The Other "Literally"

When I say "I'm sick to death of your constant bickering", I am not actually claiming that your constant bickering has left me sickly to the point of mortal death, and you know it. This is a common figure of speech. Moreover, if you were to try and criticise this statement by pointing out that it was untrue, owing to my rude health and stubborn habit of not being dead, everyone would think you were being an arse.

In saying "I'm sick to death of your constant bickering", I'm saying something I don't literally mean. I don't expect to be held accountable for it, (at least not the "to death" bit), and everyone is complicit in this. This is fine, and happens for understandable reasons. I'm also not really so hungry that I could eat a horse, and whether Colin could actually organise a piss-up in a brewery is besides the point when assessing his competence. We have an elaborate zoo of literary and rhetorical devices in which we do not say what we literally mean and yet people still understand us. Fine. Fine and peachy.

But there is another type of wordy embuggerance I see people using, where they say something they don't mean and don't expect to be held accountable for it, but they very much should be. I don't have a good name for it, and I don't know how to deal with it. If I had the former but not the latter, I would call it out every time I saw it until my lungs gave way. If I had the latter but not the former, I'd be entirely satisfied with burying it in an unmarked grave.


If you were listening to Romeo waxing poetic about Juliet, and he said "Juliet is the sun", you'd know he wasn't literally saying Juliet is an astronomical ball of plasma approximately 1.4 million kilometres across. You'd surmise that he was employing a figure of speech, because the literal interpretation is absurd.

If he were to say "Juliet is the worst person in the history of the world", you'd probably think that he was using hyperbole. You'd be on slightly shakier ground here, because there is a literal interpretation of his statement that makes sense, but to the best of your knowledge Juliet hasn't committed any genocidal atrocities or instigated any transglobal conflicts.

If he were to say "Juliet thinks she's Ginger Rogers, but she isn't", you would not think that Juliet is experiencing an identity crisis, which Romeo is dutifully reporting. Romeo is probably implying that Juliet has an over-inflated opinion of her own dancing. This is a more impressive inference on your part. If Juliet did think she was Ginger Rogers, and Romeo wanted to express this fact, he'd probably say something like this. In spite of this obstacle, we understand the sentiment he is expressing.

So far, so satisfactorily non-literal. Here is where we cross the threshold.

If Romeo were to say "Juliet expects to get what she wants because she's wealthy and pretty", I wouldn't be sure what to do with this one. On the one hand, it's a factual claim. You could plausibly (and reasonably) interpret it as Romeo asserting that Juliet expects to get what she wants, and that this is caused by some combination of her being wealthy and pretty. As a parseable declarative statement, it's vulnerable to counter-assertions: if Juliet doesn't always expect to get what she wants, if she isn't wealthy or pretty, or if the causal claim between the two doesn't hold, it is false.

On the other hand, you could imagine that Romeo is simply expressing the sentiment "Juliet is spoiled and conceited", or perhaps just "boo Juliet!" If you confronted him on the factual merit of the statement, you could imagine him defensively making further disparaging remarks about Juliet, swigging from his ninth bottle of Blue WKD and shaking his fist at the sky.

This is the wordy embuggerance I'm talking about. On one level it has a plausible literal truth condition with implications if that truth condition is met, and there is no obvious indication that it's not meant to be taken literally. On another level, it is expressing a sentiment the speaker feels passionately about, and you can't take that away from them.


Many years ago, on the lofty philosophical forum known as LiveJournal, I was in a discussion about Stephen Fry. He'd apparently said something disparaging about women, and my interlocutor claimed this was because he was a gay man.

I questioned this. If he wasn't gay, would that have made him less likely to say it? Are other gay men at greater risk of saying these things than heterosexual men? Is his gay-mannitude really the dominant causal factor in making this specific remark about women?

Her response was something along the lines of "well, if you're going to take it to ridiculous extremes..."


I don't think it's taking things to ridiculous extremes to hold people to task on the meaning of what they say and the implications thereof. But there is a problem here. If you refer to that meaning as "the literal meaning", it opens the door for arguments about implied meaning and subjective interpretation and hermeneutic witchcraft that are completely besides the point of not saying things which have readily-available meanings you don't care about.

So what do you call that interpretation if not "the literal interpretation"? The actual meaning? The for-realsies that-which-people-who-care-about-what-words-mean meaning? The woman I was talking to on LiveJournal had a legitimate sentiment about the uniquely-privileged position of gay men (the veracity of which is outside the scope of this post), yet I can't help but feel that sentiment is subordinate to what the words actually mean, and in the absence of it being a common figure of speech, or other indicators that it's not to be taken literally, that's the standard you should hold it to. That doesn't feel like the curmudgeonly insistence of a square-headed logic-warrior. It feels like a sensibly-prescribed method for using words.

My verbal response to that LiveJournal comment was something about how words mean things, how they do important work, and how they aren't just billboards for political causes. My psychological response was to have a part of my mind start screaming, and to this day it hasn't stopped.


