Containment protocol: None. Words can’t hurt you. Words aren’t real. Philosophical ideas don’t affect reality. You won’t notice any changes after reading this. You won’t find yourself, in conversation and in your own thoughts, ceasing to reach for institutionally certified sources of aggregate information of universal applicability. You won’t find yourself reaching instead for personal anecdotes or any tangentially-related connection to your own experience. You won’t gradually cease to expect that positive knowledge exists for new questions you encounter. You won’t notice the words squirming beneath your feet with their sense gelatinized, like cobblestones turned to jellyfish. “Hermeneutic” doesn’t count.
Description: “Ignorance, A Skilled Practice” is a guest blog post written by a literal banana. The banana’s tiny cartoon arms barely span the keyboard, and as a result the banana is only able to press one key at a time with each hand or foot. The blog post is offered here as an example of what bananas can accomplish when given proper access to technology.
Since Harry Frankfurt’s essay taking a rather negative view of bullshit as a category (“On Bullshit,” 1986), some philosophers have attempted to redeem “bullshit” from its negative connotations (see, e.g., Joshua Wakeham, “Bullshit as a Problem of Social Epistemology,” 2017).
What follows is my attempt to articulate the subset of bullshit that I think is bad, because I think most bullshit is good. The skillful practice of ignorance cannot be reduced to the avoidance of bullshit. It would be self-defeating to deprive oneself of bullshit completely, for bullshit is the air we breathe, and we could not learn or accomplish anything without it. Here I try to bring to salience a particular quality of a particular kind of knowledge, such that in recognizing it, much harmful bullshit can be appropriately classified, and a state of skillful ignorance can be maintained.
The Puzzle of Indexicality
This whole thing is about indexicality, and there’s no way around the word. I tried.
Unfortunately there is no such thing as indexicality, at least in the sense of there being a generally true and proper definition independent of context. (That’s the point.) There is only indexicality for this particular context – this blog post written by a banana – and possibly contexts a little bit like it. Those things are true for any word, but indexicality is special, because it is about those things.
“Indexical” is a word almost perfectly calculated to sap the morale of the reader and annihilate interest. I have seriously considered replacing it with the word “pointing,” used as a descriptor. Indexical statements are pointing statements: “I prefer this one,” “don’t do that,” “I made it for you.” These sentences have no particular meaning without reference to the situation in which they are produced. Physical pointing, as with an index finger or the lips or chin, may or may not accompany an indexical expression, but there is a sort of pointing-to-the-situation that occurs in all cases. I’ve decided to keep “indexical” for clarity, but keep in mind that indexical means pointing, in a literal and then in an extended, figurative way.
In the linguistic sense, an expression is indexical if it refers by necessity to some particular state of affairs. “This guy arrived just now” depends on the person indicated and the time of speaking; it is highly indexical. Compare “The Prime Minister arrived at 5:15 p.m.” This is less indexical, but notice that the identity of the Prime Minister depends on the country, and of course we don’t know anything about the circumstances or place of arrival from the text: 5:15 p.m., but in what time zone?
Extremely non-indexical expressions often appear as health or science headlines. These are pretty much the opposite of indexicality:
Stanford Researchers: Average Human Body Temperature Dropping
How Puberty, Pregnancy And Perimenopause Impact Women’s Mental Health
Why is air pollution so harmful? DNA may hold the answer
Predatory-journal papers have little scientific impact
Can a healthy diet reduce your risk of hearing loss? Here’s what the research says
Notice that these refer to people in general, and vague concepts in general. They take the form of objective knowledge that is true in general, for all cases, globally, universally. They “see through” to the ultimate truth of matters, unsullied by the messy realities of particular people and situations. The kind of knowledge that non-indexical statements presume to convey is timeless, and describes all of humanity or the world in general. Indexical knowledge, on the other hand, refers to specific situations, times, people, and interactions. It does not purport to apply timelessly, in general, or to all people.
There is a certain group, or school of thought, that is very interested in indexical knowledge, in a wider and more metaphoric sense than the purely linguistic sense. Not just statements, but looks, gestures, pointing with fingers and objects, and many more methods and strategies, can reveal knowledge held within contexts. This group believes that non-indexical or global knowledge is possible, but rare, and celebrates the accomplishments of science and engineering in locating information that is true across many contexts, and its twin accomplishment of maintaining stable contexts in which beneficial things happen with regularity. This group does not so much celebrate the “inexact sciences” for such accomplishments, however. The gulf between indexical and non-indexical is sometimes crossed in the exact sciences, through coordinated effort and much in-the-moment situated problem solving. (The Periodic Table of the Elements is a good, ongoing example of this.) However, the inexact sciences (the social or soft sciences) have had much less luck in their attempts to discover the universal truths of social reality, according to this mysterious group that I have not yet named. “Indexicality” is this group’s specialist jargon.
And there is further bad news, which is that this mysterious group is known by the enormous word “ethnomethodology.” (The good news is that “ethnomethodology” is the last terrible jargon word in this essay, unless you count one Peter Unger coinage that’s not even that bad.) “Ethnomethodology” scans to “Angels We Have Heard On High,” “Cult of Personality,” and “Camptown Races,” but please use this information wisely.
If you look at the etymology and various senses of the word “index,” brought together in an awkward family reunion on Wiktionary, you quickly see that it’s about pointing: pointing things out, indicating, separating some aspect of reality out from its other aspects. There is a great deal of ethnomethodological study of the act of pointing itself, which I think is useful for understanding indexicality in terms of its primordial metaphor. Sometimes pointing can be very complex, as when an archaeology student gets the attention of his supervisor (verifying her gaze direction), then uses his trowel to point out a feature on a paper map, then points out the corresponding feature in the “territory” of the dig site. In each case, his gestures, eye gazing, language, etc. are methods he uses to separate out aspects of his context. In pointing, he is not making any general or global claims. He is directing attention to something specific in context. That’s indexicality in a rather pure form. (Charles Goodwin, ”Pointing as Situated Practice”, 2003)
We’ll take a few more passes at the concept. The strongest expression of indexicality I have come across – the thing that solidified the matter in my mind rather like a flash – is David Chapman’s account (ms. in preparation):
We use words as tools to get things done; and to get things done, we improvise, making use of whatever materials are ready to hand. If you want to whack a piece of sheet metal to bend it, and don’t know or care what the “right” tool is (if there even is one), you might take a quick look around the garage, grab a large screwdriver at the “wrong” end, and hit the target with its hard rubber handle.
Words go the same way. Almost any word can be used to mean almost anything, in some context. You could play this as a challenge game… How about “The eggplant is a straw hat, and the spinach is yelling about politics”?
We’re in the kitchen of a vegetarian restaurant. A table’s entrees are ready, and the server who took the order is explaining to the one who will deliver the meal which diner gets which dish. One customer’s flamboyant straw hat is a salient, unambiguous identifying feature; you can see it all the way across the room. The other probably needs to turn up a hearing aid; you can hear their opinions about cultural appropriation all the way across the room.
We naturally use whatever word is unambiguous and obvious in context to describe something for which “literal, objective” terms would be ambiguous, unavailable, or inconvenient. It doesn’t matter what “the word means”; it matters that the customer gets their moussaka.
(Underlined emphasis mine.)
When you think of “spinach” or “eggplant” in the context of a tweet, you probably don’t think of it as an index for people (unless perhaps you know a banana and attribute personhood to bananas). But in the context of a server at a restaurant, it is natural to map food to people. Eggplants, like bananas, can yell about politics.
Words Don’t Work That Way
Keep the yelling eggplants in mind as we pass into Peter Unger’s delightful but troubling “Why There Are No People,” (1979) which can be faithfully if rudely summarized as “How Can People Be Real If Our Words Aren’t Real.” Unger is an important figure in the philosophy of ignorance, but here I am mainly interested in presenting his model of word acquisition, and the sorites-type problems that fall out of it, to elucidate the kind of meaning that is the opposite of indexical (which non-indexical meaning, by the way, Unger proves cannot exist).
Consider David Chapman’s statement: “Almost any word can be used to mean almost anything, in some context.” Unger takes this as a basically mathematical challenge to be proven, rather than as a party game.
Unger’s word acquisition model is this: you are presented with an unfamiliar object, and told its name (he offers “nacknick”). You are given to understand that things that are a little different from this example, in particular ways, will still qualify under the same name. Anything that’s pretty close to a nacknick is also a nacknick. Further, and importantly, objects that are a little different from these as-yet-hypothetical new qualifiers, the nacknicks that are out there but that you’ve never seen, but still pretty similar in particular ways, will themselves qualify as nacknicks. (See Figure 1.)
