguous at all, because signed pronouns are not. However, I can sign them so the causal attribution effect shows up. This seems provide a negative answer to the question of whether the causal attribution effect and the pronoun effect are one and the same.
@Uzza: How do third-person pronouns work in ASL? That is, suppose you're referring to someone who isn't there. Do you do this through a spatial reference (pointing to the spot where you previously signed the name of the person)? More detail would help.
Yes, the pronouns are marked for a specific location in space, and since its impossible for more than one person to occupy said space at a given time, there's no ambiguity possible. Rather than he or she, a better translation might be “person in location X”. Then, with first mention of Mary made to your right, at location R, and first mention of Sally to your left;
(a) Location-R-person loves location-L-person because location-R-person …
The only interpretation possible is: Mary is the lover, Sally the lovee, and the reason is something about Mary. This appears to eliminate the pronoun effect, no? Then, ASL has pretty free word order, so I can sign
(b) person-at-R loves person-at-L, because, really nice is person-at-R.
If I were to cut that off just before the last word, I believe that would generate your causal attribution effect, leaving the signee wondering whether Mary loved Sally because Sally is so lovable, or whether she did so because Mary herself was just a loving type person.
When I signed the last word I'd be pointing at Mary (or her location) and the issue would get resolved in favor of Mary just being so lovey. Wouldn't such a test separate the two effects you are looking at?
@Uzza: I suspect it just means that one couldn't study the pronoun effect by asking about interpretation of pronouns that have already been produced, but you could study it other ways (as you suggest, by doing a kind of "fill-in-the-blank" method; I don't use that method often, but I have used it and lots of other people use it more).
Still, it's worth thinking about whether this different pronoun system might give rise to useful effects that could be tested in ASL but not spoken languages. Thank you for bringing it up.
Post a Comment