Factive Verbs and Protagonist Projection

Throughout most of the 20^th century, the medical community knew that large amounts of stress causes stomach ulcers—or at least they though they did.  Here are some "known facts" (things that doctors thought were true) prior to 1979 about ulcers:

 

From "Helicobacter Connections"

But then, in the 1980s an Australian physician Dr. Barry Marshall, convinced that the medical community had it all wrong, infected himself with what he believed to be the real culprit responsible for the disease, H. pylori.  In order to prove his theories against the overwhelming consensus, Marshall famously drank the bacteria in question.

 

And as it turns out, he was absolutely right.  We now know that peptic ulcer disease is actually caused by bacterial infection. 

Often, when people tell this story, they say things like "Everyone knew that stress caused ulcers, before an Australian doctor in the early 1980s proved that ulcers are actually caused by bacteria."  But as philosophers we might wonder what we should make of this kind of claim.  Did the doctors really know that stress was causing the ulcers or did they only think they knew it?  Or phrased a bit more generally by Socrates “If one fails to get at the truth of a thing, will he ever be a person who knows that thing?” The resulting dialogue in Plato’s Theatetus reveals his interlocutor’s answer (186c-187b).  Knowledge cannot be mere opinion, because there may be a false opinion.  And the answer contemporary philosophers give has changed little since.  Basically every epistemic analysis today includes a truth condition for knowledge.  It’s just overwhelmingly clear to philosophers that only true things can be known. 

Nonetheless, there seem to be several examples today showing that non-philosophers do not find this thesis obvious, and that at least as far as ordinary language is concerned, people frequently use ‘know’ in what appear to be blatantly non-factive ways.  A quick Google search--from the ulcer case, to headlines in the New York Times, to major blockbuster movies--reveals that non-factivity may be all around us!

 

Examples like these have led a growing number of philosophers to begin to speculate about the role of factivity in the actual knowledge judgments people make, as well as the significance these ordinary judgments might have for traditional epistemic theorizing.  John Turri for instance, has advanced a performance-view of knowledge that allows for knowledge of “approximate truths”, which are strictly speaking, false beliefs (2011; forthcoming).  Sympathy for non-factivity has been expressed in research by Daniel Nolan (2008).  And, the possibility that epistemic contextualism might allow for contexts under which certain kinds of false beliefs qualify as knowledge has also been (at least) discussed in the works of Keith DeRose (2009).

And perhaps the most comprehensive challenge against orthodoxy to date is presented by Allan Hazlett. In two recent papers, “The Myth of Factive Verbs,” and “Factive Presupposition and the Truth Condition on Knowledge” Hazlett displays compelling evidence that people ordinarily use purportedly factive verbs like ‘know’, ‘learn’, ‘remember’, and ‘realize’ in utterances of the form ‘S knows that p’ in ways that frequently do NOT require that p is true.  Hazlett’s theory is that these kinds of sentences are acceptable to non-philosophers because the folk concept underlying the meaning of ‘knows’ allows that false things can genuinely be known.  

But given that a number of epistemologists have begun to focus more on the truth condition in light of intuitions about ordinary usage, we may wonder, could it really be that the folk concept of knowledge is truly a non-factive concept?  Afterall, if true, it seems this would have a series of important epistemic and methodological implications about the connection between the ordinary concept and the (decidedly factive) concept of knowledge philosophers have historically been interested in analyzing.

Besides non-factivity however, another possible explanatory hypothesis of these linguistic data (like the ulcer case) is that ordinary uses of ‘knows’ are highly sensitive to something called ‘protagonist projection’.  The basic idea basic idea is that non-factive uses frequently might just appear acceptable to us only because we take up the protagonist’s perspective and say what seems true from their perspective--and not because people think it's actually possible to know false things. (This is kind of like when one talks about someone else, and—with sufficient cues, e.g. imitating their bodily language and tone of voice—can then use ‘I’ to refer to this other person.)

To prove that Experimental Philosophy isn’t always about attacking the great tradition, I went ahead and ran some experiments attempting to confirm armchair intuitions about the factivity of ‘knows’.  A draft of the paper is available HERE.  The main goal of these experiments was to use explicit paraphrasing tasks (a method pioneered in the study of mental state attributions to groups) to see if the linguistic evidence about factivity collected so far is better explained by (i) the folk tendency to adopt the perspective of the putative ‘knower’ via protagonist projection when attributing knowledge to falsehoods or (ii) an underlying folk concept which really does allow for knowledge of false things.  