However, objects that are sufficiently different from the exemplar, the original nacknick as presented, are NOT nacknicks. That is part of the definition: some things are simply not nacknicks. Unger’s term for this is the final unfortunate vocabulary hurdle in this blog post, and it is “vague discriminative expressions.” They are vague (like “tall men”) in that they are imprecise about what they might include or exclude, yet they are discriminative in that they are offered to, and purport to, discriminate qualifiers from non-qualifiers.
Unger performs a sort of sorites maneuver, or perhaps Banach-Tarski for words, on various dimensions of vague discrimination. What it amounts to is that, in a global sense, in a logical sense, a word cannot be both “open” to describing minutely-different circumstances from its existing cases and “closed” in the sense of definitely excluding some cases. For however tiny your definition of a “minute difference,” with enough of them put together, a word can transit any distance in meaning space. “Eggplant” can be a person. “Person” is no better off, as hinted in the title of the paper. (I won’t say much about the sorites maneuver that Unger performs on people to prove that they can’t exist, except that it is cellwise.)
Here’s how Unger is correct that there are no “people”: words cannot have global meaning. They can only discriminate within a context.
The Twitter Heliopause
Think about what happens when a tweet leaves its zone of context, perhaps due to popularity. If the tweet is sarcastic, its content contrasting with the account holder’s usual views, the sarcasm will be visible to those with that context – the followers or in-group. But if the tweet escapes its context, it may be misinterpreted as sincere – the indexical information of sarcasm lost. “Errors” may be pointed out which make no sense in the tweet’s original context, but make sense if the tweet is viewed as a global statement in a context vacuum. People encounter the tweet and interpret it according to their own context, and put it to their own uses. The tweet began its life as all statements do – as an indexical expression, produced in a context. It can then have its context sheared off and be entered in the global knowledge game as a general statement, vulnerable to incredibly simplistic counterexamples. It can be repurposed entirely and offered in a new specific context for new specific purposes. The text of the tweet itself carries little of its context. Sometimes the replies from early in its career offer context, but if it gets popular enough, it will find itself dragging around a low-context colony of kelp and jellyfish tentacles.
I’m not sure what the proper antonym for “indexical” is. Garfinkel distinguishes indexical from “objective” and “non-indexical,” and here I have used global, general, universal, etc. There’s no good word for it and I don’t propose one. Words aren’t real.
What I do think is real is something that I will at first call a belief in global, general, objective, non-indexical meaning. And by “belief” I mean something like an aura of ordinariness that surrounds general information, the ongoing expectation that objective knowledge is available for any encountered topic. And I don’t even necessarily mean a psychological expectation that such information exists, or a psychological willingness to trust the probity of that kind of information in ordinary life. I think that global knowledge is a game (in the Wittgenstein sense explored in Philosophical Investigations). The global knowledge game is taken seriously in some ways and not in others, and probably doesn’t comprise one single game, but many interlocking games. Such games share the trait that they pretend to be engaged in a universal knowledge game.
There are particular rules and standards for every particular local scene. There may be rules of evidence (as in a courtroom) or logical fallacies (in rhetoric), rules that can be cited by game participants to negotiate bracketing out some information as “not relevant” or otherwise not appropriate.
The Global Knowledge Game
To illustrate that global knowledge is a game, consider a story about Alexander Luria, who studied illiterate Russian peasants and their semi-literate children. Consider especially this version of the story, prepared in the 1970s to provide morale and context to reading teachers (John Guthrie, 1977). Essentially, Luria discovered that the illiterate, unschooled peasants were highly resistant to syllogisms and word games. The adult peasants would only answer questions based on their own knowledge, and stubbornly refused to make deductions from given premises. “All bears are white where it is snowy. It is snowy in Nova Zembla. What color are the bears in Nova Zembla?” “I don’t know, I have never been to Nova Zembla.” Children with only a year or two of education, however, were easily able to engage in such abstract reasoning. They quickly answered the syllogisms and drew inferences from hypothetical facts outside of their own observation.
In this story, I argue, Luria’s peasants are indexical geniuses, who refuse to engage in unproven syllogistic games. They are not interested in a global, universal game. Their children, however, are easily introduced to this game by the process of schooling and literacy.
Interestingly, a more recent group of researchers claim that illiterate people do fine at making inferences against experience, if the context is given as a distant planet (Dias et al., 2005). I am not offering this as true, but as a story about how expecting people to operate in the “global knowledge game” might portray them as stupider than they really are, if they simply choose not to play in that game. This is to segue into the next hermeneutic pass, in which we are told that the hype surrounding “cognitive bias” is really a sort of science magic trick, an illusion designed to portray indexical geniuses, like Luria’s peasants and ourselves, as global fools.
The paper is “The Bias Bias in Behavioral Economics,” by Gerd Gigerenzer (2018). If you, like me, have ever been fascinated by cognitive bias research, this is a brutal paper to come to terms with. Gigerenzer examines several purported biases in what I would call analytic reasoning or the global knowledge game, and finds explanations for these purported biases in the indexical reality of humans.
For instance, some apparent “biases” that people display about probability are not actually errors. For the small (and in most cases, merely finite) samples that reality has to offer, people’s “biased” intuitions are more accurate than a “globally correct” answer would be (that is, the correct answer if the sample were infinite). In tossing fair coins, people tend to intuit that irregular strings are more probable than more regular strings (e.g. that HHHT is more probable than HHHH in a sequence of coin flips). This simple intuition can’t be correct, though, because given infinite coin flips, each string is as likely as any other, and if the sequence is only four flips, after HHH, each outcome is equally likely. But for small, finite numbers of flips greater than the string length, Gigerenzer argues, it is the human intuition that is correct, not the naive global solution: HHHT does take less time to show up than HHHH in repeated simulations, and is more commonly encountered in small samples. To drive home his point, he offers a bet:
If you are still not convinced, try this bet (Hahn and Warren, 2010), which I will call the law-of-small-numbers bet:
You flip a fair coin 20 times. If this sequence contains at least one HHHH, I pay you $100. If it contains at least one HHHT, you pay me $100. If it contains neither, nobody wins.
More broadly, cognitive bias proponents find fault with their subjects for treating “logically equivalent” language statements as having different meanings, when context reveals that these “logically irrelevant” cues frequently do reveal rich meaning in practice. For instance, people react differently to the “same” information presented negatively vs. positively (10% likelihood of death vs. 90% likelihood of survival). Cognitive bias proponents frame this as an error, but Gigerenzer argues that when people make this “error,” they are making use of meaningful context that a “bias-free” robot would miss.
I am reminded of the time I was reading Rousseau’s entry on Wikipedia, and the article mentioned that he had abandoned his children, and that this fact had been used in “ad hominem” attacks against him, since he wrote important treatises on child-rearing. I think it can be seen from this example how both the “logical fallacy” project and the “cognitive bias” project are bracketing projects – negotiating the bracketing-off of parts of reality, as in a court room, in order to play a particular language game. In ethnomethodological terms, rules don’t define a game. Rather, they are resources that players of the game can use in order to make sense of their situation. Referring to the Rousseau scandal as “ad hominem” is a move in the game, asserting that a particular rule applies, and hence that the move it refers to is invalid. By making fun of this move, I am making a move myself. (If it is still there, please do not edit it out – it’s too illustrative to be destroyed.)
To summarize the associations we have collected thus far for indexicality: indexicality is a linguistic concept building on the metaphor of pointing, meaning a statement whose meaning is specific to a situation, and whose meaning cannot adequately be reconstructed from text without importing aspects of the context. Indexicality is further extended in ethnomethodology to refer to knowledge and methods expressed and contained in particular contexts. The opposite of indexicality is the objective, general, universal, global word game. Science has celebrated many victories in wresting objective knowledge from indexical observations, but these victories are rare, and are the product of concerted group activity backstage, rather than a natural consequence of scientific ritual. Overconfidence in the global word game, especially in social sciences, threatens the production and appreciation of the genuine kind of indexical knowledge that humans are geniuses at producing and using. The framing of “cognitive bias” privileges the global knowledge game at the expense of a richly situated rationality (what Gigerenzer and others call “ecological rationality”).
Behind social science, there is the assumption that there must be a fact of the matter for various social science questions – a fact of the matter in a particular global, floating, timeless, objective sense. The problem is not that there is no fact of the matter in some particular case, though this is frequently true. The problem is the assumption that global knowledge can be found in every case, coupled with an easy certainty that one has found it when certain rituals are performed and certified.
