Tuesday, April 14, 2020

The Epistemic Puzzle of Rashomon

Rashomon Gate - from the movie

[What follows are some thoughts on the movie Rashomon. These were prepared for a class I teach on critical thinking in law. If you haven't seen the movie, you can watch it for free, online, here.]

Rashomon is one the great films of the 20th century. Directed by Akira Kurosawa and released in 1950, it is a film that is perhaps known more by reputation and allusion than by actual viewing. The central plot device — multiple characters recounting the same events all with slight but significant variations — has passed into the popular imagination. So much so that the word “Rashomon” is now synonymous with the kind of epistemic problem that arises from conflicting testimony. This problem is faced in many disciplines, perhaps most notably in law.

I recently watched the movie Rashomon with one of my classes. I did so as an exercise in critical thinking. I wanted to use the movie as a way to easily illustrate the problem that we often face when dealing with conflicting testimony from different witnesses. I wanted students to approach the movie as an epistemological puzzle in which they had to use their critical faculties to figure out which narrative was most likely to be true.

I appreciate that some people will see this exercise as completely wrongheaded. They think the movie is intended to present a kind of epistemological nihilism. The point is not to ask which story is true but, rather, to realise that the ’Truth’ is elusive. Perhaps there is no such thing. Donald Richie, for example, in one of his essays about the movie suggests that this is the major theme of all of Kurosawa’s work: the world is an illusion that we construct through our interpretations of it.

I’m not sure what the correct analysis of the movie is. I suspect, like many artistic works, the movie allows for multiple interpretations and I think the ‘epistemological puzzle’ interpretation is a valid one. The movie is asking us to understand different errors and deceptions that might underlie people’s testimony about the world. If we reflect on these errors and deceptions we might be able to get closer to the truth.

In what follows, I want to do three things. First, I want to explain, in abstract terms, the basic epistemological problems that arise from all forms of testimony. Second, I want to use this abstract framework to explain the epistemological puzzle at the heart of Rashomon. And third, I want to present some of my own thoughts on how we might resolve that puzzle. I won’t, of course, claim to present a definitive solution to that puzzle. Part of the fun of Rashomon is leaving this as an exercise for each individual. Nevertheless, I think there are better and worse ways to analyse the different narratives within the movie.

1. The Epistemological Puzzle of Testimony
Testimony is what other people tell us about the world. It comes in different forms. In law we draw distinctions between ‘eyewitness testimony’, which is what happens when someone tells us about an event that they witnessed (or participated in) in the past, and ‘expert testimony’ which is what happens when an expert (e.g. scientist) tells us about some discovery they made or theory they have about how we should understand some issue that is relevant to a particular legal trial. Either way, the key feature of testimony is that it is information about the world that comes to us through other people. It is not something that we experience or understand directly, for ourselves.

We rely on testimony all the time. In fact, in many cases we are quite credulous about testimony. If someone tells me that they just saw a car accident down the road then, nine times out of ten, I will believe them. That said, there are times when our credulity lets us down. Perhaps the person is high on some psychedelic drug and so imagined the accident; perhaps they are lying for some reason. If we want to be reasonable and rational about our reliance on testimony we need to think carefully about the ways in which it can be misleading and modulate our credulity accordingly.

There are two basic kinds of problem that can arise when we confront testimony, particularly eyewitness testimony concerning events that happened in the past (which is the kind of testimony at the heart of Rashomon):

Encoding Errors: These are errors that arise when the person is witnessing or participating in an event that they later recount from memory. That is to say, they are errors that arise when the person is encoding the events into their memories for later retrieval. There are many different forms that encoding errors can take. The person may not be a good position to see exactly what is going on; they might mishear or misinterpret what someone else is saying; they might be drunk or intoxicated; they might have only participated in or witnessed some of the events; they might have their own biases that lead them to interpret events in a particular way; and so on. All of these encoding errors prevent their testimony from revealing the truth to us.