In the paper above, there's work looking at the ulcer case, as well as some other examples from the web and from movies.  However, here is another experiment from the paper looking specifically at the role of false complement clauses on people's knowledge judgments.  In a 2 (truth) x 4 (verb) between-subjects design, participants (N = 217, 102 men, median age = 34) were presented with the following story about two police offers relaying some testimony.  While each participant only saw one story, the stories varied by the verb used in bold (‘know’, ‘learned’, ‘realized’, or ‘believe’).  Below is an example of the vignette involving ‘knows’:

Officer Ted asks the Police Sergeant, “Is there any information from the FBI about how the bomb was constructed?”

The Sergeant told him, “From the investigation, they know the bomb was homemade.”

After seeing one vignette, roughly half of participants were presented with a complement phrase expressing a false proposition.

False Complement: But actually, the bomb in question only appeared homemade.  Instead, professionals constructed it in a high-tech chemical plant.

The other half received a complement expressing a true proposition: 

True Complement: And as it turns out, the bomb in question was homemade. Non-professionals had constructed it in a basement apartment. 

They were then asked to select one of the explicit paraphrases provided regarding the usage of that verb (order of these answer choices randomized in all experiments):

Which of the following do you think best describes what is meant in the BOLD portion of the above sentence:

A) The FBI thought they knew (projectionist answer)
B) The FBI really did know (non-factive answer)

If participants are reading these sentences and interpreting the verbs ‘know’, ‘learned’, ‘realized’ as factive, then we would expect them to be more likely to adopt the projectionist answer in false complement cases—that from protagonist x’s perspective, ‘x thought that x Φ’.  Alternatively, if the sentences really are rightly interpreted as counterexamples to Factivity, then we would expect participants to give the non-factive answer—that from the participant’s more informed perspective, “x really did Φ.”  We would also expect a null effect for ‘believes’, whereby the truth or falsity of the complement still results in a majority of non-factive answers (and of course, high “x really did Φ” answers for all verbs in the true complement conditions).

And that was exactly what was found.  Most participants gave the ‘really Φ’ answer for ‘know’ (86%), ‘learn’ (78%), ‘realize’ (90%), and ‘believe’(80%) when presented with the true complements.  But participants were significantly less likely to  give the ‘really Φ’ answer for all factive verbs in false complement cases: 12%, 21%, and 31%, respectively.  And as expected, the false complement case had no significant effect on judgments about ‘believe’ (61%).  Lastly, the difference between projectionist answers in all the factive verb sentences compared to the ‘believe’ sentences—within just those cases with false complements—continues to be significant: 

Percent of participants giving “really Φ” answers grouped by true or false complement.

 So one interpretation is that these data begin to call into question whether people’s treatment of these particular sentences really is best interpreted as a counterexample to Factivity.  On the other hand, it looks like people may be engaging in projective readings, rather than actually attributing knowledge to subjects with false beliefs.

Of course, the data (here and throughout the paper above) do not show that all would-be non-factive uses of ‘know’ are cases of protagonist projection.  Nor do they show that there cannot be evidence against factivity independently of projection phenomena.  What they do suggest however is that absent strong evidence in favor of the non-factive concept, orthodoxy must still stand.  In the meantime, it looks as though the folk concept might be more closely related to the philosophical one (regarding knowing false things) after all. Or as Joe Friday used to say, when it comes to knowledge perhaps “All we know are the facts, ma'am”.

 

 

So we might conclude that without further evidence to the contrary, the data suggest that philosophers who choose to justify provisions of factivity in their philosophical analyses of knowledge by appeal to the ordinary usage or the ordinary concept of knowledge now have empirically supported reasons for doing so.  And--perhaps more importantly--we might also conclude that this result serves to highlight the need for more empirically informed research in epistemology when making inferences about ordinary concepts or language practices.

Comments

[Eddy Nahmias]

Hi Wesley, since no one is weighing in, I'll just say that I really like these experiments. You get very clear results using what seems to me to be good materials to test the hypothesis. Did you use more than just this scenario? I ask just because this one includes a non-standard feature--you have an organization (FBI) as the subject of the mental states. Do you have others with individual persons as the subject? (Yes, I know Mitt, corporations are people too.)