Folktale, Metaphor, and the Construction of Non-Indexicality
I warn the reader that despite the polysyllabic suffering that has been endured thus far, things will actually get more unpleasant from here. Things will get more unpleasant not for vocabulary reasons but for subject matter reasons. The object-level topics to choose from in social science are often depressing to the degree that they are interesting, and the ones I present here are all moderately to severely depressing. However, these are the topics I have grappled with for the purpose of this investigation, and I don’t think that dealing with only wholesome, pleasant topics will improve the quality of my message, such as it is.
First, here is a story in the style of a folktale:
A baby elephant was training for the circus. His trainer led him around on a rope, and the rope was stout enough that the baby elephant could not get away. When the baby elephant grew up, he still believed that the rope was strong enough to hold him, and never tried to resist. Fire broke out one night when he was tied up, and he made no effort to escape, believing himself to be trapped by the thin rope. The elephant perished in the fire.
This is my rendition of a popular tale, one that you might encounter from a motivational speaker or addiction counselor. I have made it extra bleak so that its ridiculousness compared to reality is more visible than in the typical example, in which there is no fire. Is this story indexical? Does it purport to contain global knowledge? What about folktales, fairy tales, Aesop’s fables?
I think that stories like this exist outside of the global knowledge game. Proverbs and folktales, like rules, often exist in mutually contradictory forms; that’s because they are not a set of rules that govern from the top down, but metaphors to be picked up and used to illustrate concepts and elucidate privileged properties, or to function as attractors to collect and compare relevant examples within memory. They function like Douglas Hofstadter’s “Danny at the Grand Canyon” – a photo of his son as a toddler at the Grand Canyon, totally bored by the majestic canyon, examining some tiny rocks in a crouching position on the ground. Hofstadter notices that the photo/story “catches” similar situations with the same “gist” – people ignoring what’s around them and focusing on something unusual (given the context), for idiosyncratic or relatable reasons.
These stories aren’t “true,” and they are not told with the expectation that they be believed. (If a story is meant to be believed, or if its truth is up for debate, some folklorists would instead classify it as a “legend” rather than a fairy tale or folktale.)
Here’s an Aesop’s Fable illustrating the opposite point from the popular elephant story:
The King’s Son and the Painted Lion
A king, whose only son was fond of martial exercises, had a dream in which he was warned that his son would be killed by a lion. Afraid the dream should prove true, he built for his son a pleasant palace and adorned its walls for his amusement with all kinds of life-sized animals, among which was the picture of a lion. When the young Prince saw this, his grief at being thus confined burst out afresh, and, standing near the lion, he said: “O you most detestable of animals! through a lying dream of my father’s, which he saw in his sleep, I am shut up on your account in this palace as if I had been a girl: what shall I now do to you?” With these words he stretched out his hands toward a thorn-tree, meaning to cut a stick from its branches so that he might beat the lion. But one of the tree’s prickles pierced his finger and caused great pain and inflammation, so that the young Prince fell down in a fainting fit. A violent fever suddenly set in, from which he died not many days later.
We had better bear our troubles bravely than try to escape them.
Now consider a real elephant following a human on a thin rope. Is the elephant really stupid? Does the elephant “believe” that he’s not strong enough to get away, based on his childhood experiences? Would the elephant really stay trapped by the rope in the event of a fire or emergency? Or is the elephant engaged in a more complex, ongoing relationship with the human? The “learned helplessness” picture makes for a good story to illustrate a point in a talk, but it’s certainly an inaccurate and biased picture of a human-encountering elephant’s actual circumstances. The thin rope is offered only as a metaphor for something that might happen in human emotional life, not as a depiction of elephant behavior.
Unfortunately, in the field of psychology, it became very popular to construct gruesome animal metaphors for people to project human emotional and social ideas onto, instead of just writing animal fables or creepypastas. These have not been treated in the way that animal folktales are treated, but rather as scientifically certified bases for making inferences about humans.
In the 1960s, Martin Seligman and colleagues performed a series of investigations into the phenomenon they called “learned helplessness.” It is the same phenomenon illustrated in the elephant story, above. Previous researchers had found that when they subjected dogs to painful electric shocks, the dogs quickly learned to efficiently escape from the source of the shocks. Seligman and colleagues introduced a new complication into the metaphor: if dogs were first held immobile by researchers, tied up and in some cases given paralytic agents, and then administered extremely painful shocks lasting for 50 seconds each, then when the dogs were later freed from the immobilizing harness, they no longer tried to escape from the source of the shocks, which would require leaping over a barrier. In 1968, Seligman and colleagues further extended their metaphor with a mechanism to repair “learned helplessness.” (Seligman et al, The Alleviation of Learned Helplessness in the Dog, 1968) They tied four “mongrel dogs” up in harnesses and administered unavoidable, extremely painful 50-second shocks to them, as usual. (Try counting out 50 seconds for context if you like.)
The dogs demonstrated “learned helplessness” when freed from the restricting harnesses, as usual. However, when researchers verbally encouraged the newly freed dogs to jump over the barrier, one of them eventually responded. Three more of the dogs were cured of their “learned helplessness” by being dragged over the barrier, using two leashes per dog, by the researchers.
This research is a bummer, but I promise that it gets worse. The juicy animal metaphor was now certified as global knowledge. Now it could be applied to human situations with unassailable warrant, in the global knowledge game.
In 1979, Lenore Walker would adapt the dog metaphor for use in the context of battered women, in her book The Battered Woman. For example, on page 53, in a section titled “Stopping Learned Helplessness,” she says:
If battering behavior is maintained by perceptions of helplessness, can this syndrome be stopped? Turning back to the animal studies, we see that the dogs could only be taught to overcome their passivity by being dragged repeatedly out of the punishing situation and shown how to avoid the shock. Just as the dogs have helped us understand why battered women do not leave their violent situations voluntarily, perhaps they can also suggest ways the women can reverse being battered. A first step would seem to be to persuade the battered woman to leave the battering relationship or persuade the batterer to leave. This “dragging” may require help from outside, such as the dogs received from the researchers.
In a section on psychotherapy, on p. 239, she says:
The battered woman needs to recognize concrete steps she can take to improve her situation. Like Seligman’s dogs, discussed in Chapter 2, she must be dragged over her escape route numerous times before she can be expected to do it on her own.
Of course, many 21st century academic feminists are not thrilled with Walker’s characterization of the problem. But remember that Walker considered herself a feminist, and her work was lauded by feminist activists at the time. Learned helplessness is still a going concern in the field of psychology. The learned helplessness animal metaphor, and various facts that the metaphor gave life to, still live on in the official global word game.
Here we must introduce The Fact.
A Grain of Sand in Which to See Non-Indexicality
The Fact is just a little fact. It’s not a big deal, and probably nobody bases any serious policy decisions on it, although it is disseminated by policy advocates. If you search any term related to the domain of the fact on Google, Google will become concerned, and will offer you as the first result a hotline to call if you are personally having trouble in this domain. This semi-official hotline offers The Fact as a true fact, but it is not alone. CNN and The Guardian have both published articles claiming it is true. Different versions of The Fact are on offer in the United States, Canada, and the UK. Political groups sometimes tweet The Fact. People often tell The Fact to people involved in violent relationships.
The domain of the fact is domestic violence.
Here’s a version from The Washington Post.
Five Myths About Domestic Violence – The Washington Post
February 23, 2018
“Victims leave and return to a relationship an average of seven times before they leave for good — or are killed. Departure is the most dangerous moment for a victim, because the abuser suddenly faces a loss of control and may lash out. It’s a Hobson’s choice: Stay and possibly die, or leave and possibly die.”
(In this and all the versions, the underlined emphasis is mine.)
This is not presented as itself a myth, but as a true fact in opposition to a myth. The word “leave” links to a source – a CNN article. The claim is otherwise unsourced. The version in the CNN article looks like this:
When a Friend Won’t Walk Away from Abuse – CNN
January 10, 2013
Katie Ray-Jones, president of the National Domestic Violence Hotline and the National Dating Abuse Helpline, says there are countless reasons a victim returns to or stays in an abusive relationship: low self-worth, financial worries, fear of what the abuser would do if the abused left, even love….
Ray-Jones said an abused woman will leave a relationship approximately seven times before she leaves for good. The back and forth can frustrate friends and family.