Decoding Errors: These are errors that arise when the person is recounting the events from memory. Again these errors can take different forms. We are told, repeatedly, by memory scientists that our brains are not like video recorders. We do not ‘store’ perfect records of past events for later retrieval. Memory is, rather, reconstructive. We reconstruct past events based on present knowledge, capacities and biases. So, for example, someone might have forgotten exactly what happened in the past and fill in the details in a way that seems plausible to them; or they might be suffering from some mental impairment in the present that causes them to distort the past. Most dramatically and obviously, they might deceive us as to what happened in the past because it suits their own interests. Sometimes this deception might be active and intentional; other times it might be inadvertent and arise because the person has such a distorted self-image or view of the world that they delude themselves into thinking that what they are saying is true.

Whenever we hear someone’s testimony we need to think about whether what they are saying is susceptible to these possible errors. As we shall see below, figuring out the potential decoding errors — particularly the motivations for deception — might be the key to unravelling the epistemological puzzle in Rashomon.

Is there any general test we can apply to testimony that will enable us to figure out whether it is true or not? Perhaps. The modern view is that we should apply Bayes Theorem to all evidence, including testimony. This means that we should weigh the probability that the testimony is true against the background probability that the event being testified to actually occurred. This can be useful but it is important not to mislead yourself into thinking that we always have reliable quantitative estimates for those probabilities. I personally think that David Hume got the basic picture right in his essay ‘Of Miracles’. That essay is famous for presenting a sceptical view about the occurrence of religious miracles, but in presenting this view Hume tackles the problem of testimony and proposes a simple two-part test we can apply when deciding whether or not to believe in it.

I outlined this test in some of my previous writings, relying on Robert Fogelin’s interpretation of Hume. Hume’s test does, roughly, follow the structure of Bayes’s Theorem but does so in qualitative, descriptive terms and not mathematical terms. What Hume says is that when assessing testimony we should first apply a ‘direct test’, which focuses on how credible the testimony is, and then a ‘reverse test’ which focuses on how probable the event being testified to is, given our prior knowledge. That works as follows (some of the language used in this test is directly from Hume's essay):

Direct Test: Testimonial evidence for some event X is generally more reliable (i.e. more likely to be true) if it exemplifies the following (non-exhaustive) list of properties: 

  • (i) There are many witnesses, not few. 

  • (ii) The witnesses concur with one another rather than contradict one another. 

  • (iii) The witnesses are of unimpeachable, rather than of doubtful character. 

  • (iv) The witnesses are disinterested, not interested, parties. 

  • (v) The witnesses present their testimony in measured tones of confidence, rather than with hesitation or too violent asseveration.

Reverse Test: The probability-raising potential of reliable testimonial evidence for X must be assessed relative to the prior probability of X. If X was highly improbable, then the effect of reliable testimony is proportionally diminished.

The importance of the two parts of this test varies depending on the context. In Hume’s case, where the focus was on the plausibility of religious miracles, most of the emphasis was on the Reverse Test. Indeed, Hume’s main insight into the debate about miracles was his observation that even if the testimony provided was credible it still not did make it rational to believe in the historical occurrence miracles, given their low prior probability. In the case of Rashomon, the direct test really becomes the more important one. The challenge of Rashomon is that we have contradictory testimony coming from interested, not disinterested, parties.

2. The Epistemic Puzzle of Rashomon
Let’s turn now to the plot of Rashomon and the epistemic puzzle it presents (there will be spoilers from here on out). Rashomon starts in the rain. Three people — the commoner, the priest and the woodcutter — huddle under the ruined Rashomon gate for shelter. Earlier that day, the woodcutter and the priest watched a trial concerning the murder of a man in a grove in a wood. The murder took place three days ago. At trial, three witnesses gave evidence: a bandit (Tajōmaru - the only named character), the murdered man’s wife, and the murdered man himself (via a medium). The trial perplexed the woodcutter and the priest because each of the witnesses presented a very different version of events. They tell the commoner about the trial and their own role in it. It turns out both had tangentially witnessed events related to the murder. The woodcutter discovered the body and the priest had witnessed the man and his wife journeying through the wood several days before. It later turns out that the woodcutter also witnessed the murder, but this is not revealed until close to the end of the movie.