[Wesley Buckwalter]

Hey Eddy, thanks for weighing in! So in the paper I do four different types of scenarios. And the scenarios are tied very closely to the work presented by Hazlett (2010) and discussed by Turri (2011). The general thinking was that knowledge sentences in those specific cases constituted the evidence that a defender of factivity need explain. But I totally agree with you that there's a worry in this case that talk of 'the FBI' generally might cue participants to not give strict literal readings. So I ran a follow-up experiment to make sure the results for 'know' replicated when people were asked about a specific person and not just a group:

Officer Ted asks the Police Sergeant, “Is there any information from your contact at the FBI about how the bomb was constructed?” The Sergeant told him, “Yes. Agent Smith told me that, from his investigation, he Φs the bomb was homemade.”

This was followed by the same true/false complement clauses as before, and then people were asked if 'Agent Smith thought he Φ the bomb was homemade’ or ‘Agent Smith Φ that the bomb was homemade’. And with these changes in place the results seemed to clean up a little more. So for instance, 92% of those in the false ‘know’ condition gave the projectionist answer ('thought he Φ'). But alternatively 93% gave non-factive answers (‘Agent Smith Φ-ed’) for false ‘believe’, 83% for true ‘know’, and 94% for true ‘believe’. So it could be that people's tendency to favor projectionist readings are even more pronounced the more explicitly described the particular protagonist is to be projecting into.

[Dave Maier]

Here's another data point, of fairly limited significance, but it has stuck in my mind. When I was about 10 or 11, I had a conversation with a peer which went something like this:

Me: You can't *know* something unless it's *true*.

Him: I can *know* that 2 + 2 = 5.

Me: No you can't, because 2 + 2 = 5 is false. You can't *know* it unless it's *true*.

[repeat until frustrated and bored]

In any case I don't think it has that much philosophical significance that philosophers often use words in particularly circumscribed senses, while others sometimes use them differently in other contexts. Words have different senses; you just have to keep them straight.

Still, I agree with my 11-year-old self that there's no point in using "know" that way (i.e. non-factively) when we have a perfectly good word for that already (= "believe"). But it seems like some people do.

[Eddy Nahmias]

Similar to Dave's case, I'd say about 10% of my undergrad students seem to use 'know' in the sense of 'justified in believing', since they agree with (and strongly defend) statements like: "People living thousands of years ago *knew* that the earth was flat and that the sun revolved around the earth." I feel the pull of this intuition, perhaps in part because people back then seem at least as justified in believing those things as we are justified in believing many of the things we seemingly know. (It might also be that some people are using 'know' to mean 'believe really strongly' which would also explain their response to 'knowledge' about the earth being flat, not to mention "I know that God exists," etc.)

This will betray my ignorance of contextualism, but can someone tell me why it is (as I think it is) that contextualists do not say that those people living long ago *knew* the earth was flat.

[Wesley Buckwalter]

So the precise details about how factivity (in the sense that no false things can be known) is implemented might depend on the particular epistemic theory considered. But in many cases, I think a lot of contextualist (/non-contextualist) theories just end up stipulating it. There is a nice discussion of this question in the first chapter of DeRose's book. When considering just how low the low epistemic standards can go, say, past just having a true belief only, "Why stop there? Why for instance, think that even the truth of p is always necessary for the truth of 's knows that p'? DeRose ends up appealing to ordinary language and PP here when setting the floor for 'knows'. Though one could defend a c-theory with low standards that low if one wanted!

[Robert Hvistendahl]

Sorry I don't think this is relevant to this thread, but I wrote my opinion on knowing truths.

Does something need to be true for one to know it?

Does it depend on the way we define know?

Let's not forget that we produced the word 'know', we created it for our own use, we defined it. So we need to be careful that we are not arguing that the definition is one such thing or another, as we choose it's definition. And we can choose for it to mean that something would have to be true in order for one to know it, should we? Or do we? Very interesting questions and I love your experiments. After thinking about it for a while, in order to reach a conclusion about whether we do or should, one would actually have to define truth, but I haven't given that a go for a number of years. So I'll talk about what goes on in one's head when he 'knows'.

It's important to realize that talking about knowing is talking about a process of the mind, it's very relevant to think about the way people feel when they think they 'know' something.

If you tell me the FBI knows the bomb is homemade, and later we find out that it is not, then either
1.The FBI only thought they knew.
Here, knowing --> truth, in this case, you lied to me because you should have said, the FBI thinks they know that the bomb was homemade.