Once you start looking for truth conditions, it's hard to stop. The process is quite simple: just ask yourself "what would it mean for this statement to be true?"

Here is an opinion piece from The Independent about how masculine stereotypes are horrible and damaging and generally boo-worthy. I'm not a massive fan of opinion pieces, since reading them will probably not make me smarter, happier, or more well-informed. That said, I'm pretty on-board with this one. A friend of mine shared it on Facebook, along with the following singled-out quote:

...young men need to understand as early in their lives as possible that men have a long history of getting their way for no good reason.

There's a charitable sentiment to extract from this about male privilege, and a slightly less charitable sentiment along the lines of "Go Team Feminism". I don't object to either of these sentiments, but I do object to the statement itself. It seems indefensible from a variety of angles. There are many adjacent statements on much sturdier ground, but the author didn't choose any of those. There are also many less-objectionable sentences in the piece, yet the friend who shared it on Facebook didn't choose any of those either.

Most importantly, if you try and reason with this statement, weird answers will fall out of your reasoning process. This doesn't seem to stop people, though they may as well be reasoning with a flag or a Beach Boys song for all the good it will do. Try it yourself: what would it mean for this statement to be true? What are the necessary conditions for this to be the case? What are the necessary consequences of this? Are these conditions and consequences met and born out by other observable facts about the world, or do they generate conflict? If you assume its truth, can you only generate reasonable implications from this, or does it produce implications that are inconsistent with other things we hold to be true?

If this feels like it's being taken to a ridiculous extreme, I have to ask: why give someone the opportunity to take it there? Reductio ad absurdum isn't something Alice does to Bob that reflects badly on Alice; it's something Bob allows Alice to do that reflects badly on Bob. If you can express a sentiment using literally true statements, (as is quite achievable in the above case), why don't you? If you can't express a sentiment in this way, maybe there's a reason for that.


There's a simple take-away from this post: don't make words that don't mean what you want them to mean. In the words of H.P. Lovecraft, do not call up that which you can not put down.

I would like to think that there are two sorts of people in the world: those who value precise, meaningful discourse evaluated by a common standard, and those who don't. That way we could slowly convert the latter into the former. Unfortunately, I don't think the world is that tidy. Plenty of smart people will throw words around without stopping to consider what those words mean. I can't see this being good, and I'm not sure how to convince them to do otherwise.

It feels like it's a straightforward question: "what would it mean for this statement to be true?" Having an answer for this question in response to the statements you make feels like a reasonable minimal standard, but I can appreciate it's quite hard to live up to. Still, as mental habits go, it's a powerful one to cultivate.

Say it to yourself a few more times: what would it mean for this statement to be true? For the next five claims someone makes, ask it to yourself and see what happens.

Imaginary Expertise

I'm currently reading The Sense of Style, partly because everyone else seems to be reading it, and partly because it's so good I have trouble putting it down. The book is a contemporary style guide, combining linguistics, cognitive science and writing convention in a compendium of advice on writing well. Linguist, writer and cognitive scientist Stephen Pinker is obviously well-placed to write such a book, and I'm glad he has.

Aside from being an interesting and useful guide on forming pleasant English, this book has also given me something else: it's allowed me to capture an example of my own bad behaviour out in the wild. You see, I think I know about grammar, but I don't. Not really. Beyond secondary school I have no formal education in the subject, but I've read enough around it to feel like I know what I'm talking about. This is useful, because I have an in-situ example of what this behaviour feels like from the inside.

It's not that I don't know much about the subject. Far from it. Compared to Average McStreetPerson I know an awful lot about grammar and linguistics, but that's a lousy scale to hold oneself against. Someone who reads a lot of popular science knows much more about physics than Average McStreetPerson, but that doesn't mean they know a lot about physics compared to the field of academic inquiry known as physics.

There are a few features of Pinker's book that alerted me to my imaginary expertise in grammar, most notably his talk about contemporary theories of grammar overthrowing incumbent ideas I thought of as canonical. In the abstract I was dimly aware that people must be doing modern research into the subject, but I could't have told you what any of it was.

I can't remember where, but I distinctly remember reading about some research on how humans have trouble distinguishing between familiarity and knowledge. This is presumably responsible for the experience of looking over what you've just been revising, and feeling like you know it without actually having retained anything. I wonder if imaginary expertise is like this. I'm familiar enough with, say, verb forms to follow a discussion about them, but I don't have command of them. My familiarity is reinforced every time I follow such a discussion, but my command faculty is rarely challenged, as it would be if, say, I took an exam.

If this is the case, it gives a useful rule of thumb, which in retrospect seems obvious: take tests on subjects and see where you fail. The less obvious aspect of this rule of thumb is presumably to avoid questioning the need to take a test, or to make excuses as to why they're not necessary.

The Other Groupthink

[Experimenting with bashing out a few hundred words on something I'm thinking about. I start writing too many things before abandoning them for not being immaculate. Perfection is the enemy of the tolerably mediocre.]

