Wednesday, January 6, 2010

How do you think about natural kinds?

Although there is not consensus about what would make a natural kind natural, most traditional views agree that naturalness is a monadic feature; ie, "K is a natural kind" can be true or false of a given kind without specifying any further parameters. Call this the monadic presumption.

A few philosophers of science have denied this assumption and insisted that a kind is only a natural kind relative to a specified enquiry; ie, it's a relation of the form "K is a natural kind for E." (Proposals of this kind have been made by Dupre and Boyd.)

Consider an example like 'race.' There is no essential biological difference between members of different races, and so it may be tempting to say that race is not a natural kind. This flatfooted conclusion that race is not a natural kind only makes sense given the monadic presumption. On the relational conception, all that follows is that race is not a natural kind for biology.

A sociologist trying to understand social stratification and discrimination in the US South (for example) might need to recognize race, at least in some form. If so, then race would be a natural kind for that sociological enquiry.

It's tempting to say that race is not a natural kind because we want to deny the bogus rationale for discrimination. Recognizing race as a natural kind for sociology doesn't undercut that, however, since the sociologist's 'race' category couldn't justify the practices that it is used to explain.

To take a different example, biological kinds will not be natural kinds for particle physics - but they are nevertheless natural kinds for appropriately specified enquiries.

Although Dupre proposed a relativized conception of natural kinds over twenty years ago, the monadic presumption is still alive; eg, Bird and Tobin, in the SEP entry on Natural Kinds, simply presume it.

What I'm wondering is whether you, reader of this blog, consider the monadic presumption to be the default view of natural kinds. How heterodox is the relativized conception? Do you even consider the relativized conception when you think about natural kinds?


  1. Hi P.D. --

    How (if at all) do you think the relativized conception fits with the standard Kripke-Putnam account of how natural kind terms acquire their reference? Are baptisers indexed to type of enquiry as well?

  2. I thought the focus on natural kinds was the focus on a metaphysical fact about some thing. For instance, under the monadic view water is H20, regardless of inquiry. It would seem to me, that the relativized conception about natural kinds stipulate that to be a natural kind is to be observed by an agent. This leads me to believe that there are two implied counter-intuitive consequences. First, a natural kind, say 'race,' can be a natural kind for one discipline and not for another (or it can be of a completely different structure). Second, natural kinds are only natural when humans exist.

    I admit I am ignorant of the relavent debate, and possibly I am understanding the situtation incorrectly. However, if I am correct, then I do not see how the relativized notion is not heterodox.

  3. I take it that the more antirealist your approach to natural kinds is--the more you take facts about them to depend on our interests as inquirers, rather than being brute, metaphysical facts about the world--the more you'll be sympathetic to the domain-relative rather than monadic account of natural kinds. While I'm sympathetic to this sort of antirealist picture, I suspect it's not the majority one.

  4. @Daniel - I would think that the anti-realist account presumes the monadic one (i.e., scientific kinds are enquiry-relative, and natural kinds are brute, metaphysical facts about the world, so there are no real natural kinds). And Dupré explicitly says that he is a realist, as Bird and Tobin point out in the article P.D. linked above.

  5. Maybe perspectives on natural kinds depend on one's starting point in philosophy of science. If one begins at physics, it might be more tempting to see a certain fixity in the identity of objects in the universe (like the molecular structure of water) and hence a monadological conception of natural kinds. If one follows Dupre and your own example of race, PD, and starts from biology, then perhaps greater sympathy for the relativized view is engendered. Thinking about natural kinds in this way might make views on identity the determining factor on views towards natural kinds: depending on one's attitude towards change and complexity, one might be more or less sympathetic to the monadological view. I tend to come at things from a perspective closer to the biological, so you can guess where I fall.