The claim is otherwise unsourced. The version given by the National Domestic Violence Hotline, the one offered by Google in response to every search for any content related to domestic violence, is this:
50 Obstacles to Leaving – The National Domestic Violence Hotline
“Leaving is not easy. On average, it takes a victim seven times to leave before staying away for good.”
The word “seven” links to the CNN article above, and the word “times” is a broken link to a PDF from Planned Parenthood that I was not able to locate. Here’s another version, referring to The Fact itself:
Eliminate That Seven Times Statistic: How To Stay Away For Good
“On average, a woman will leave an abusive relationship seven times before she leaves for good, according to The National Domestic Violence Hotline.”
The link is to the above-mentioned National Domestic Violence Hotline article which links to the CNN article. Another popular source – offering the numerically highest version of the fact that I have encountered so far – is from an episode of the radio show This American Life. It is from a story about a high school girl who was beaten by her older boyfriend. She left the relationship, but got back together with him:
What’s Going On In There? This American Life Episode 567
“I insisted that, if we were going to keep producing this story together, that she meet with someone from an organization called Day One. They work with young survivors of intimate partner violence. I brought her to meet with their community educator, Sarah Gonzales….
Sarah Gonzales: The only thing that I would change in that is the “allowed” part. People, even with good intentions, unintentionally blame victims or survivors. So we say things like, if that person just had self-esteem, it would be OK. What that tells someone is that it’s your fault because you don’t have the esteem to leave, when that’s not really the case.
On average, it takes seven to nine times for someone to leave. So just because someone went back doesn’t mean that they’re never going to leave.
The claim is not sourced. In context, it is being used to reassure a teenage victim of domestic violence that her returning to her abusive boyfriend is normal. It is offered, apparently, to give the teenager hope that she could one day leave the relationship, even though she went back once already.
Here’s a version of the fact from Canada (recently tweeted by the local National Organization for Women chapter):
Silence Leads to More Violence in Abusive Relationships – Now Magazine, Toronto
“It takes an average of seven attempts before a woman leaves her abuser for good – but once she leaves, her chances of being murdered increase nine-fold”
“According to Statistics Canada, 70 per cent of all spousal violence is not even reported to police. A woman is assaulted an average of 35 times before she calls the police. It takes an average of seven attempts before a woman leaves her abuser for good. But once she leaves, her chances of being murdered increase nine-fold. A woman is murdered by her current or former partner every six days in Canada. All of these numbers increase significantly for Indigenous women, women of colour, immigrants and women with disabilities.”
The “seven times” claim is not sourced. No link is provided for “Statistics Canada,” if that is intended to refer to the claim.
The Ministry of Children and Family Services of Ontario provides a slightly different number:
The average woman will make up to five attempts to leave her abuser before ending the relationship permanently.
The source for almost every claim in this fact sheet is linked and labeled “Statistics Canada.” However, source 40, for the “five times” claim, has no link or page number. It is given as:
40. Okun, L. as cited in Yamawaki, N., Ochoa-Shipp, M., Pulsipher C., Harlos, A., and Swindler, S. (2012). Perceptions of Domestic Violence: The Effects of Domestic Violence Myths, Victim’s Relationship with Her Abuser and the Decision to Return to Her Abuser. Journal of Interpersonal Violence 27 (16). Pg. 3196.
That is to say, the authors cite a paper that in turn cites Lewis Okun’s 1986 book, Woman Abuse: Facts Replacing Myths. The paper is from 2012, suggesting that the reference is reasonably up to date, rather than based on data collected in the 1970s. There is a timelessness about non-indexical knowledge. The Fact seems to discourage asking When? much less Whom? and Based on what? The global knowledge game has its own peculiar temporality.
With Okun we can segue into the academic career of The Fact.
Here is the context for the Okun citation in the Yamawaki et al. paper, cited above:
Of the large number of women who are victims of domestic violence, a majority experiences difficulties when deciding to leave the violent relationship and may temporarily, but unsuccessfully, leave their partner many times before fully severing ties (Sullivan, Basta, Tan, & Davidson, 1992). Okun (1986) reported that the average woman will make five attempts to leave her abuser before successfully ending the relationship. Similarly, Peled, Eisikovits, Enosh, and Winstok (2000) noted that 50% to 60% of battered women returned to live with their abuser after being discharged from a shelter.
Yamawaki et al. are not the only recent academic authors to cite Okun on The Fact. In “Give Me Shelter: Temporal Patterns of Women Fleeing Domestic Abuse,” by Kathryn S. Oths and Tara Robertson (Human Organization, Vol. 66, No. 3, 2007), the authors present The Fact as follows:
Many, if not most, physically abused women eventually leave their partners permanently (Okun 1986; Strube and Barbour 1984), though in doing so they attempt to leave an estimated four to six times (Gondolf 1988; Okun 1986; Walker 1979) over an eight-year period before they finally leave for good (Horton and Johnson 1993). Thus, leaving an abuser, which often entails the aid of a safe house, tends to be a process rather than a discrete event (see Anderson and Saunders 2003 for thorough review). This process is the focus of the research reported here.
Oths and Robertson’s claim is a little different: they give “four to six” instead of “five,” and cite Lenore Walker’s The Battered Woman (mentioned above), Lewis Okun’s Woman Abuse, and Gondolf’s Battered Women as Survivors. Each source is a book rather than a paper. Notice that no page numbers are given. Nonetheless, we will look at each of these sources in turn for any mention of The Fact, and the context in which it arises.
Okun has been cited multiple times. Okun himself cites two sources in his literature review section, in page 56:
Women rarely quit their violent relationship after the first separation. It generally takes repeated beatings, preceded and followed by repeated promises to reform, before the victim becomes disappointed and scared enough to give up on the relationship. Walker believes that a woman will generally flee her abuser three to five times before finally severing their relationship. [Note 179.] The women in Hilberman and Munson’s group who left permanently averaged four to five previous separations apiece. [Note 180.]
Note 179 cites Lenore Walker’s 1979 book, The Battered Woman, that we have encountered several times before. Note 180 refers to at paper titled “Sixty Battered Women,” published in Victimology: An International Journal (Volume 2, Numbers 3-4, pp. 460-470, 1977-78).
Here is what Walker actually says in her book, in a section called “Tertiary Intervention,” regarding the provision of emergency shelter services (p. 189). The Fact occurs at the end of a long paragraph about the need for safe houses in the late 1970s:
Here the battered woman needs a totally supportive temporary environment before she can attempt to make any independent decisions. Safe houses, immediate hospitalization, and long-term psychotherapy provide such an environment. Most often this tertiary intervention is required immediately in order to provide optimum safety. These women will be unable to arrive at realistic decisions or act upon them unless they have this safety. The safe houses, refuges, and/or shelters which have sprung up across England and more recently in this country [note: the United States] have been an essential element in the treatment of battered women. The length of time a woman spends in such a controlled environment depends upon the individual. This time is used to regain the natural resources lost during her battering experience. Many women do become frightened at the prospect of being totally responsible for their lives and return to their violent relationships. However, such women, if permitted, will quickly leave their battering situation again and return to the safe environment. This back-and-forth movement often occurs three to five times before a woman is able to detach herself permanently.
Notice that The Fact is not here offered as a population average of any data, but as a number that “often” occurs, according to the author’s impression of, as she reports, 120 “detailed stories” and “fragments” of “over 300 stories.” The Fact is not here presented as an average, but the rest of the paragraph takes a decidedly non-indexical tone. The author can predict everything that “such women” will do, and knows that they are motivated by fear of responsibility, among the more surprising bits of positive knowledge offered. Recall that this is the cheerleader for “dragging” of various sorts.
I tracked down what seems to be one of the only existing copies of Hilberman & Munson, and here is The Fact in its context, in a section late in the short paper called “Treatment Considerations”:
The work with these sixty women has been without benefit of protection or shelter. It is slow, frustrating, and intensely anxiety-provoking, with the knowledge that a client’s self-assertion could prove fatal. It is helpful to keep in focus the following two realities. First, the client’s progress may seem small or inconsequential, for example, when a wife asserts her right to attend church. For the middle-aged woman who has been battered and isolated for twenty years however, this may be a monumental step forward. Second, it may take an average of four or five separations before the last hope for changes within the marriage ends and a decisive separation occurs. The clinician can expect a series of stormy trial separations and should not view this as a treatment “failure.” The battered woman is well aware of what she should or ought to do, and if she feels she has disappointed or failed the clinician, she will be reluctant to return to treatment in the future. The clinician’s insight and understanding will make it possible for her to resume the psychotherapeutic work when she is ready to do so.
