Is Gold a Real Category or Just a Word We Agreed On? The Answer Matters
By the BrainSnail editorial team. How these articles are written and checked, and how to tell us when one is wrong.
Some groupings appear to carve the world at its joints while others are conveniences we invented, and telling which is which turns out to be extremely difficult.
The distinction being drawn
Some categories seem to correspond to genuine divisions in the world rather than to human purposes. Gold, water and electron look like groupings that would exist whether or not anybody had noticed them, because members share an underlying nature that explains their common properties. Weeds, furniture and things heavier than a kilogram do not, since their members have nothing in common beyond the criterion used to assemble them. The claim is that the first sort supports reliable generalisation and prediction while the second does not.
What such a category is supposed to have
Several features are usually required together:
- •Members share an underlying essential structure
- •That structure explains their shared observable properties
- •The category supports generalisation to unexamined members
- •It appears in genuine laws of nature
- •Its boundaries are discovered rather than chosen
- •Scientists investigating independently would converge on it
Why it matters for meaning
The idea did substantial work in a famous argument about how words refer. If a word like water picks out whatever shares the underlying nature of the stuff around us rather than whatever matches our description of it, then people who knew nothing of chemistry were still talking about the same substance we are, and the meaning of the word was not fixed by what anybody had in mind. Hilary Putnam and Saul Kripke developed that position in the 1970s, and it changed how philosophers think about reference, about necessity and about what scientific discovery actually discovers.
Why the question is not merely academic
Decisions with real consequences turn on whether a category is genuine or constructed, which is the practical stake in an argument that looks abstract. Medical diagnosis assumes that conditions are real groupings rather than lists of symptoms that happen to co-occur, and psychiatry has spent decades arguing about exactly that. Legal categories of race and of disability raise the same question with obvious consequences. Conservation law protects species, which requires knowing what a species is. And machine learning systems discover clusters in data that may or may not correspond to anything beyond the data.
Where the idea struggles
Biology is the standing embarrassment for the whole framework. Species look like the clearest examples of genuine natural divisions and turn out not to have essences, since members of a species vary continuously, boundaries between species are genuinely fuzzy, and biologists use several incompatible definitions depending on the question. Chemistry cooperates better and still has awkward cases, including isotopes and mixtures. Many philosophers have therefore weakened the requirements, allowing a category to count if its members share a cluster of properties that tend to occur together, without requiring any essence at all.
The takeaway
Some categories appear to track real divisions that support prediction, with members sharing an underlying nature, while others are conveniences assembled for a purpose. The distinction underpinned an influential account of how words refer to substances regardless of what speakers know about them. Biological species resist the framework entirely, which has pushed many philosophers towards a looser version.