2.The FBI really did know
Here, knowing does not imply truth and people can know false things. You did not lie to me.

I wouldn't agree with an FBI agent who said, I know the bomb is homemade in any case, he may strongly suspect it was, but he shouldn't claim to know such things. Maybe my overall conclusion here is that we have no way of understanding what knowing is, so I guess claiming to know is the closest we can get, in which case, one can claim to know something that is not true.

We can't map our thoughts, we don't have an encyclopaedia with different types of thoughts in them. Both the specific physical process by which electrical impulses cross synapses and such (im not a neural doctor so I don't know how it works exactly) in the particular part of the brain that is excited when we feel that something is true (if this mechanism in the brain exists) and the sensation/s (I use the term sensation very loosely, just to mean any of the experience we have, including our thoughts/feelings, emotion, touch, any information that is available for one's interpretation) we experience (our own feelings) when we feel something is true, are too complex and not within our grasp to reproduce, not like words we can list in a dictionary. I can't give you a thought, you can't feel what I feel. This is a hindrance in our quest to understand 'knowing' and truth.

Were we able to distinguish between what goes through the head of an FBI agent when they *know* the bomb was homemade in a case where it actually was and a case where it was not, we would immediately be able to define two different words, one meaning 'to know because it is true' and the other being to know even though it is not'. If there was a difference in the two cases, this would be strange as the brain would have to know the real truth at the same time the FBI agent did not. There cannot be a difference in these two processes as the brain does not know everything (I assume it doesn't)

If there was a particular neuron in the brain that was excited every time we experienced a truth and we could see into somebody's brain and see when this neuron was excited, we would know if things were true.

So in the end it does come down to how we define 'know', maybe it's that simple.
What are we doing when we try to define this concept, we are trying to put a label on a particular process, it's a persistent process that occurs every day in our life. But since we are unable to find any grassroots property/s that are common to each time we know, it may be fruitless to define such a word. For example I feel a certain way when I see my red curtain and think, this is red. And I feel a certain way when I think, it is blue. I can't experiment with these two different thoughts because I don't have magic experimental tools and the thoughts are not like material objects in front of me. If I could I would dissect them with a scalpel, or boil them in a beaker to find boiling point, or bounce alpha radiation off of them to try to understand what the differences between the two are.

[Warren Olson]

There is of course the old argument here that people used to KNOW the world was flat!
My interest in the subject focus's on how epistemology differs between cultures.
I found a project quoted in Nisbetts Geographpy of thought ( and no doubt elsewhere ) was very enlightening ; It concerned placing Asian and then US students in front of an aquarium.
Around 90% of the Asian contingent commented on the overall picture, range of colours etc ; whereas a similar % of the US Students, commented on the 1 especially large fish in the tank.

Sorry if I have strayed a little from the subject, but any comments/similar research along these lines wouold be much appreciated.

[Jennifer Nagel]

Wes, thanks for the paper.

Warren, here's another line on the Nisbett program: http://languagelog.ldc.upenn.edu/nll/?p=478

[Allan Hazlett]

Very cool experiments. Experiment 1 (from the paper) is probably the most interesting, for my part, because it involves an alleged non-factive use of "knows." The other experiments don't, or at least don't obviously, involve uses of "knows" that look non-factive. (I don't think we can expect to get intuitive non-factive uses by taking an intuitively factive use of "knows" and appending to the story an ending that says the proposition known was false, as in the Crab cases and the FBI cases in the paper.)

I'm not sure that all this ends up supporting the truth condition on knowledge. It seems consistent with the idea -- this is what I want to say in the new paper -- that when "A" is factive, and when "S A's p" factively presupposes p, it isn't because "S A's p" entails p. We can get factive and non-factive uses of "learns"; but this means that factive presupposition isn't in general down to entailment. (If the connection between "S knows p" and p's truth is a matter of factive presupposition, then we don't, so I argue, have support from that connection for the truth condition.)

It would be really interesting (for epistemology) to see how people respond with some classic factives like "perceives," "remembers," and "sees." My hunch is that these are going to fall in with "learns" and "realizes."

Given the option of (A) "Everyone thought they knew" and (B) "Everyone really did know," the correct answer is obviously (A). (Experiment may confirm that, but one could have seen it coming.) It's interesting to think about what people would say in response to an open-ended question, like "In the story, did people know that stress caused ulcers?" Given the contrast between "really knowing" and "merely thinking they knew," the answer is obvious. Given the (A)/(B) choice, the contrast for "really" is made explicit. There are other possibilities, though. As in:

(1) Did Zinn really learn that World War II made the world safe for democracy? Or did he just use that as an example of educational corruption?