  6. I don't tend to think of natural kinds in a binary way; naturalness seems to me to come in degrees. It seems to involve clusters of interesting, correlated similarities, and the bigger the clusters, and the more interesting and tightly correlated the similarities, the more natural the kind is. I hadn't really thought of putting it in relativized terms, but obviously my version of things would seem to lend itself to things being more or less natural depending on which similarities you were interested in.

  7. I don't like the idea of non-monadic "natural kinds". It sounds too much like "Relative to E, X is absolutely Y." I mean, what does "absolutely" mean there? I think a relative-to-E "natural" kind is similar.

    We might be able to say that "race" in a biological sense is not a natural kind, but "race" in a sociological sense is a natural kind. Here were are not talking about the same kind relative to two different kinds of enquiry, however, but simply two different kinds that happen to use the same word as a label.

    But ultimately I'm a kind of Kantian realist about monadic natural kinds. The world, fundamentally, is indeed differentiated into kinds. But all our categories are inexorably cultural, inexorably human (that includes molecular and even sub-atomic structures). We carve up the world in enquiry-relative ways all the time, but along "fibres" or "joints" that are unknowable to us.

  8. This comment has been removed by the author.

  9. Thanks to everyone for replying. Two remarks:

    First, Greg asks: 'How (if at all) do you think the relativized conception fits with the standard Kripke-Putnam account of how natural kind terms acquire their reference? Are baptisers indexed to type of enquiry as well?'

    Separate baptisms would only be required if 'race' as used by eugenicists and sociologists of eugenics are different words. The relativized conception of natural kinds requires that the word pick out a scientifically legitimate category for one enquiry but not for another, which is compatible with it being the same word in both contexts.

    Second, many of the replies seem to assume that the relativized conception of natural kinds is antirealist. It doesn't need to be.

    Thomas writes: 'We carve up the world in enquiry-relative ways all the time, but along "fibres" or "joints" that are unknowable to us.'

    This is entirely compatible with the relativized conception of natural kinds. There are so many joints in the world that we can't carve the world up along all of them at once. So different enquiries may require cutting along different joints.

  10. P.D., perhaps we can achieve your aims while maintaining monadicism. If biologists don't recognize race as a natural kind, it is because there are no biological properties that all and only members of any putative race shares. E.g., there is no genotype or pattern of reproductive relations that marks out 'caucasian' uniquely. If sociologists, however, recognize race as a natural kind, it is because there are sets of properties including non-biological ones that all and only members of corresponding races share. 'Race' is then being used ambiguously. There is bio-race, which is apparently non-natural (or very weakly natural?) from the perspective of any field, and socio-race, which is natural (or strongly natural?) from the perspective of any field. The biologists happen not to care much about the socio-races, because they are individuated in part by properties they don't study. But this doesn't mean biologists must count socio-races as any less natural. (All of this can be adjusted to fit with recent attempts to divorce natural kinds theory from traditional essentialism).

  11. Suppose I stack ten lego blocks on top of one another, five blue blocks and then five red ones. I then split it into two equal stacks, one blue and one red. I'm here "carving" along a "natural" joint (between the fifth and sixth blocks, the last blue one and the first red one). For the purposes of a particular enquiry I might say that the world divides into blue and red kinds of stuff. And this is of course actually also true. But the "joint" is not "really" between the blue and red stacks of blocks but between each individual block.

    The division into red and blue "kinds" does follow the "real" structure of the ten-block stack, but that real structure would also have allowed us to "break off" other kinds of objects (some of which would have been both red and blue). The original, colour-oriented enquiry did not discover this underlying structure (i.e., the blocks "themselves"), only the truth that there are blue kinds and red kinds.

    When I say I'm a Kantian about this I mean that I think that parts of reality can be "broken off" along pregiven (i.e., "natural") boundaries that determine the possible enquiries of dogs, monkeys, humans, and alpha centaurians alike. But the inquirer will not, ultimately, know what the natural fibres of reality are. The known kinds are never the natural ones we might say.