From the start then, Rashomon is structured in an unusual way. It’s not just that we get different witness accounts of the murder; it’s that these witness accounts are revealed to us second-hand via the woodcutter and the priest and, in one case, third-hand because a medium provides the testimony from the deceased husband. Only the woodcutter provides direct eyewitness testimony at the very end. Thus, the reality of what actually happened in the grove is filtered through several layers of  potentially misleading testimony. The viewer has to peer through these layers of testimony, as is illustrated below.

You could mull over this curious structure for a long time. Since the story comes at us through these layers of testimony it makes it more difficult to get at the truth. You don’t just have to think about the testimonial errors that might have been made by the original witnesses to the events; you also have to think about the testimonial errors that the people recounting the testimony of the original witnesses might be making.

But adding that extra layer of critical analysis might be going too far. Initially, I think it makes most sense to assume that the witness testimony is recounted by the priest and the woodcutter in a reasonably accurate way. The puzzle then is in trying to reconcile the four different accounts of the murder. Doubling back and considering whether the priest and the woodcutter might be misremembering or deceiving us with respect to the testimony of the three witnesses at trial might be fun, after we have attempted to resolve the discrepancies between the four different versions of events.

So what are the four different versions of events? Let’s start with the testimony from the bandit, Tajōmaru. He was sleeping under a tree in the wood when the husband and wife passed by. He runs after them and then lures the husband away with the promise of showing him some weapons. He tricks the husband and ties him up. He then proceeds to rape (or ‘seduce’ - if you believe what he is saying) the wife. He begs her to leave her husband and come with him. She tells him he must kill the husband first to preserve her honour (she cannot be tied to two men at the same time). He and the husband fight nobly, crossing swords 23 times. He kills the husband with his sword. The wife runs away and so does the bandit, taking the husband’s horse but not a valuable dagger that was also in his possession (the dagger plays a crucial role in subsequent narratives). The bandit drinks some poisoned water and is later discovered by a policeman.

Next we get the wife’s version of events. She takes up the story after the rape. She tells the court that the bandit runs away after the rape and she is left with the husband. She unties him. He is contemptuous of her. She asks him to kill her with the precious dagger. He does nothing. She goes into a fugue holding the dagger before her and threatening the husband. She then blacks out. When she wakes up the husband is dead with the dagger sticking out of him. The implication, though not clearly stated, is that she may have killed the husband in her fugue state.

The next story comes from the husband himself. You have to suspend disbelief at this point and assume that the medium at the trial really is channeling the husband’s voice. If you don’t do that then you may as well discount this version of the story in its entirety. He tells us that he is tied up and witnesses the rape of his wife. The bandit then pleads with the wife to come away with him. She agrees, but only if the bandit kills the husband. The bandit rejects this as dishonourable. The wife runs away. The bandit tries to find her but fails. The husband is left alone and, in his grief and shame, he kills himself with the dagger. Later, someone takes the dagger from his body but he cannot tell who it is.

Finally, we get the woodcutter’s version of the story. He comes upon the scene after the rape. He witnesses the bandit begging the wife to come with him. She will not. She argues that the men have to decide her fate by fighting over her. The husband doesn’t want to, but the bandit is disgusted by the husband’s attitude and they reluctantly fight each other. The fight is not noble; it is farcical. They do not cross swords 23 times. After a comedy of errors, the bandit kills the husband with a sword.

This prosaic summary does not do justice to the way in which the stories are told in the movie. There are some interesting aspects to performances by the actors that could affect our interpretation. For example, the bandit is a little bit eccentric and quite proud in the way he recounts events; the wife goes through different extremes of emotion, displaying quite an odd attitude to events at times. The medium is pretty creepy and ominous but she presents the husband in a somewhat pitiful light. And the woodcutter is sheepish and meek. It’s worth watching the movie to appreciate the different performances.

