Today I’m interested in the—sometimes strained—relationship between two aspects of philosophical products: analysis and metaphor. An analysis is an account of some notion or problem we’re interested in. This often involves highly technical machinery and is as precise-as-can-be. Philosophical metaphors are often how we present, sell and label our analyses: metaphors are in a sense a matter of philosophical style; they’re what makes an otherwise dry, precise analysis vivid and fun. I’ve been thinking about this because, well, metaphors sometimes play a prominent role in my own work. As Derek Skillings says in his perceptive review of Rock, Bone and Ruin, “[Currie] never wastes an opportunity to use a turn of phrase”. I use metaphors, ‘bumper-stickers’ and other catchy stylistic things pretty often. Indeed, in Joyce Havstad’s challenging discussion of Rock, Bone and Ruin here in Extinct, she picked apart one of my metaphors: the ‘ripple-model of evidence’. In introducing this model, I ask us to imagine throwing a pebble into a lake and then taking snapshots of the ripples at subsequent intervals, and use this to think about evidence degrading over time. Joyce suspects something shifty has occurred with the metaphor:
There is sometimes a distance between the metaphors we use to explicate and make vivid our philosophical analyses, and the analyses themselves. And sometimes, as Joyce suspects with the ripple-model of evidence, these differences can be pernicious. So what is the value of philosophical metaphors and when are they valid? Instead of hitting Joyce’s criticism head-on today (that’s for another venue!), I thought I’d ask these questions in light of another metaphor, one extremely prominent in the philosophy of historical science: Carol Cleland’s ‘smoking gun’.
Let’s start with Cleland’s analysis. In a nutshell, Cleland wants to argue that there are two archetypical ways of going about your scientific business: experimentally or historically. The latter (which we’re interested in here) involves a three-step process. We notice a surprising correlation or phenomenon, we then generate a bunch of hypotheses about the past which could lead to that correlation or phenomenon, and we then search for further traces which support one of those hypotheses above the others. Consider Dartmoor, the rather sublime granite country in Devon, near where I live.
Dartmoor is distinctively boggy, split by rivers, and marked by ‘tors’, rounded hills covered in exposed granite boulders (in the photo there’s one in the distance). Dartmoor makes for fine walking country and is dotted with funky archaeological remains. What kind of events shaped this distinctive landscape? The traditional view has it that fairly undramatic processes of freeze-and-thaw shaped the tors: they are the mark of ice-age permafrost. It was thought that the glacial activity which shaped more northerly parts of Britain would have flattened the tors. However, in 2012 a local group of geologists argued that there is bountiful evidence for glaciers reaching this far south (see this rather delightfully drab BBC short). They point to classic signs of glaciers: drumlins, moraines, and glacial valleys. They also point out that the distribution of tors suggests that some were crushed by glaciers after all.
On a Cleland-style reconstruction, we have two hypotheses concerning Dartmoor’s distinctive landforms. By one, they were formed by permafrost, by another, by glaciation. According to recent studies, glaciation wins out because it explains both tor distribution, as well as the drumlins, moraines and so forth. As Cleland says,
So, historical reconstruction proceeds by picking hypotheses which best explain traces: permafrost explains the tors, but only glaciers explain how the tors are distributed, and the other glacial features. For Cleland, the mode of evidence testing is inference to the best explanation, especially focused on appeal to common causes. Appealing to a glacier is a better explanation than permafrost because it posits a single cause—glaciers—to accommodate a whole bunch of traces.
So much for the analysis, now for the metaphor.
Cleland names the traces which decide between hypotheses smoking guns. The notion of a ‘smoking gun’ appears to have originated in one of Conan Doyle’s Sherlock Holmes stories in 1893, but really entered public consciousness during Watergate. There, the notion of a ‘smoking gun’—something which would show incontrovertibly that Nixon had obstructed justice—became a very common way of thinking about legal evidence. A ‘smoking gun’ is, in popular parlance, something similar to what Nancy Cartwright has called a clincher: a piece of evidence which establishes a hypothesis. The June 23, 1972 Whitehouse tape, where Nixon agreed that officials should recommend halting Watergate, was taken to establish Nixon’s entering into a criminal conspiracy. It was the smoking gun. The contrast for Cartwright’s clinchers are ‘vouchers’, which simply support one hypothesis over another.
So, the smoking gun metaphor implies that historical scientists search for clinchers rather than vouchers. Before looking at the relationship between Cleland’s metaphor and her analysis, we should think a little more about philosophical metaphors.
Metaphors are stand-ins, they are a kind of representation for something. In the present discussion, should we consider philosophical metaphors as representing the target of analysis, or the analysis itself? That is, does Cleland’s ‘smoking gun’ represent historical method directly, or is it a representation of her view? I think the latter. The term ‘smoking gun’ for Cleland is not a short-hand for the method of historical science, but of Cleland’s analysis of that method. One reason for thinking this is that when I disagree with Cleland, I don’t argue that she has misconstrued smoking gun reasoning, but that she is wrong in thinking that smoking gun reasoning (that is, her analysis) captures how palaeontologists, geologists and archaeologists reason. The function of the metaphor is to make the analysis more vivid and understandable, not to represent the target of the analysis.