    Knowing is, finally, an artificial business. That is, I don't think my view is compatible with a relativized conception of natural kinds. Enquiry brings knowledge. No known kinds are natural.

  12. PD, when you wrote that, "There are so many joints in the world that we can't carve the world up along all of them at once" you captured my thoughts exactly. The problem with the world being such a complex place is our own epistemological difficulty in divining its structure. Perhaps the best way to go at this is to adopt one of Putnam's (admittedly varied) views where we can be realists about the fact that there is a structure of the world with joints to be cut, but our abilities to know and cut them properly is limited. This doesn't entail antirealism, as you pointed out, but it does entail a mismatch between our metaphysics and epistemology that's hard to resolve.

    My next question, then, is to ask: are thusly committed to this mismatch if we do not buy into a more traditional attitude regarding natural kinds? And, if so, how do we resolve the gaps between what our picture of the world and our fragmented, limited knowledge of it?

  13. Thanks for the interesting thread everyone. For what it is worth, I think the traditional natural kinds picture involves two elements: (i) kinds are not mind-dependent, but exist independently of our interests and concerns - that is, there are objective "joints"; (ii) kinds are relatively *sparse* - I take it that this is largely what makes them useful for theories of reference and especially interesting to metaphysicians.

    An anti-realist about kinds (say a nominalist) might deny both of these, insisting that we are free to classifying things in any way we like, and so it doesn't make sense to suppose that there are some privileged kinds waiting out there to suck up reference, or to be disinterestedly investigated.

    If that's right, then there seems to be room for a theory (which I take it some people here are working towards/developing, including PD I think) that seems to me to simply not fit neatly into either the realist or anti-realist camp. It would be realist-like in suggesting that there are objective, mind-independent kinds, but also anti-realist-like in denying that there are *sparse* kinds independently of our concerns/frameworks/disciples etc.

    That seems to me actually quite a reasonable position to take as it promises to combine the virtues of both traditional realism views and traditional anti-realist views. (Although I suspect that those who have been led to natural kinds by Kripke style concerns won't be satisfied even by such semi-natural kinds.)

  14. As far as I know (and I was working on a critique of biological essentialist views of 'races' as natural kinds), the sociologist would not recognize race as much as they would socially construct race - it's not clear by any means that races are directly referential in the naive sense. Sure, there are psychological studies that babies can recognize differences in contrasting skin colors, but I would strongly argue that one cannot tell the 'race' of another strictly by looking at them.

    I really don't think that race would be a natural kind for sociology, since the term 'natural' is loaded with a biologically essentialist component. In order for something to be a 'natural' kind for sociology, one would have to come up with necessary and sufficient conditions (I do believe) for what constitutes a 'natural sociological phenomenon'. I'm not too sure that there's a non-circular way to do that.

    Sociologists might have to posit 'races' not because they are actually there, but because they are using an IBE model, and the theoretical division of the human population into 'races' would help to scientifically explain (as much as it possibly could) the phenomenon of social stratification.

    I consider (for my research) that most competent English language users think of natural kinds in the monadic sense. I'm also not too sure that the term natural kind can be relativized.

  15. Jeffrey, I think the point against sparse kinds might be a good direction to head if we're going to avoid anti-realism. It can be connected to a topic in my comments above regarding the acknowledgment of complexity in the world and our understanding. It's also an echo of an objection I just found in Sandra Mitchell's latest book, where she criticizes Kim's view on emergence by arguing that he conflates "compositional materialism... with descriptive fundamentalism" (2009, 33).

    In these comments, there seems (unsurprisingly) to be agreement on assuming compositional materialism, or a monadicism about the world. Your point about sparse kinds, Jeffrey, is a similar one in that compositional materialism can give rise to a variety of taxonomies. These taxonomies are not necessarily artifacts of limited understanding, as I might have been taken to mean, but might be tied to an actual plurality of kinds. The upshot of this might be a direction of positive research that reveals the world to be a richer place than previously assumed.