Again, note that this is not presented as statistical information about a population, but in a very qualified way. It “may take” an average of four or five separations – or it may not. The paper gives no indication that this version of The Fact is based on data collected from the study, rather than the author’s impression.
In addition to the odd temporality of the global knowledge game, it presents us with an odd sense of identity. Hilberman & Munson are cited for The Fact in general, without regard to their very unusual study population. They were studying a sort of Silent Hill town, a rural and very poor place. In a twelve month period, they say, “half of all women referred by the medical staff of a rural health clinic for psychiatric evaluation were found to be victims of marital violence.” That is quite a study population! But in the world of the global knowledge game, “people” are “people.” A statistic based on an impression of sixty women living a very particular local nightmare in the 1970s is laundered into a fact about women and domestic violence in general.
But good news: Okun performed a study of his own.
Okun has the largest data set of any encountered so far: 300 intake forms from consecutive residents of a domestic violence shelter in the Ann Arbor, Michigan, area between 1978 and 1980. He reports on page 160 that
The shelter residents had separated from their partners an average of 3.1 times apiece prior to taking shelter. Twenty-seven percent of the women had never left their male partners for a conflict-related conjugal separation of at least a day. Twenty-six percent had experienced a single previous separation. 17% had been separated twice, and 30% had separated from their partners three or more times. Over two-thirds of the women who had previously separated from their husbands had experienced at least one conjugal separation within one year of their entry into SAFE House.
According to the text, Okun gets an average of 3.1 prior separations per woman, which, based on the other numbers, seems to be driven by high numbers in the “three or more” group, comprising only 30% of the sample. But Okun is cited, repeatedly, as giving an average of 5. Why?
Later, on page 198, Okun re-summarizes Walker and Hilberman and Munson and offers a new analysis of his own data, as follows:
Both Walker, and Hilberman and Munson, suggest that a battered woman will tend to leave her abusive partner four or five times for temporary separations before the relationship is dissolved. These authors also imply that, on the average, women who continue cohabiting with their violent partners have not previously experienced as many as four to five temporary separations from the violent mate. Separations as used here means the halting of cohabitation by the partners for at least one day in the context of conflict in the relationship. This does not refer to the other context of separation such as business travel by one of the partners. [Note: the intake form, reproduced by Okun, makes no such distinction.]
The present data clearly support the published hypotheses. The average number of previous separations among women who resumed cohabiting with the batterer was 2.42, compared to 5.07 among shelter residents who never resumed cohabiting after they left the shelter. This difference is statistically significant (t=1.85, df=1.63, p<.05, one-tailed). Women who had experienced one or no separations previous to entering SAFE House were only about half as likely as those who had experienced two or more prior separations to terminate their cohabitation immediately (23% vs 44%).
So the figure of five times per woman is only for a subgroup of women who convinced the staff that they would not return to their abuser. According to Okun, women who have not yet left five times are still in the process of going through that clockwork cycle. Since over 50% of his sample had made zero or one attempts to leave, most of his sample, according to to his theory, found themselves at the beginning of an unavoidable clockwork cycle of leaving five times.
I found Okun’s reported numbers confusing, so I looked at the data table at the end of the book, and my confidence in everything that had come before, low as it was, fell even further.
In Appendix A, beginning on page 234, Okun gives a summary of the different intake form items and the data collected for each. In some cases, the data are the same as that given in the text. For instance, the text on page 158 gives the average age as 27.7 and median age as 26. On page 234, “Age of shelter resident” gives a mean of 27.7, median of 26, and range of 16-55.
However, there are several discordances. On page 159, Okun says that unmarried cohabitants make up 14% of the shelter sample, whereas in the data appendix on page 235 gives 12% “living together.” Sometimes a median is given for a quantity I would expect to be given in integers, and the median is a decimal expression other than .5. For instance, the median for “elapsed time since last beating,” given in days, is 1.1. No average is given, but it is provided that the “range” is from one to 547. Similarly, the median for “Time elapsed since last separation” (in months) is 6.1, and for “elapsed time since first assault” is 38.3. That’s okay – perhaps a lot of people gave decimal or fraction answers, like “1.7 days” or “.75 months,” such that a “median” that isn’t a whole number or a decimal ending in .5 would make sense.
That possibility isn’t available for “number of previous separations” – The Fact. This is the most confusing entry in the entire Appendix, and probably the entire book. It gives the coding criteria – “Minimum 24 hours apart from husband/partner. Not legal separations necessarily, but in context of conflict.” (Again, nowhere is this mentioned on the intake form, which merely gives “Any previous separations? How many?” – presumably a diligent worker gave the same instructions about 24-hour separations in the context of conflict every time an intake form was filled out.)
There are no “remarks” that would otherwise indicate that the analysis was performed on a subset of the data, as is present for many other entries.
Here are the numbers for Number of Previous Separations, in the Appendix:
Mean = 0.9
The median of an inherently indivisible quantity, the number of previous separations, which must appear as a whole number or possibly a decimal of .5 if the median occurs between whole numbers, is .9. The average – previously given as 3.1 – is .9, less than one separation per woman. The range goes all the way up to 75 separations, which might lead us to suspect that the high numbers are driving the high averages previously mentioned elsewhere in the text, rather than a clockwork pattern heavily centered on five separations. There’s just no way to know what’s going on, except that it’s shady and unbelievable, and there are layers to its badness.
So we’ve gone through every text cited for The Fact, at least its five-times-per-woman version, except one – Gondolf 1988, Battered Women as Survivors, a study of over six thousand Texas women who used battered women shelters in 1984 and 1985. Edward Gondolf’s commitments are clear in his title, modeling women as survivors with agency rather than helpless victims. He thoroughly rejects the Learned Helplessness model. He mocks it repeatedly. Out of principle, he doesn’t even try to replicate The Fact, whose production was driven by the Learned Helplessness model in the first place. Instead, he asks about “helpseeking efforts” made by victims of abuse, including not just separations but a variety of helpseeking efforts, including calling friends, family, clergy, and police.
Gondolf does find that women in his large Texas sample made an average of five types of helpseeking efforts prior to intake at the shelter (see p. 29-30). But he wasn’t asking about separations at all, except as one of many types of helpseeking efforts. Even as he rejects on principle The Fact and its Learned Helplessness roots, poor Gondolf is cited nonetheless because he found the special number “five” for a vaguely related category. In the global knowledge game, it’s close enough.
There is one academic reference for the seven-times version of the fact. In Frasier et al, “Using the stages of change model to counsel victims of intimate partner violence” (Patient Education and Counseling, Volume 43, pp. 211-217, 2001), the authors say:
Some physicians may feel a sense of failure if the victim returns to the partner; however, victim advocates and shelter staff anecdotally report that victims of partner abuse relate that they make an average of seven to eight attempts to leave the abusive partner before they permanently leave. Empirical evidence of this report, however, could not be found in the literature. Nonetheless, just as relapse and recycling are an expected part of the stages of change process, so is returning an expected part of the process for abuse victims. Their return indicates neither a physician’s lack of skill in counseling and referring victims of abuse nor a failure on the patient’s part.
There’s a popular sound effect made by a trombone that comes to mind when I read this paragraph. What surprises me is that they bothered to report that they couldn’t find support for the fact, rather than relying on a citation with little probative value, as others did.
There’s also an alternate universe UK statistic that’s much lower, but seems to have arisen from the popularity of The Fact and the perceived need to say something about it. Here it is:
68% of high-risk victims try to leave in the year before getting effective help, on average 2 or 3 times each.
Source: SafeLives (2015), Insights Idva National Dataset 2013-14. Bristol: SafeLives.
This claim – “on average 2 or 3 times each,” qualified to “high-risk victims” – seems much more reasonable than the versions encountered in the US and Canada. It is based on data from a few thousand intake forms of a DV group, half referred by police. Unfortunately, the intake form itself is not provided.
Victims report making 2.5 attempts to leave on average before contacting the organization. They also report contacting the police an average of 2.5 times, identical to the previous figure. More women made reports to the police than made attempts to leave (73% vs 62%) and a huge percentage describe the perpetrator as an ex (60%), indicating that they did successfully leave. Many of these women are not “learned helplessness” victims going back again and again; it seems that they are women who left abusive partners, and the abusive partner kept coming after them. 75% of the sample were living away from the perpetrator at intake, and 81% were living apart from the perpetrator at exit.
It is not clear whether the statistic refers to only “high-risk victims,” but it seems not from the numbers, despite the claim of the linking page making that qualification. It is not clear if zero values are being included (zero prior attempts to leave), or if this is the average number for women who make some attempt at some time.
