Adrian Currie writes...
An evolutionary perspective is often crucial for understanding the living world—ancestry combines organisms into lineages, each with their own unique, contingent histories. These histories are central to making life what it is. And central too are the evolutionary relationships between lineages: who is more related to whom, when did evolving populations split from one another, and so on. Without that comparative information, we’re flying blind.
But figuring out such relationships for extinct critters is really, really tricky. So tricky in fact, that one might argue that unless some fancy new technologies or techniques are developed, fine-grained knowledge of ancestral relationships between extinct organisms will remain severely limited. If there’s good reason to think we’re not going to know the ancestral relationships between extinct lineages, why do palaeontologists spend so long trying to figure it out anyway?
This is where I left us hanging in part 1.
I apologise for nothing!
My task today is to drag us from the cliff edge by suggesting an answer. It’s pretty speculative, but I think quite interesting. Indeed, if you stick with me through to the end, I’ll suggest a general strategy for exonerating scientific practices which appear hopeless.
It also involves some conceptual machinery, which I’m going to approach by simply presenting in all its jargony glory, before spending a bit of time unpacking it. With that in place, I’ll then give you some reasons to think that maybe it’s true. Sounds like fun? No? Well. Um. Sorry.
Hypothesis: practices of morphological phylogenetics in paleontology often express extrinsic epistemic values, by examining the consequences of our commitments to the evolvability of characters.
Clear, yes? No? Okay: let the unpacking begin...
Unpacking the Jargon
Morphological phylogenetics is the practice of inferring ancestral relationships between lineages on the basis of their morphological features: basically, how they look. Say we were considering these three crocodilian skulls:
Look at the differences and similarities between these skulls. Some ‘characters’—traits—are common across all three and serve to group them together; other characters are not shared by all three, and enable us to distinguish between them. For instance, two of the three are ‘longirostrine’, that is, they have elongated snouts compared with the third, who I’m going to call ‘Snubby’. Under the right circumstances (which we’ll return to below!), the possession or otherwise of such characters can be used to infer ancestry. Perhaps their being longirostrine is evidence that Snubby belongs in a different taxonomic category. That is, our two long-noses are more closely related to each other than to Snubby, who split off from their lineage earlier.
So, here’s the first bit of unpacking: morphological phylogenetics involves (1) inferring the ancestral relationships between organisms using (2) morphological characters as evidence.
Okay, then. What’s this talk of ‘epistemic values’, and when are they ‘extrinsic’? ‘Epistemic’ is a fancy way of saying ‘about knowledge’. Sometimes philosophers distinguish between epistemic values—things we value for their knowledginess, valuing ‘truth’ say,—and non-epistemic values, like say, um ‘friendship’ (aw). There’s reason to be pretty worried about that general distinction, but let’s not fuss about that now. ‘Extrinsic’ epistemic value is introduced in a nice paper by Dan Steel. Steel distinguishes between two different kinds of epistemic values: intrinsic and (you guessed it) extrinsic. Suppose I value precision: if I achieve that value, it seems like that in itself is an epistemic good. Precision, then, has intrinsic epistemic value. Now, perhaps I value having shiny, well-funded laboratories. Simply having shiny labs doesn’t in itself give me epistemic goods, but maybe it facilitates the getting of those goods. If it does, the valuing shiny labs is an extrinsic epistemic value.
Wrong kind of shiny lab - these are still potential bearers of extrinsic epistemic value, though!
Here’s the distinction from the horse’s mouth:
“… a value is intrinsically epistemic if exemplifying that value either constitutes an attainment of truth, or is a necessary condition for a statement to be true”, while extrinsic epistemic values “… promote that attainment of truth without themselves being indicators or requirements of truth” (18).
So, something has extrinsic epistemic value if my valuing it promotes the attainment of truths.
Let’s connect this discussion of value to phylogenies. To begin, we’ll have to make a slight modification to Steel’s distinction. By his account of extrinsic epistemic value, something can’t be extrinsically valuable if it is intrinsically valuable (they cannot ‘themselves be indicators or requirements of truth’). But if I get an accurate phylogenetic tree, that is an indicator of truth. But I think something can possess both extrinsic and intrinsic epistemic value, so long as the epistemic goods are different. Say, I might value accuracy, but in doing so also promote the attainment of good explanations. Or – as I’ll argue – valuing phylogenetic reconstructions promotes the understanding of the evolvability of characters.