So, philosophical metaphors are stand-ins, simplified representations, of philosophical analyses. Is Cleland’s metaphor a good stand-in for her view? Note that Cleland is very clear that she doesn’t mean ‘smoking gun’ in a casual way:
Cleland then, has a special, technical meaning for ‘smoking gun’. What is that meaning? I think this is somewhat ambiguous, and has led to confusion about Cleland’s view. In earlier work in particular, it does look as if a ‘smoking gun’ is a clincher:
On this view, a smoking gun is the piece of evidence which establishes a hypothesis about the past. We see this reading in Patrick Forber and Eric Griffiths’ critique of Cleland. On their view a smoking gun is “... a naturally occurring experimentum crucis” (3, Forber & Griffith 2011). Infamously, a crucial experiment (experimentum crucis) is some experimental result which establishes beyond the shadow of a doubt one hypothesis over another. Newton claimed as much for the experiment in his first optical paper (Kirsten Walsh has a particularly clear discussion).
Notice the disconnect between Cleland’s analysis and her metaphor. The analysis didn’t say anything about establishing a hypothesis. It was rather a view that historical hypotheses are tested by their capacity to explain and unify new traces. The analysis doesn’t require that evidence clinched those hypotheses. This leaves us with two versions of Cleland’s position: a stronger and a weaker one. In the stronger version, we take the smoking gun metaphor seriously—smoking guns are clinchers. In the weaker version, we don’t—smoking guns are new trace evidence which help discriminate between hypotheses.
Personally, if I were Cleland I would definitely gun (as it were) for the weaker view. When we think about Dartmoor’s glaciers, it really doesn’t look like there is a single piece of evidence—a smoking gun—which makes the difference between the permafrost and glacial hypotheses. Rather, the scientists build a case for glaciers by showing how better they account for a variety of Dartmoor’s features. This fits the weaker view, and not the stronger view. Happily, in more recent work Cleland puts her view in explicitly weaker terms. For instance:
Here, the process of discovery seems to matter: the smoking gun is the piece of evidence which shifts the scientific community in favour of some hypothesis. If the drumlins were found first, then perhaps moraines would have clinched the hypothesis in favour of a glacial Dartmoor; if the morains were found first, then perhaps the drumlins would have clinched it. This is a more subtle notion. In the same piece, Cleland also clarifies what can count as a smoking gun:
So perhaps we should say that the drumlins, the morraines and the distribution of tors count as the ‘smoking gun’. After all, it was the capacity of glaciers to best explain all of those features which made it more attractive than permafrost.
But now we see that ‘smoking guns’ are simply new trace evidence which support one hypothesis over another. They don’t need to be a single piece of evidence, and they needn’t establish a hypothesis incontrovertibly (after all, they only support one hypothesis over another). The metaphor of a ‘smoking gun’, on this interpretation of Cleland, seems like a bad one. Cleland’s real view is actually relatively thin: historical science proceeds by testing hypotheses about the distribution of traces by looking for new traces, and evaluates hypotheses on the basis of their capacity to explain those traces. There might be occasional ‘clinchers’—smoking guns in the normal usage—but these are not definitional of historical science. Trace-based reasoning is.
What went wrong with Cleland’s metaphor? The metaphor implied things which the analysis did not. According to the analysis, the investigation of Dartmoor looks like boiler-plate historical science. According to the metaphor it doesn’t: where is the clinching evidence? This leads to problems. Forber and Griffith disagree with Cleland in part because they read her in the stronger sense. But that stronger view (even in her earlier papers) is not so well supported by how she presents her analysis (despite appeal to ‘so-to-speak cinching’).
There are two lessons I take from this discussion, one for the metaphor-producer, the other for the metaphor-consumer.
(1) If you use philosophical metaphor, ensure that the metaphor implies nothing which is not implied by the analysis.
The point of the metaphor, of course, is to make the analysis catchy, memorable, or to just have a nice label (‘Cleland’s view’ is not anywhere near as fun as ‘smoking guns’!). As such, it doesn’t have to imply everything the analysis does. My claim here is that it shouldn’t imply things the analysis doesn’t. It should imply some subset of what the analysis does, and hopefully in a way which captures the essential spirit of the analysis. This of course raises the question of what a metaphor does, and doesn’t, imply. Obviously Cleland’s metaphor doesn’t imply that historical scientists are literally looking for guns. But it does seem to imply that they’re looking for ‘crucial, deciding’ traces. Philosophers have said a fair bit about what metaphors are and how to decode them. For the purposes of this post, let’s just say that whatever view we have, it had better come out saying that (a) Cleland’s metaphor doesn’t imply that paleontologists are digging up recently-fired firearms and, (b) Cleland’s metaphor does imply a kind of ‘experimentum crucis’.
(2) Critique analyses, not metaphors.
One of my least favourite philosophical tendencies is arguing against a view’s title, or the metaphor it is presented in terms of, as opposed to the details of the actual view itself (that is, the analysis). Charitability is a cardinal philosophical virtue. And being charitable requires going beyond the bumper-sticker version of the view and interpreting the analysis itself. So, instead of getting hung up on labels or metaphor, we should understand and critique analyses.
Philosophical metaphors are I think pretty important. They make our work more striking and engaging (and making philosophy fun matters: philosophy is too important to be boring), it also potentially helps with understanding, shaping a quick model of the view in our minds. The best metaphors are generative. Reflecting on them brings us to a deeper understanding of the analysis. But as we’ve seen, the use of metaphor is also risky. I think we should use metaphors which only imply what our analyses make explicit, and critique the analysis, not the metaphor.
Have I always met these two guiding principles? I doubt it: there’s every chance I’m happily chucking stones all about this fine glass house I’ve built for myself. If Joyce is right, my metaphor in the ‘ripple model of evidence’ smuggles in a bunch of notions unjustifiably. I also wonder what other philosophical metaphors there are out there, and whether they meet my criteria or not. Regardless, the next time I think about throwing pebbles into a lake, I’ll make damn sure the metaphor is apt…