3. Resolving the Epistemic Puzzle
So who is telling the truth? Clearly, the four narratives present very different and irreconcilable versions of events. There are, however, some common points of agreement. All seem to agree that the bandit rapes the wife (though he suggests that she ultimately fell for him), all agree that the husband was tied up, all agree that the husband was, ultimately killed. They then differ quite substantially on the details. Two of the witnesses (the bandit and the woodcutter) suggest that the bandit killed the husband in a duel; the husband claims to have killed himself; and the wife is unclear but hints that she may have killed the husband. There is also disagreement about the murder weapon. Two of the narratives suggest the husband was killed with a sword; two suggest that he was killed with a dagger.

Is there anyway to sort through this mess and get to the truth? Well, let’s start with sketching the possible answers to this question. There are only three:

Possibility 1: None of the witnesses presents us with anything that can get us close to the truth.
Possibility 2: One of the witnesses is telling the truth and all the others are lying.
Possibility 3: Some or all of the witnesses present part of the truth and it is possible to triangulate on what really happened by carefully selecting relevant portions of the testimony.

The first of these options seems unlikely given that there is some agreement on what happened across the multiple witnesses. In addition to this, the stories told do seem to exhaust some of the relevant possibilities when it comes to who the murderer might have been. The only possibility not covered is that the woodcutter might have murdered the husband (perhaps to get his precious dagger). Consequently, it seems more plausible to suppose that either one of the witnesses is telling the truth or there is partial truth across the different stories and that if we interpret it carefully this might be enough for us to get close to what actually happened. Let’s work with that possibility for now.

We can start to sort fact from fiction by thinking about the possible errors that the witnesses might be making. As noted above, there are two basic types of error: encoding errors and decoding errors. It’s certainly plausible to think that there are some encoding errors affecting the testimony in Rashomon. The woodcutter, for example, seems to have only witnessed some of the events and the wife doesn’t remember a crucial part of what happened. That said, decoding errors seem to be the more likely cause of the discrepancies between the stories. When you look at it, it’s not just the case that the witnesses misremember or misunderstand events; they have very distinctive ‘slants’ on what happened. Indeed, their narratives typically paint some of the participants in a favourable light and others in a less favourable light. What’s going on?

One of the major themes in Rashomon is that of honour. The movie depicts a Japanese society that is very much bound up with a culture of honour. The bandit makes it a point to suggest that he and the husband fought honourably; the wife cares about her honour as a ‘defiled’ woman; the husband is a Samurai and, according to his own testimony, appears to have killed himself out of a sense of shame; the woodcutter may, if we believe what the commoner suggests at the end of the movie, have provided his testimony in order to cover up his own dishonour in stealing a dagger from the corpse of the dead man. Given all this, it seems plausible to suggest that the characters are retelling events in a way that serves their perceptions of honour.

Another major theme in the movie is that of love and loyalty. The bandit appears to have fallen in love with the wife (and he suggests that she has fallen in love with him); the wife cares about losing the respect and love of her husband; the husband seems heartbroken at having lost the love of his wife. Love can make people do irrational things and it may well be affecting the testimony we hear. The characters could be covering for one another in order to protect the people they love.

In short, each of the witnesses could be deceiving us, for different reasons, associated with honour and love. We can try to work out the truth by figuring out the motivations for deception and seeing which story is most plausible after we have accounted for the possible lies. I’m going to suggest one way of doing this but leave it open to readers to suggest alternative analyses.

Let’s start with the bandit. Why might he lie and can we believe his story? The obvious answer is that he might be lying to protect the wife. It could be that she really killed the husband but he is in love with her and wants to protect her. He certainly professes a strong affection for her in his testimony. He may also reason that his lie is likely to be believed because he is a notorious bandit who would commit a crime like this, and he has less to lose for his lie since he is presumably wanted for other crimes that may attract the death penalty. (A more subtle interpretation is that he may be in league with the wife. They want to sow enough reasonable doubt in the courtroom so that neither of them gets convicted.)

What about the wife? Why might she lie and can we believe her story? The obvious answer is that she might be lying to protect either the husband or the bandit. She wants to protect the husband from the shame of suicide; or the bandit from the charge of murder. That said, she tries hard to paint herself in a sympathetic light in her testimony (as a victim) and she conveniently blacks out at the moment of the murder. This may be a hedge on her part to make it less likely that she will be convicted. There may also be a gendered aspect to this insofar as women might be less likely/more likely to be believed or treated sympathetically in the society at the time. I don’t know enough about the history of Japanese society to comment meaningfully on this.

