Varieties of motte and bailey

Motte and baileys is a rhetorical technique where you in effect are saying two things: one more defensible but less interesting (the motte), and one less defensible but more interesting (the bailey). The concept was invented by Nicholas Shackel, and was popularised by Scott Alexander. Ben Garfinkel provides some good examples:

“Reality is socially constructed”: On the “motte” reading, our beliefs about reality and the concepts we use to make sense of it are heavily influenced by our social context. On the “bailey” reading, reality itself — for example, the chemical composition of water — is determined by one’s social context.

“Killing is killing”: On the “motte” reading, killing is the same thing as itself. On the “bailey” reading, some relevant form of killing (for example, the killing of farm animals, the killing of embryos, or the killing of people who have received the death penalty) is morally equivalent to any other act of killing.

Shackel and Alexander give similar examples. They’re all pretty philosophical, and they all trade on key concepts being used ambiguously, in a way that’s clearly pretty dodgy. Most sensible people would agree that this isn’t a great way of arguing. I take it that Shackel, Alexander, and Garfinkel all agree on that.

But motte and baileys, or what reasonably could be seen as such, is a much wider phenomenon. People are going back and forth between interesting but indefensible claims and defensible but uninteresting claims all the time. And they make analogous moves relating to speech-acts like requests.

*

An important case is implicatures and other pragmatic effects, as Bryan Caplan points out. Take, for instance, indirect requests, analysed by Pinker, Nowak, and Lee:

Gee, officer, is there some way we could take care of the ticket here? [a bribe]

Would you like to come up and see my etchings? [a sexual come-on]

Why do people make these requests in this indirect way, instead of asking directly? Because indirect requests allow for plausible deniability: “a cooperative listener can accept the request, but an uncooperative one cannot react adversarially to it”. That is, if the listener is cooperative, you stay in the bailey, but if they’re uncooperative, you retreat to the motte.

These motte and bailey-requests are quite different from the standard examples of motte and baileys (“Shackelian motte and baileys”). There’s nothing particularly philosophical about them.

There are many other examples of how implicatures and other pragmatic effects are used for motte and bailey-purposes. Humour is one. People make jokes-cum-arguments where the argument is shaky. If they just get laughs, they stay in the bailey, letting the argument interpretation stand. If challenged, they retreat to the motte: “it was just a joke”.

*

In academic writing, you’re supposed to be more literal and explicit, which means that it’s harder to make use of those particular devices to construct motte and baileys. But there are other devices. Subtle use of caveats and hedges is one. Many claims are interesting but indefensible when stated without caveats, and less interesting but more defensible when the right caveats are added. How can you have the cake and eat it?

One way is by relying on others who will report on your claims, such as the media. Since they need a snappy version, the caveats and hedges drift away when they report on your paper. This way, most people read the interesting bailey version, even though you sit safely in the motte, having included all the right caveats and hedges in your paper.

But you can also pull off the same trick without relying on others to remove the caveats and hedges for you. You can add the caveats and hedges further down in the paper — maybe in a section for that particular purpose (“Limitations”). Most people don’t read the whole paper, so won’t see them. And even if they do, I suspect that many won’t change their interpretation of the key claims as much as one ought to. If the strong, uncaveated version is front and centre in the paper, I think that that’s what most readers will remember, even if they at some point saw caveats and hedges.

*

Such hedge-based motte and baileys are quite different from Shackelian motte and baileys and motte and bailey requests. The latter two are based on ambiguity. And you don’t need to be ambiguous to convey what you want to say. If you want to say that some relevant form of killing is morally equivalent to any other form of killing, as Garfinkel puts it, you can say that explicitly. You don’t need to merely imply it through saying “killing is killing”. So we can demand that people be explicit. We can prevent ambiguity-based motte and baileys in academic writing through having norms against ambiguities that enable motte and baileys. Shackel is effectively arguing for such norms.

But removal of caveats and hedges is different. We do need brief summaries, where most or all of the caveats and hedges are removed. For instance, we need titles and abstracts. So we can’t have a blanket rule against removal of caveats and hedges. The solution must be more nuanced than the solution to ambiguity-based motte and baileys.

Brief summaries are supposed to convey the gist of the full story. Conversely, the full story should fill in the details of the brief summary. Importantly, those additional details should be relatively symmetric, in the sense that if some of them pull the story in one direction, then others should pull them in another direction. It shouldn’t be the case that all of these additional details make the story less striking, compared with the brief summary. If that’s the case, then the brief summary doesn’t convey the gist of the full story, but rather an exaggerated version of it.

So the norm shouldn’t be that we can’t have brief summaries, but that those brief summaries should convey the gist of the full story. If so, in an important sense there aren’t two different versions, but just one version, which you stick to. The full version just fleshes out the details of the brief summary. That’s in general the solution to motte and baileys: norms that we stick with one version, instead of jumping around between several. No doubt there are other forms of motte and baileys besides those that I’ve covered here, but they all have one thing in common: that people jump between different versions; between motte and bailey. If you have to stick with one version, then you can’t make a motte and bailey.

Slightly edited 5 June.