I think this is interesting as an example of a statistic being a “meme,” and the mistaken claim in the meme getting rehabilitated into something supportable. It’s much smaller in magnitude and more qualified than the original form of the claim, but it has some empirical support, if understood in proper context.
Why these numbers? Why five, and why seven? It seems likely that Walker’s “three to five” and Hilberman and Munson’s “four or five” versions of The Fact reflect an oral tradition that existed during the late 1970s. The oral tradition now, as evidenced by the radio show and the news articles including oral citations of The Fact, seems to be anchored on the number 7 – and perhaps growing, as suggested by the “seven to nine” and “seven to eight” versions.
The Fact may be classified as a Policy Legend, according to “Policy Legends and Folklists: Traditional Beliefs in the Public Sphere,” by Gary Alan Fine and Barry O’Neill (Journal of American Folklore 123(488):150–178, 2010). A policy legend is
a traditional text that describes an institution’s action in a context that supports a particular policy choice or presents social conditions in a context that calls for governmental or collective action. It may be transmitted in oral or written form. It often has a historical theme, presenting a “fact” or set of “facts” relevant to an ongoing public debate
Fine and O’Neill focus on lists (for example, that list of the top problems teachers used to have in their classes back in the good old days, like gum chewing and tardiness, compared to the top problems now, like guns and arson). The policy legends that Fine and O’Neill examine frequently end up reproduced in high-status sources like CNN, and the highest-status source is usually the most popular one to cite. We can see this pattern with The Fact: the CNN version specifically, and other high-status sources, are cited without regard to the fact that they offer little probative value. Again, it’s good enough for the global knowledge game.
Fine and O’Neill’s lists seem to self-repair to particular numbers: three, seven, and ten are particularly represented. I wondered if the movement from five to seven in The Fact might have some kind of folkloric significance. For some reason, I made a spreadsheet of the number of Google Scholar results for certain search strings in academic papers – “two myths,” “three myths, “four myths,” etc., as in papers with titles like “Eleven myths about snail breeding.” I obtained the following results, and I don’t think it proves anything, but it’s kind of neat. Shitposts are innocent, and inhabit the same epistemic ground as folktales. Shitposting is how we learn.
Is The Fact A Lie?
There’s a reason that The Fact feels like a lie, even though in most of the contexts in which it is offered, it may not function as a lie. It feels like a lie to me, because based on my experiences and background, when I hear a claim like “on average this happens this many times,” the theater in my head shows a normal distribution centered on the given average value. This is true even when no information about the distribution underlying the reported average is given. I argue that this is a reasonable mental unfolding, because there is a presumption that when information is offered, it is meaningful. An average gives a lot of information about a normal distribution, but not much information about a highly skewed or weirdly scattered underlying distribution.
If the average woman leaves seven times, says my brain, then it must be extremely rare to get out of an abusive relationship on the first try. For every woman that gets out early, assuming a nice symmetrical normal distribution, another woman will go through the cycle fourteen times or more.
I don’t think the picture above is the one that The Fact is calling to mind in most of the occasions for its use, such as victim counseling. (Which is good, since calling such a normal distribution to mind and unpacking its consequences instantly renders the fact ludicrous, and therefore presumably of little counseling value.) Using The Fact to reassure the victim that there’s always hope, even if she has tried to leave her abuser and failed, seems mostly fine to me. No matter how dubious this reassurance may be, and no matter how many ways it might be hypothesized to backfire, it is not necessarily harmful in this context. I certainly have no evidence that it’s harmful. But consider whether this story is more or less appropriate, more or less insulting, and even more or less of a lie than The Fact:
A princess was captured by a dragon and held in a dungeon for ten years. The princess explored all the tunnels, but never found a way to escape. One day, the side of the mountain caved in, opening an entrance to the outside world. But the collapse happened far from the princess, and she neither heard nor suspected it. She had long since given up trying to escape. A few days later, a mouse found his way into the princess’ chamber in the dungeon. The mouse gave her a kernel of corn. “If only I were as small as a mouse,” thought the princess, “I could escape.” The mouse returned and told the forest creatures about the princess. A robin flew into the cave to the princess and brought her a worm. “There’s a way out,” said the princess. “If I were as small as a robin, I could escape.” A rabbit brought her a daisy, a porcupine brought her a grub, a badger brought her a trout, and a coyote brought her an old sock. Always she cursed her size. Finally, a moose made its way to her, bringing a smartphone equipped with GPS in his mouth. The phone didn’t have any reception in the cave, but the princess got the message and escaped.
Science Magic Tricks with the Assumption of Non-Indexicality
I’m not actually on a crusade against The Fact. There are millions like it, and many of them are a lot more harmful in their use and effect. I’m just annoyed with the assumption that non-indexical knowledge always exists and is easy to get, and the shady tricks that are pulled in order to supply this impossible information. There’s a particularly egregious example involving the indexical and non-indexical meanings of words like punch, kick, hit, pull hair, shove, slam, throw. (I told you it would get worse.)
The psychologist Murray Straus developed a survey instrument called the Conflict Tactics Scale in 1979, around the time that Lenore Walker’s The Battered Woman was published. Straus’ goal was to develop a neutral, scientific survey instrument to detect abuse, excluding all the messy judgments that people make about what does and does not “count” as “abuse.” To this end, the instrument lists specific acts – slap, kick, punch, pull hair, throw something, etc. – and asks participants to endorse or decline to endorse whether each act in the (often quite long) list had happened to them in the context of a romantic relationship. Certain acts are categorized as “severe abuse” based on a global view of the word denoting the act. “Kick,” for instance, is denoted as severe abuse.
You may wonder what an instrument developed in 1979 has to do with contemporary Facts. In fact, the National Intimate Partner and Sexual Violence Survey, the survey administered to produce statistics for the Center for Disease Control, uses an act-based survey based on the Conflict Tactics Scale. I reproduce it in part:
This is the survey that’s used to produce such Facts as “One in five women and one in seven men have experienced severe physical violence at the hands of an intimate partner.” (Imagine connecting the two Facts: each of these one-in-five makes on average seven attempts to leave…)
From a global knowledge game perspective, it’s hard to see what’s wrong with this. Victims may tolerate violence without realizing it’s abuse, and an act-based questionnaire will pick up incidents that we all know are abuse, but the victim might not. Furthermore, imagine someone cruelly kicking a victim while the victim lies on the ground. Surely that’s severe abuse?
But a more indexical (mixed methods) investigation reveals a startling flaw in the methodology. Dobash, Dobash, Daly & Wilson warned about this back in 1992 in their “The Myth of Sexual Symmetry in Marital Violence” (Social Problems 39:1). (In addition to high absolute rates of “abuse,” act-based CTS-style surveys also produce surprisingly high levels of women abusing men compared to men abusing women, and it is this latter, rather than the former, that Dobash et al. were concerned with primarily in this paper.) I quote their argument from twenty-eight years ago at length:
Consider a “slap.” The word encompasses anything from a slap on the hand chastizing a dinner companion for reaching for a bite of one’s dessert to a tooth-loosening assault intended to punish, humiliate, and terrorize. These are not trivial distinctions; indeed, they constitute the essence of definitional issues concerning violence. Almost all definitions of violence and violent acts refer to intentions. Malevolent intent is crucial, for example, to legal definitions of “assault” (to which supporters of the CTS have often mistakenly claimed that their “acts” correspond; e.g., Straus 1990b:58). However, no one has systematically investigated how respondents vary in their subjective definitions of the “acts” listed on the CTS. If, for example, some respondents interpret phrases such as “tried to hit with an object” literally, then a good deal of relatively harmless behavior surely taints the estimates of “severe violence.” Although this problem has not been investigated systematically, one author has shown that it is potentially serious. In a study of 103 couples, Margolin (1987) found that wives surpassed husbands in their use of “severe violence” according to the CTS, but unlike others who have obtained this result, Margolin troubled to check its meaningfulness with more intensive interviews. She concluded:
While CTS items appear behaviorally specific, their meanings still are open to interpretation. In one couple who endorsed the item “kicking,” for example, we discovered that the kicking took place in bed in a more kidding, than serious, fashion. Although this behavior meets the criterion for severe abuse on the CTS, neither spouse viewed it as aggressive, let alone violent.
