When Dmitri Mendeleev arranged the elements into his periodic table in 1869, he did more than create a useful chart. He predicted the existence of elements not yet discovered, based on gaps in the pattern. Gallium, scandium, and germanium were later found with properties remarkably close to his predictions.

This raises a profound question: did Mendeleev discover a structure that nature itself possesses, or did he invent a convenient organizational scheme? The answer matters enormously. If scientific categories like electron, mammal, or chemical element track real divisions in reality, then science is genuinely uncovering nature's joints. If they are merely useful fictions, the picture of scientific knowledge changes dramatically.

Plato suggested that good thinking divides reality the way a skilled butcher carves an animal—along its natural joints, not through arbitrary cuts. This metaphor captures a central question in philosophy of science: do scientific categories reflect divisions that exist independently of us?

Consider the difference between gold and jade. Gold is a single element with atomic number 79, and its properties follow from this underlying nature. Jade, however, refers to two chemically distinct minerals—jadeite and nephrite—grouped together by cultural convention. One category tracks a real boundary in nature; the other reflects human interests in appearance and use.

The scientific realist holds that successful sciences progressively identify genuine kinds. Chemistry's elements, biology's species, and physics' fundamental particles are not arbitrary groupings but reflect structures that would exist whether or not anyone classified them. When we mistake a conventional grouping for a natural one, our predictions and explanations falter.

Takeaway

Some categories cut reality at its joints, while others impose lines that serve our purposes but tell us little about how the world actually works.

Philosopher Nelson Goodman noticed something puzzling about induction: not every property supports reliable generalization. If you discover that several samples of copper conduct electricity, you can confidently project this to all copper. But knowing that several objects in your kitchen are green tells you little about other green things.

The difference lies in what Goodman called projectibility. Properties tied to natural kinds support inductive inferences because they connect to underlying causal structures. Copper conducts electricity because of its atomic configuration—a feature shared by every sample. Greenness, by contrast, can result from countless unrelated causes and predicts nothing else.

This is why natural kinds are the workhorses of scientific reasoning. When biologists study one population of a species, their findings often generalize because membership in that species reflects deep similarities in physiology and ancestry. Science succeeds, in part, because it identifies categories whose members share enough underlying structure to make prediction possible.

Takeaway

Reliable prediction depends on identifying categories whose members share genuine causal structure—not just surface similarity.

What makes water water? Hilary Putnam argued that water is essentially H₂O—a substance with that molecular structure would be water even on a distant planet, and any substance lacking it would not be water no matter how watery it appeared. The essence lies in the underlying structure, not the surface properties.

Identifying essential features transforms scientific understanding. Once we know that water is H₂O, we can explain why it boils at 100°C, dissolves salt, and expands when frozen. The essence grounds the explanations. Without it, we have a list of correlations; with it, we have a theory.

This is how mature sciences progress. We move from classifying tigers by their stripes to understanding them through evolutionary lineage and genetics. We shift from cataloging diseases by symptoms to identifying their molecular causes. Each shift reveals that surface features were tracking something deeper—and that the deeper feature is what science was after all along.

Takeaway

Scientific understanding deepens when we move from surface features to the underlying structures that make things what they fundamentally are.

The question of natural kinds is not merely academic. It shapes how we interpret scientific success and what we expect from future inquiry. When science identifies genuine kinds, it gains predictive power, explanatory depth, and the ability to discover what we did not yet know to look for.

Mendeleev's empty cells were filled because the periodic table tracked something real. The challenge for any science is distinguishing categories that carve nature from those that merely organize our experience—and recognizing that the difference matters for what we can know.