Finally, to this ‘evolvability’ business. Evolvability is a tricky concept (like most concepts in biology – why can’t things just be simple at least once?), and I don’t want to get too complicated here (see Rachael Brown's excellent paper for the gory details). Here, I’ll go for an intuitive gloss.
Evolutionary Developmental scientists are (among other things) interested in explaining how characters transform over evolutionary time. How a scale might transform into a feather, for example. Such explanations are often presented as sequences where small developmental changes lead to changes in the character’s morphology (Brett Calcott calls these ‘lineage explanations’).
A transformative sequence of feather evolution (from http://people.eku.edu/ritchisong/feather_evolution.htm)
Some characters are easier to evolve than other characters, and some lineages are better at evolving particular characters than other lineages. The ‘evolvability’ of a character is, roughly, that character’s disposition to transform over evolutionary time. It’s at least possible for scales to turn into feathers, but it might not be possible (or it at least might be very hard) for mammalian hair to turn into feathers.
Enough unpacking
Let’s re-express my claim. Morphological phylogenetics doesn’t just teach us about the ancestry of extinct lineages. It also is extrinsically epistemically valuable. Specifically, it promotes epistemic goods pertaining to the evolvability of characters.
Ok then, how? To see my argument, let’s consider the evolvability of that longirostrine character—those long snouts—and how that relates to the reasoning underlying morphological phylogenetics. I’m basing my discussion on Eric Wilberg’s rather epic discussion.
As I summarized in Part 1, there is a bit of a phylogenetic mystery surrounding the Thalattosuchia clade, a bunch of marine crocodile-like not-crocodiles from the age of the dinosaurs.
Dimitry Bogdanov's reconstruction of Dakosaurus maximus, a representative Thalattosuchian. Not a crocodile.
The ‘longirostrine problem’ revolves around how related these critters are to true crocodiles who also boast long snouts. Specifically, this appears to be highly sensitive to how that character is treated, as well as (as Wilberg emphasizes) choice of outgroups. Let’s focus on a single paragraph (from Wilberg pp 8-9, I’ve removed the references because wow). He’s summarizing reasons for being suspicious that all of the longirostrines are closely related. I’ll quote each reason and I’ll summarize as we go (*deep breath*!).
“The three contentious longirostrine groups are all found in marine environments suggesting they share similar ecological/functional pressures. Early phylogenetic analyses recovered all three long-snouted groups as a clade. This grouping was treated sceptically from the beginning and was originally dismissed as the result of convergence.”
In early studies the marine longirostrines were unified in a single ancestral group. However, this was taken instead to be a convergence. That is, the long snouts were taken to have evolved separately—like bat wings and bird wings, for example—rather than signalling common ancestry. Wilburg goes on,
“Later works no longer dismissed the longirostine clade but it has remained suspect for three primary reasons. First, thalattosuchians possess the plesiomorphic condition for many characters (which must be optimized as reversals when sister to other longirostrine groups) …”
‘Plesiomophic conditions’ refer to the basal state – what the ancestor of all the crocodiles and crocodile-like-things looked like. This is another way of saying that the thalattosuchians possess ‘primitive’ traits. If they were closely related to true crocs, they would have had to evolve more derived traits, and then revert back to the primitive states later. This would be akin to a bird’s feathers re-evolving into scales.
“Second, because crocodyliforms have demonstrably evolved similar skull shapes numerous times, it was assumed that snout shape is not a reliable phylogenetic character.”
This is crucial: no matter how you shake the family tree, crocodile skull shape changes all the time (well, all the time over evolutionary time anyway). As I’ll expand on below, this suggests that skull shape is highly evolvable, and there’s a link between a character’s evolvability and its suitability for phylogenetic reconstruction. But we should finish the paragraph:
“Finally, the sister group relationship between thalattosuchians and [crocodilian longirostrines] requires a long ghost lineage… Additionally, analyses have demonstrated that when characters suspected of correlation with a slender snout shape are removed, thalattosuchians are recovered as either basal mesoeucrocodylians or as the sister group to Crocodyliforms”
The last two points are that if the long-snouts are closely related, then there’s a great big gap in the fossil record, and that their clustering like this depends on us taking snout shape as a character.