Back in 1992, psychology researchers were already concerned that act-based inventories fail to adequately discriminate abuse from non-abuse. In 2014, researchers Amy Lehrner and Nicole Allen published their “Construct validity of the Conflict Tactics Scales: A mixed-method investigation of women’s intimate partner violence” (Psychology of Violence, 4(4), 477–490).
Lehrner and Allen followed through on Dobash et al.’s suggestion to verify in an interview format what respondents “mean” by their responses to the CTS survey. As Dobash et al. report that Margolin found in 1987, Lehrner and Allen found that a “kick” can have extremely indexical meanings. They provide detailed examples of interview responses from women who endorsed “severe violence” on the survey (IPV is intimate partner violence, CTS is the Conflict Tactics Scale):
Descriptions of play fighting included behaviors coded as severe assault by the CTS, such as kicking, punching, or resulting in injury to self or partner. An exemplar of this pattern is participant 106 (level 4), who endorsed significant frequency and severity of violence on the CTS, including items such as throwing something, twisting arm or hair, pushing, shoving, grabbing, slapping and two instances of having had an injury for herself and her boyfriend. When asked about times when arguments or fights have ever “gotten physical,” she replied:
P: Honestly, not when we’re fighting have we ever been physical. I mean when we’re joking around and you know, things like that? We like to just play around and like pretend to beat each other up. But it’s not anything that would really inflict pain on each other. And if it is, it’s very minor and it’s accidental. But…
Q: Not in the context of like an argument or a conflict?
P: No. I honestly cannot remember a single time where we were fighting and it had gotten physical. Usually if anything we’re on the phone or we are 5 feet away from each other.
When questioned more directly about her CTS endorsements she responded, “I think I might have been thinking in context of playing sort of thing. . . it probably, I thought it was in the context of just at all?. . . And when we are like at all, you know, physical like that it’s just like all in fun.” When asked for examples she reported:
A: Well I’ll either just like go up to him – “let’s fight” or something – and then I’ll just like you know lightly punch him or something like that. And then you know he would like pick me up and throw me on the couch, and like start tickling me. Usually that’s what ends up stopping it is that he’ll just keep tickling me until I can’t breathe anymore. But I mean that’s pretty much it.
Q: And so you’ll get to that point where you’re trying to really overpower him and see if you can do it, and you never can?
A: Yeah. Usually the only thing I can do is like pinch him or something? I pull his hair and he’ll stop tickling me. But that’s about it.
Q: So when are the times where you’d be in the mood to say, “hey let’s fight?”
A: I don’t know, a lot of times when I’m at home, I get bored and there’s really not a lot to do in my town. So watching TV and movies gets kind of old. So we’re just you know, something to do? Like kind of like a brother/sister type of thing, where you just have nothing else to do so let’s beat up on each other kind of…
This participant’s description of her relationship across the interview is consistent with this report, and she provides no indication of IPV by either partner in the interview data. Although coded as severely and frequently assaultive by the CTS, her endorsements on the CTS appear to miscategorize her playful behavior as IPV.
Playful violence and mock violence are both mistakenly captured by act-based, CTS-style instruments. From Lehrner & Allen, con’t:
Almost 20% of the interview participants (n=5) coded as violent by the CTS clearly described mock violence.
Participant 90 described a pattern of her boyfriend playfully provoking her and apologized when she thinks she has misled the researcher to understand her behavior as an act of violence. She provides an example: “[I say] ‘you’re getting on my nerves,’ like, ‘stop’ [indicates a playful shove]. He plays around and stuff like that so I’m sorry but I didn’t…” Mock violence generally involves a “minor” form of violence, but as with playful violence, it can include more severe items. An exemplar is provided by participant 239 (level 4). In the interview she described multiple incidents of serious violence, which she identified as such, against her boyfriend. However, she laughed when she realized her endorsement of “kicking” her boyfriend (a severe item) had been interpreted as an act of violence. She described the incident as follows: “he came to visit and we were just like messing around. …he was like on the floor and I like kicked him like ‘get up, like be serious, like just get up…’ So that’s probably like a tap, like that was probably…I shouldn’t have wrote that.” She continued, “Well that was like a joke. I wasn’t yelling or anything.” She distinguishes her real violence, which was intended as violence, expressed in anger or frustration, and described in detail, from this kind of “messing around.”
Each of the incidents reported above would be classified as “severe violence” according to the NIPSVS scale. A “kick” is not necessarily a kick.
We might further attempt to measure how wrong the CTS-style survey instruments are for a given population, and devise a correction. But if the answer to the question matters, why not use an instrument to measure it that works the first time? Why use something that’s demonstrably been garbage for thirty years? It’s hard not to conclude that true, accurate information is not what is required in these survey procedures, but something else – probably something different in each case.
“Word laundering” is the term that comes to mind with survey instruments in general.
These survey instruments, these facts, these studies are not special or unique – at least, I have no reason to suspect so. Domains that widely rely on survey instruments – depression inventories in mental health, happiness studies, education, nutrition – seem particularly vulnerable to taking words wrongly seriously.
And you can’t tell from looking just how bad a survey instrument is. A lot of the badness is how the instrument is used. Unwarranted inferences, suspicious groupings of unlike things, and layers of other tricks and malfeasances can be buried along the path from survey instrument to Fact. The problem is not the survey instrument itself. The problem is the expectation that this kind of knowledge can be had, and confidence that a survey is the way to get it.
I will leave the Big Five Factors of Personality and the lexical hypothesis in general as an exercise, except for a few comments. The Big Five is interesting for what a crappy metaphor it provides. Compared to Seligman’s dogs, it’s a nothingburger, a mere list of five vague qualities. At least the 1960s version of the qualities (a set of words used to summarize dozens of inane questions, thereby unlocking the mysteries of the human soul) sang to the imagination a bit: ”surgency”, “agreeableness”, “dependability”, “emotional stability”, and “culture.” It’s interesting that the Big Five manages to be popular despite not offering much narrative hook at all. It’s incredibly boring and it’s STILL probably wrong and/or meaningless.
Skillful Ignorance and Transmissible Cancer
It’s hard to avoid knowing things.
One of the titular studies in Kenneth Liberman’s More Studies in Ethnomethodology (2013) is the study of crossing a street. The street is called Kincaid, and it passes by the bookstore at the University of Oregon. Liberman says that colleagues describe it thus:
“It seems to be a real confusion there in front of the bookstore – I try to avoid crossing there if I can.”
“As a driver I find it frustrating at that intersection because so many kids just seem to walk without paying any attention.”
In watching the intersection over many hours, videotaping all the while, Liberman and his students discovered practices that people used in order to cross Kincaid. One of the most delightful “ethnomethods” they discovered is styled [BEING OBLIVIOUS].
(Note: It’s put in brackets to show that it’s an ethnomethod – a practice people use to deal with the reality of human interaction, often specific to their contexts. [APOLOGIZING WHEN SOMEONE BUMPS INTO YOU] might be an ethnomethod.)
When the driver quoted above saw pedestrians step out into the street “without looking,” perhaps looking down at a phone the whole time, the driver interpreted this as genuine obliviousness. However, [BEING OBLIVIOUS] appears to be an effective strategy for crossing the street – Liberman even suggests [DOING OBLIVIOUS], to emphasize that it is a skilled practice and not a state of being. In watching people use the intersection, Liberman and his students discovered a language of gazes, motions, speeds, and accessories (sunglasses, phones, enormous posters) that drivers, pedestrians, and cyclists used to negotiate passage. A student who steps off the corner appearing oblivious – wearing sunglasses, looking down at a phone, carrying a huge poster perhaps – is usually not actually oblivious of the risks to him from traffic. Generally, he will have carefully selected a moment, taking advantage of the exact position and speeds of cars, cyclists, and pedestrians around him. The student is using his apparent, visible obliviousness as a weapon against the drivers. If they can’t catch his gaze, they have to stop! It’s very similar to the strategy in the game of Chicken of tossing one’s steering wheel out the window, and the whole topic brings Thomas Schelling’s work firmly to mind.
Of course, sometimes the cars win, and often everybody wins. There were many cases of cars, people, and cyclists using “illegal” methods in a way that seemed to minimize the stopping and wait times of all involved. In past decades, the Kincaid intersection has been temporarily modified with various top-down control solutions, including traffic lights and stop signs and painted intersections at various places. But Kincaid seems to work better without these interventions (or with fairly minimal interventions). Chaos, organized mainly by intelligent people using available methods, seems to outperform naive top-down analytic solutions in this case.
Appearing oblivious in traffic is a skilled practice. Novice crossers of Kincaid are not able to perform it. Similarly, remaining oblivious to the global knowledge game is an ongoing skilled practice. The first step, I think, is to develop an aesthetic for it, to make it less invisible.