(2) Did Zinn really learn that World War II made the world safe for democracy? Or was his “learning” really ideological misinformation?

Question (1) is a question about Zinn's personal history; question (2) is a question about global military history. The answer to (2) is: Zinn didn't really learn. The answer to (1), we can imagine, is: Yes, Zinn really did learn that.

Upshot: it's not surprising that people say that people didn't really know stress caused ulcers, when the contrast for the "really" is such that the question, of whether people really knew, is a question about ulcers. The answer then is obvious: they didn't really know (i.e. stress doesn't cause ulcers). But it's also possible to imagine contexts in which the contrast for the "really" is such that the question is about people's mental state. E.g. "Did they really know that stress causes ulcers, or were they just pretending to think that, to get funding for their anti-stress drug?" I'm not sure what people's answer would be there, but the hunch is that you could get more people saying that they "really" knew (i.e. they weren't pretending), when the contrast is that they were pretending to believe.

[Wesley Buckwalter]

Hey Allan, thanks for these comments! I think you’re totally right that nothing in these experiments claims to comment on the issue of presupposition and semantic entailment (so there’s no evidence here against the hypothesis that uses of 'know' presuppose, rather than entail, that the propositional complement is true). That surely remains an alternative explanation of these data that the non-factive person could evoke. Though given the strong revisionist aims here, the burden of proof seems to be on the supporter of the existence of a non-factive folk concept to (i) provide some solid evidence that holds up experimentally that non-philosophers (well, in my studies English-speaking, 35 y/o Americans, at least!) have a concept of knowledge that is non-factive (say, in the face of protagonist projection concerns or whatever else) and then once that is done, do this further thing (ii) showing its at the level presupposition. It's not clear that there's been sufficient empirical evidence for these claims.

The latter thing seems more difficult to test, though at least one quick thing we could try is to see whether it’s possible for someone to know propositions that couldn’t possibly be true.

About the contrast issue, I’m pretty interested to know the effect different answer choices in the DVs might be having on people’s responses here. So I’ll try going ahead and using qualitative measures like you suggest (this has also been suggested to me by Eric Schwitzgebel). My prediction FWIW is that we’ll still get highly projectionist answers. In the paper (and above in response to Eddy), I also mention one follow-up experiment that uses slightly different choices (‘Which do you think best describes Agent Smith?’…‘Agent Smith thought he Φ the bomb was homemade’ or ‘Agent Smith Φ that the bomb was homemade’). The result of that study was that people ended up giving more non-factive answers for ‘believes’ as the previous studies across true/false complements. So that’s at least some evidence that the contrast, as far as using ‘really’ is concerned has some effect here (though this particular one at least didn’t seem to be something that was affecting people’s projectionist judgments about ‘know’).

About one of your specific examples, I think people would tend to say that everyone ‘really knew’ (i.e. they weren't pretending) more than ‘they were pretending’ about the ulcers simply for pragmatic reasons (i.e. it’s really clear that whatever they were doing they weren’t pretending). Just for fun, I'll mention that if you watch Marshall’s Nobel talk on this (I’ve actually heard some of this talk at OSU recently last year) you get a sense of just how genuinely convinced the medical community was that they were right, and ironically, how horrified Marshall would be if he read that people think that that's what counts as knowledge!

[harvey brockman]

Hi Wesley,

A few interesting phrases from sentences at the end of the first paragraph of your post, and then at the beginning of the second: "he was absolutely right"; "we now know"; "did the doctors working before Marshall really know" -- How do you mean the "really" to be functioning in this latter sentence? If its function is to mark a kind of historically-developing CERTAINTY (regarding, in this case, a causal explanation of peptic ulcer), would it be your suggestion that we say: "In the pre-Marshall past, doctors 'merely knew' (or just 'knew') that stress was the cause of peptic ulcer, whereas today we 'really know' the cause to be H. pylori"? And with "he was absolutely right" is it your further suggestion that Marshall's explanation has "passed over" (maybe there's a better phrase for this!) from the "empirically-true" (so potentially revisable) to the a priori "logically-true" realm (that is, the Marshall explanation isn't just likely not to be over-turned, but it can make no sense of talking about its "being over-turned")?
Cheers, hb