I have spent a considerable amount of time, effort and resources learning about subjects such as social choice theory, international development and government regulatory mechanisms. These are diverse areas of inquiry which, for historical and administrative reasons, are placed under a common category called "economics", but which otherwise don't resemble one another very much. So when someone comes along and makes broad sweeping statements about "economics" or "economists", as if these words point to coherent homogeneous groups of people and activity, I want to scream in their face until the seas boil and the skies fall down.

There may very well be all sorts of legitimate criticism to be levied at some subset of economists, but if you try and reason non-trivially at the group level about all economists, the chances of you not being somehow wrong are as good as zero. I've made this case in a couple of places on the subject of economics, (c.f. "screaming", "seas boiling", etc.), but the general pattern keeps on coming back to haunt me for groups in general:

Although [group] sounds like it's referring to a coherent class of people or activity, [group] doesn't actually capture a homogeneous category in general, and doesn't define the features you're arguing about in particular. As a consequence, thinking and talking about [group] as if this wasn't the case is going to have undesirable results.


In a few places recently, I've heard complains about superfluous use of the word "white". This is probably best exemplified by the term "white nerdy males". The word "white" in this construct isn't really doing any work in whittling down the referent. Someone making a statement about "white nerdy males" isn't trying to exclude ethnic-minority nerdy males from their statement, but trying to stick as many "privileged-group" labels onto the referent group as they can before making claims about it. There may very well be a group of people to whom their statement or reasoning applies, but all literal members of the group defined by "white nerdy males" probably isn't it.

Scott Alexander made a very interesting post on the subject of unhelpful political coalitions accidentally falling out of perceived threats to a wider group identity. Then he shot himself in the foot a bit by reasoning about feminism as a coherent homogeneous category of activity. More recently, I think he's been getting better at this.


For the past couple of weeks I've been trying to notice whenever I, or anyone else, makes claims about a defined group that probably isn't the intended group to whom the claim applies. This has been quite humbling. I've received specific training in figuring out to what extent individual behaviour generalises and generalised behaviour is applicable to individuals, and yet I clearly don't apply this as a matter of course.

A lot of very silly disputes go away when you realise the object of dispute doesn't really exist in any meaningful sense, and if I can get into the habit of questioning whether any given group I find myself reasoning about is fit for purpose, I hope to make a lot of problems go away, or at least flatten out into more manageable ones.

The Arse-Cheeks of Statistics, and Titillating one's Smart-Parts

[This post is brought to you by me being grumpy about Epidemiology: A Very Short Introduction.]

One of the reasons I went off popular non-fiction a couple of years ago was that it got repetitive. There are only so many times you can read an introductory explanation of Prisoner's Dilemma without wanting to chew up the pages and spray them like confetti out of the window. It makes me wonder how many pages I've read, redundantly reintroducing me to something I know already. How many books do they all add up to?

I can forgive something like Prisoner's Dilemma because game theory is slightly esoteric, but I seem to have a lot less patience when it happens with statistics. I have lost count of the number of places where I've been "introduced" to Pearson's r, or regression to the mean, or sampling biases. My dear host, we've already met, and I know them well. Very well. In fact, I can tell you what kind of underwear they prefer and which songs they end up singing when they're drunk at 3 o'clock on a Saturday morning.

It's not just me. A lot of people are very friendly with introductory statistical concepts. It is not a shy discipline. Anyone who's completed a quantitative research methods course (the majority of science graduates?) has seen at least one arse-cheek of statistics. This is a fairly standard and cross-domain body of knowledge, so let's not be coy.

I assume authors don't want to spend their time torturing metaphors for linear regression either. Nate Silver must prefer to state facts about his material in the obvious language for doing so. I can't help but feel that there's the opportunity for a much better class of popular science writing, if we could just nudge the general public's statistical literacy up a couple of notches.

This would possibly seem a bit elitist as recently as ten years ago, but there are so many resources for gaining statistical literacy in 2014. If you're in a position to read Freakonomics, you're probably in a position to take one of the dozens of MOOCs on introductory stats, or to pick up the quite fine Cartoon Guide to Statistics.  If everyone learned this once, we could factor out hundreds of pages of insipid statistics in all our future reading.

In my ideal world, authors should be able to stick a logo on the front of their book. Maybe a nice red bell-curve. That logo will say:

To read this book you should know, more or less, what a probability distribution is. You should understand measures of dispersion and central tendency. You should understand logs and e. You should know what 'population' and 'sample' are referring to in context. 'Significance testing', 'statistical power' and 'regression' should not be alien terms to you. It would be nice if you had some idea what 'Bayesian' means. We trust that if you don't know a particular piece of terminology, you can look it up on Wikipedia. This book is not just here to titillate your smart-parts. It's here to teach you something.

I think a lot of people read pop-sci books to feel smart, and to signal their smartness to other people. If you make those little red logos an intellectual status symbol, you're going to seriously put the boot up the level of public discourse.