So, we should read the long-snouts of true crocodiles and of thalattosuchians as convergent, as opposed to a signal of ancestry—that is, we shouldn’t count them as phylogenetic characters—because:
· A bunch of primitive characters would need to have re-evolved;
· Crocodile skull shapes are highly evolvable;
· There’d need to be a massive hole in the fossil record; and
· The clustering depends on taking snout shape as a character.
Notice that whether or not we should take snout shape as a character depends partially upon its evolvability. If the character is highly evolvable, then it doesn’t make very good evidence for ancestry. To establish ancestry, you want characters to be relatively inert over evolutionary time—stable characters will be reliable indicators over time (see Arnon Levy and my discussion for more). This means that what I take as a character, and how I choose to weight those characters* (and here’s the crucial point) encode a set of hypotheses about the evolvability of characters. Those I include as characters, and weight high, are characters which I think would be relatively stable over evolutionary time. Those characters I exclude, or weight low, are the ones I expect to be highly evolvable.
Ok, so what? Well so far I’ve summarized a pretty complex discussion about the evolvability of snout-length in crocodile-like things, and the relationship between snout-length and other characters. Phylogenetic reconstructions use many characters. Wilberg’s analysis, for instance, used 394 morphological characters. That’s a lot of hypotheses about evolvability! Moreover, running various trees tells us about the dependencies between various characters – in other words, the dependencies between various hypotheses about character evolvability. Indeed, one reason to doubt that snout-length is a good character is that it requires that we say rather implausible things about other characters – lots of primitive characters would need to have re-evolved.
Here’s the clincher: even if we’re no closer to knowing the fine-grained ancestral relationships between crocodiles and crocodile-like things, the practice of phylogenetics has extrinsic epistemic value. It promotes our understanding the evolvability of characters by (1) motivating research into the evolvability of particular characters by identifying sensitivities between ancestral relationships and decisions about characters, (2) exploring the sensitivities between different hypotheses about character evolvability, (3) exploring the consequences of our hypotheses about character evolvability in an evolutionary context.
So, there’s my defence of morphological phylogenetics. Even if it doesn’t get us any closer to working out fine-grained ancestral relationships, it nonetheless plays an important role in structuring and motivating investigations of character evolvability. That is, it has extrinsic epistemic value.
It strikes me that looking to extrinsic epistemic value might be a general strategy for understanding apparently hopeless scientific strategies. Let’s sketch that.
A General Strategy for Exonerating the Hopeless
Often, when we consider whether some scientific practice is warranted or not, we focus on the explicit appearances of the investigation. We ask, do these studies really test for what they’re supposed to test? A further question we should ask is: how does this practice structure investigations—what epistemic goods does it indirectly promote? That is to say, in addition to asking after the intrinsic epistemic value of scientific practices, we should also chase up the extrinsic epistemic value.
One example (inspired by Marta Halina) is in comparative psychology. Various experiments are designed to tell whether or not various critters—parrots, corvids, mice, non-human primates, and so forth—have certain higher-level cognitive capacities. Do they, for instance, have a theory of mind? That is, does the animal attribute beliefs, desires, and so forth to other organisms with which it interacts? (A long time ago I wrote a little piece about this in kea which I still quite like.) The standard that appears to be set for achieving this is pretty hard-core: experiments are expected to show— pretty much conclusively, and almost on their own—that the only hypothesis which could explain the subject’s behaviour is such fancy cognition. That’s a high bar! Which is odd since, considering the evidence as a whole, it does seem pretty clear that various animals do indeed have such high-level cognitive capacities. What, then, could justify the practice of attempting to find a single, critical test to establish theory of mind in non-human animals?
Given our lesson here, one approach regarding the exoneration of the practice is to ask after the indirect goods which setting the bar so high might encourage. How does it structure the investigations which comparative psychologists engage in? I haven’t a clue about the answer to this question, (any thoughts?), or if there are any other examples of this (I’m all ears!).
Generally speaking, the lesson here is that, in considering the justification of various scientific practices, things are potentially more complex than they appear. The direct outputs, such as working out the phylogenetic relationships between lineages, are perhaps not where the action is. Instead, we might do well to understand some scientific practices in terms of what they indirectly achieve.
*Some phylogenetic analyses assign weight to characters, while others do not. Those that do not still use weights, but only implicitly!