Struck by a semirealism

Thu 12 Feb 2015 09:57 AM

In a number of recent articles, Bence Nanay has argued for singularist semirealism. It's an anti-realist view about natural kinds which holds that particular tokens of properties exist with various degrees of similarity and dissimilarity among them, but that there are not any natural property types. The view is similar to Anjan Chakravartty's semirealism, which holds similarly that the world consists of property instances more or less sociable with one another, and that the clusters of sociability which science picks out are not somehow special in nature.

Nanay writes:

Some pairs of property-tokens are closer together in the property-space; they resemble each other more than others. But property-types are our arbitrary ways of delineating regions of this property-space. The property-space does not have joints: it consists of lots of property-tokens, some close together, some further away from each other. (2013, p. 377)

His approach seems to be more deeply metaphysical than mine. Nanay is most centrally concerned with whether a natural kind is a thing in the world that exists. I am concerned centrally with the extent to which the world constrains scientific categorization. I am happy to say that categories which uniquely allow successful science would be natural kinds regardless of whether there is an entity the deep ontology of the world which corresponds to that category. I am also willing to allow that kinds can be more or less natural, to the degree that the world condemns alternative taxonomies to failure.

Nevertheless, Nanay argues that singularist semirealism coheres with scientific practice. The reason is "that the two main tools of actual scientific practice, experimentation and measurement, are practices involving property-tokens and not property-types" (2011, p. 189; 2013, p. 383). This seems wrong to me for at least two reasons.

First: If a scientist were given a table of data which was just numbers or magnitudes, she'd have no use for it. Measurements necessarily have units. So measuring the masses and lengths of 10 samples necessarily requires measuring the masses as masses and the lengths as distances. Each singular property property must fit into a category scheme, and so measurement is impossible without kinds.

Second: It ignores the distinction between what Bogen and Woodward call data and phenomena. Singular measurements are data which are always subject to error and variation. Although data play an important evidential role, scientists don't primarily care about data. They care about phenomena which data instantiate. The phenomena are the curves or patterns which we think the data would trace out if it weren't for noise and error. When scientists repeat an experiment, they do not expect to produce precisely the same data as earlier experiments. Rather, they expect to get data which (once reduced by standard formal methods) will yield the same phenomena. So measurement and experiment are about general phenomena-types rather than singular data-tokens.


Bogen and Woodward 1988. Saving the Phenomena. PhilRev, 97(3): 303-352.

Chakravartty 2007. A Metaphysics for Scientific Realism, Cambridge University Press.

Nanay 2011. What If Reality Has No Architecture? The Monist, 94(2): 181-197.

Nanay 2013. Singularist Semirealism. BritJPhilSci. 64: 371-394.


from: Derek C.

Fri 20 Feb 2015 06:20 AM

Hey, PD.

Interesting post, but here's a couple of questions.

First of all, do you think that Nanay would be perfectly happy to say "Well, if your criteria for a natural kind being real is just that natural kinds are, in some sense, constrained by the way the world is, then I would agree with you. Of course, that doesn't mean they are things that exist in the deep ontology of the world." He does say that "property types are our arbitrary ways of delineating this property-space," but surely he doesn't mean totally arbitrary. For instance, wouldn't he agree that there are reasons--and not just pragmatic reasons--for grouping together mallard-tokens under the type mallard vs, say, grouping together all bear-tokens and mallard tokens under the type mallabear? Perhaps all he means by 'arbitrary' is that the boundaries are fuzzy, but would you disagree with that?

Second, why does he have to ignore the distinction between data and phenomena? Could he not concede that what scientists really care about are the phenomena and still go on to argue that this does not mean measurement and experiment are about phenomena? In other words, he might argue that, since you only measure or experiment on data-tokens, measurement and experimentation are, to that degree, about data-tokens. And so data-tokens are what Really exist. Now it's true that scientists care more about the curves and patterns which can be abstracted away from these data-tokens and that these curves and patterns can be used in any number of applications. Nonetheless, these are mere constructions and do not, in any deep sense, exist.

What do you think? As an aside, I'm not entirely sure I've understood your argument and I haven't read either of the Nanay articles.

from: P.D.

Fri 20 Feb 2015 08:10 AM


Regarding your first point, that might be right. Nanay just be denying that kinds are ultimate elements of the deepest ontology. But it's not clear to me that single property tokens will meet the strictures of fundamental ontology, either. Maybe the only real thing is the entirety of being. In any case, this ceases to be a view in the philosophy of science.

Regarding the second point: If scientists are interested primarily in phenomena and phenomena aren't real, then the immediate consequence is that scientists aren't primarily interested in real things. I take it that Nanay wants to avoid that consequence, and that's why he appeals to measurement and experiment as singular.