William James, in his 1895 lecture “Is life worth living?” takes a metaphorical turn in contemplating the Problem of Evil. He invites us to consider, as metaphorical targets for ourselves, dogs who are used for medical experimentation:
Consider a poor dog whom they are vivisecting in a laboratory. He lies strapped on a board and shrieking at his executioners, and to his own dark consciousness is literally in a sort of hell. He cannot see a single redeeming ray in the whole business; and yet all these diabolical-seeming events are usually controlled by human intentions with which, if his poor benighted mind could only be made to catch a glimpse of them, all that is heroic in him would religiously acquiesce. Healing truth, relief to future sufferings of beast and man are to be bought by them. It is genuinely a process of redemption.
You have to wonder about this in the case of Seligman’s dogs. For what “relief to future sufferings” were they tortured for? For what “healing truth”? What kind of immortality is it to live on as a self-replicating misleading metaphor?
Consider The Fact that was once a suffering dog. The canine transmissible venereal tumor was once a dog, too, and offers a like kind of immortality.
Three Ways Forward
I see at least three possibilities toward a less cancerous form of inexact science.
The first is the attempt to restore lost context to celebrated results in the social sciences. For example, Gina Perry’s book The Lost Boys: Inside Muzafer Sherif’s Robbers Cave Experiment (2018) adds back context to the classic work of Muzafer Sherif: an earlier trial run of the experiment, aborted because the children would not produce the results that Sherif wanted; interviews with the child subjects, decades later; and details of the narrative and economic incentives facing Sherif. Every month or so, a new classical result of psychology or sociology seems to disintegrate when someone brings it together with the context of its production. The one about normal people posing as fake psychiatric patients seems to be getting contextualized out of existence. The strategy of adding back context produces two benefits: it reduces the hypnotic power of universal, context-free facts without biographies (it is corrective), and it also provides interesting case studies on how people manage to produce these context-free facts in the first place (it is informative on its own).
The second possibility is to engage in forms of research that seek to locate and preserve the information that is indexical to a situation, rather than attempting to create global knowledge at all. That is the goal of the tiny field of ethnomethodology, for example. This kind of science is still rare. There’s a proverb to the effect that biology without Darwin was just stamp collecting. I don’t know how much hope there is for social science to stop hallucinating Darwins everywhere, and go back to honest stamp collecting, but I suspect that it is the only way toward social science knowledge that would be genuinely valuable. These methods look more like watching reality with heightened attention, acquiring from each situation the tools with which to understand it, rather than devising contrivances to advance narratively compelling theories (and sometimes even ostentatiously boring ones).
Tal Yarkoni opens the paper “The Generalizability Crisis” (preprint, 2019) with a joke about generalization (I give the Wikipedia version):
An astronomer, a physicist and a mathematician are on a train in Scotland. The astronomer looks out of the window, sees a black sheep standing in a field, and remarks, “How odd. All the sheep in Scotland are black!” “No, no, no!” says the physicist. “Only some Scottish sheep are black.” The mathematician rolls his eyes at his companions’ muddled thinking and says, “In Scotland, there is at least one sheep, at least one side of which appears to be black from here some of the time.“
The mathematician is acknowledged in this joke as the master of being aware of generalizations, which are often disguised, invisible, hiding. Is this extreme? Yarkoni argues that inexact scientists should take a page from mathematicians and get more aware of their generalizations, their careless laundering of words. He says:
Let us assume for the sake of argument that there is a genuine and robust causal relationship between the manipulation and outcome employed in the Alogna et al study [in which the subjects viewed a video and had poorer recall if they’d subsequently done verbal tasks before memory testing]. I submit that there would still be essentially no support for the authors’ assertion that they found a “robust” verbal overshadowing effect, because the experimental design and statistical model used in the study simply cannot support such a generalization. The strict conclusion we are entitled to draw, given the limitations of the experimental design inherited from Schooler and Engstler-Schooler (1990), is that there is at least one particular video containing one particular face that, when followed by one particular lineup of faces, is more difficult for participants to identify if they previously verbally described the appearance of the target face than if they were asked to name countries and capitals. This narrow conclusion does not preclude the possibility that the observed effect is specific to this one particular stimulus, and that many other potential stimuli the authors could have used would have eliminated or even reversed the observed effect.
Instead of trying to prematurely shave the indexicality off of observations, rendering them useless and meaningless, the context could instead be maximally preserved from the beginning.
So far I have discussed the corrective project of adding back indexicality to global knowledge, and the ethnomethodological project of preserving indexicality from the beginning. I think that a third alternative is honest fiction – stories that do not pretend to be reports of reality, but admit that they are the constructed products of minds.
In realms of guaranteed ignorance, what do the “experiments” offer (other than ritual sacrifice, if not of animals then of undergraduates’ time) that folktales or thought experiments with similar content couldn’t provide just as effectively? It is only folkloric and narrative value that the classical experiments retain.
Fairy tales and the normal distribution that The Fact calls to mind are not really indexical, since they do not refer to any particular set of circumstances. They are free-floating categories, or metaphors, that can function as attractors to collect like situations. Sometimes these stories, metaphors, categories, etc. are useful, and sometimes they can introduce error. They must be deployed with skill, and, more importantly, skillfully bypassed when appropriate.
19 thoughts on “Ignorance, a skilled practice”
This is a minor point, but it’s what led me here:
I think HHHT vs HHHH is not a great example, because statements that sound similar have different odds, and it’s not clear to me that intuition is any better than analysis for this.
In particular, what seems to me the “most natural” bet would be something like: “Sam flips a coin repeatedly, if they get HHHT first I win, if they get HHHH first you win”. The odds in that case are 50% each.
The situation where the odds are different is something like: “Tev and Una flip a coin repeatedly at the same rate. If Tev gets HHHT first I win, if Una gets HHHH first you win”.
The odds are different in that case because Tev ignores getting HHHH first, and Una ignores getting HHHT first.
HHHT has a shorter expected time than HHHH, because the expected time for either HHHH or HHHT is the same, but once you do get HHHT or HHHH, the expected time for the other sequence is different:
When you have HHHT, you need at least 4 more flips to get HHHH.
When you have HHHH, you have a 50% chance of getting HHHT on the next flip.
This feels to me like cherry-picking an example with unusual rules to demonstrate the point.
It’s possible to choose pairs of sequences that have various outcomes, and most of these do not seem to me “intuitive” in any particular way.
For HHTH vs HTHH:
HHTH is more likely to appear first,
but they both have the same expected time to appearance.
For HHTHT vs HTHHT:
They are both equally likely to appear first,
but HHTHT has a shorter expected time to appearance.
The main reason for this being unfair is that HHHH will almost always result in a HHHT, unless it is at the very end of the sequence. It’s an unfair bet, because it is betting on it showing up at least once, rather than the number of times it will show up.
HHHH and HHHT are equally likely to appear on average, but the distribution is not the same; HHHH is much more likely to appear twice or more in a finite-length sequence than HHHT is, because HHHHH will create two instances while HHHTHHHT is the shortest double sequence. HHHT can show up without HHHH but HHHH can only show up without HHHT if it is in the last few characters.
If it was “every time HHHH appears, I give you $100, every time HHHT appears, you give me $100”, it’d be closer to a fair bet. In reality, HHHT still wins because you can’t have an arbitrarily long string of heads, but it has much, much less of an advantage.
I’m pretty sure the latter bet is fair, by linearity of expectation. You can have an arbitrarily long string of heads, it happens with probability (1/2)^(# of heads).
Ah, I just saw this on twitter an hour ago.
Just wanted to say how much I enjoyed it. One of the most wonderful things I’ve read recently!
Slightly related, you might enjoy this post on Gelman’s blog: https://statmodeling.stat.columbia.edu/2018/03/20/statistical-controversy-trophy-wives/
From the abstract:
It isn’t an indexicality issue exactly, but I thought it was interesting how they tried to examine an individual phenomena — that is, there exists some women who have access to relationships with high(er) achievement men due to their beauty — with aggregate data, that is, seeing if the representative woman is involved in such a relationship. I see this strategy a lot, and it seems so obviously missing the point that I don’t understand why people do it.
Of course there is a fairly large statistical error here that vitiates their results also, but even absent that I don’t think the study isn’t really investigating what they claim it is. Really shocking the amount of things going wrong here.
The blog post also contains a photo of two cats, who are clearly both maximally beautiful and high status. very cute.