  16. It strikes me that we need to distinguish between:
    (1) P.D.'s proposal that 'K is a natural kind' is underspecified (like 'Hilary is taller'), because 'is a natural kind' is a dyadic predicate, so that the complete specification is 'K is a natural kind for E' (like 'Hilary is taller than Jo').
    (2) 'K is a natural kind' is fully specified; 'is a natural kind' is monadic. But it is sometimes useful to indicate which part of science informs us that K is a natural kind and describes the properties of K.

    Assuming, for sake of argument, that there is something to P.D.'s race example, then the relevant issues can be understood by adopting (2). In such a case there would be epistemic utility to indicating that sociology discovers the relevant kinds and describes their properties (and consequently the basic properties of the kinds are going to be sociological rather than biological properties).

    Consider for example 'sodium is a natural kind'. Is this really underspecified? Do we have to understand this as really saying 'sodium is a natural kind for chemistry'? If so, what do we make of the propositions 'sodium is a natural kind for biology' or 'sodium is a natural kind for astronomy'? On the dyadic view is seems quite unclear. On the one hand, it is neither biology nor astronomy that determine that sodium is a natural kind. On the other hand both sciences make use of the fact that sodium is a natural kind (in cell biology and stellar spectroscopy). Best to say that 'sodium is a natural kind' is true simpliciter, while adding that this fact is one discovered by chemistry and exploited by related sciences.

  17. I think the motivation for the relational view of kinds can be best seen where we might take different inquiries to disagree about the constitution of the "same" kind, or about whether the kind is a natural kind or an accidental arrangement.

    To take a Dupré-inspired example, think about the kinds "fruit" and "vegetable" in botany versus gastronomy. "Tomato" switches categories relative to the inquiry we're in. We can remove the appearance of disagreement by adopting (1). (This seems appealing to me when I imagine a biologist and a gastronome arguing about whether a tomato is "really" a vegetable.)

    Obviously, the different classifications of "species" among different parts of biology are prime candidates for such relativizations.

    Thinking about sodium, it might have been the case that, from the point of view of our best particle physics, sodium was an accidental arrangement of fundamental particles (my understanding is that this is not actually the case). Nevertheless, it might be quite clearly a natural kind from the point of view of our best chemistry. In such a case, (1) would again be appealing.

  18. I think Alexander Bird's distinction between (1) and (2) is one way of capturing what I said earlier (see "A." at Jan/7/10, 10:36). My suggestion was in line with choosing (2) rather than (1) as the best way to interpret apparent inter-disciplinary disagreements about certain kinds. To apply this in the more recently mentioned cases of species and tomatoes, the idea is that there isn't really disagreement about a single kind in either of these cases. Rather, the kind terms are ambiguous in each case. 'Species' for example is multiply ambiguous. For some time now biologists have been remedying this by using prefixes. They speak of bio-species, eco-species, morpho-species, etc. To more thoroughly remove the ambiguity, we could just drop the term 'species' in favor of, e.g., 'bio-unit', 'eco-unit', 'morpho-unit', etc. Then it seems plain that a researcher whose projects tend to concern "natural" bio-units could nonetheless grant that other researchers are interested in other sorts of things that are also "natural", such as eco-units and morpho-units. Why must we say that naturalness is discipline specific in the way that (1) does? The examples we're discussing are natural from the perspective of each discipline; they just happen to be more interesting in some disciplines than others. Naturalness is typically discipline-independent, salience is not.

  19. Thanks to everyone for commenting. The take away lesson for me is that the view I describe is not seen as one of the standard options. I also need to think more about Alex's distinction.

  20. The take-home lesson for me is that you people are trying to impart a specific meaning to a word which is used non-specifically by non-scientific people.

    All the linguistic and logical jumping through hoops won't be useful against people to whom the word means whatever they want it to mean at the time.