Turning our attention to the husband, why might helie? Well, he might be lying to protect the wife. Perhaps he is still in love with her, despite her rejection of him (if you believe his story), and even though he knows she killed him, he wants to take the blame away from her. He may also feel guilty about how he treated her (if you believe her story) and has nothing to lose since he is already dead. There is also a possibility that he may lie to protect the bandit if, as seems to be the case from his story, the bandit has won his respect. But this strikes me as being less likely given that the bandit has humiliated him in other ways.

Finally, why might the woodcutter lie? The obvious answer — suggested to us by the commoner at the end of the movie — is that the woodcutter lies to conceal the fact that he stole the dagger from the corpse (the theft of the dagger is hinted at in the husband’s testimony). He knows the dagger was the murder weapon and needs to tell a story that makes the sword the murder weapon. He already heard the bandit’s testimony at trial and so gives a slight variation on it to make it sound plausible and cover his own tracks.

You might think this analysis does little to shed light on the matter. And in a sense this is true. We see now that each character has a plausible motive to lie and it is possible to construct an argument in favour of each as a potential killer. That said, my own take on it is that we can discount some stories more than others. First, I think we can discount the woodcutter’s story. He had the benefit of hearing all the previous testimony and could easily have concocted a version of events that helped to conceal the fact that he stole the dagger from the corpse. He doesn’t deny that he stole the dagger when challenged by the commoner. This has the knock-on effect of undermining the bandit’s story and making the dagger the likely murder weapon. This strikes me as plausible since the bandit’s story is a bit too self-serving at times, e.g. the suggestion that the wife fell in love with him and that the duel was a noble one. That said it is not entirely satisfying to discount the bandit’s story since he does have a lot to lose by confessing to the murder, even if he is a notorious criminal with other potential convictions against him.

If you buy that, then it’s either the case that the wife murdered the husband or the husband killed himself. Of these two narratives I tend to favour the view that the wife murdered the husband. She has some motivation for doing so (he didn’t protect her from the bandit and he scorned her afterwards) and her blacking out at the time of the murder is quite convenient. Also, given how she is portrayed across the different stories, including her own, it doesn’t make sense to me to suppose that she is lying because of any great love she has for either the bandit or the husband. The husband’s testimony I tend to discount. He may be feeling some kind of guilt for how he treated his wife and wants to protect her. It may also just be worth discounting his testimony entirely because of its dubious origin.

For what it is worth, when I ran this exercise with one of my classes this tended to be the favoured interpretation. This was before I presented any interpretation of my own. That said, it is not an aesthetically pleasing interpretation. The wife is, after all, the victim of a crime and suggesting that she is the murderer plays into certain stereotypes about vengeful women that doesn’t sit well with modern audiences. Furthermore, even though I think this is a plausible reading of the events, it is not one that would have sufficient credibility to cross the burden of proof in a criminal trial: beyond a reasonable doubt. In other words, I wouldn’t convict anyone on the basis of the evidence provided in the movie nor my interpretation of it. In some ways, that's why the movie is such a good one. It challenges us to reconsider the weight we place on testimony in certain contexts.

What do you think? Is there an interpretation of events that you favour?

Further Reading


  1. This comment has been removed by a blog administrator.

    1. This comment has been removed by a blog administrator.

  2. Epistemological Nihilism fiercely defended in debate https://youtu.be/sz5jTaMD3IQ

    Epistemic nihilism is a rational construct, and as such it can only come second in our determination as to what we should do next. Rationality is a tool we have available to us to achieve what we want, it does not decide what we want.

    Rationality is at the surface of our brain's activity. Most of what we understand about the world is intuitive. Intuition is essentially logical, but it works from our representation of the world. Some of that is essentially based on past perception, but, crucially, some is determined by our rational musings. Spend enough time convincing yourself that the world is a simulation, and your brain will take it on board that the world is a simulation, and from there your intuitions will logically reflect this assumption. We can convince ourselves of anything if we try hard enough.