Good to continue from the family-resemblance challenge. Last lesson established a logical and semantic point: a category can appear to work through overlapping similarities rather than a feature shared by every member. The unresolved question is metaphysical: does that show that there is no objective common nature to discover, or only that some words do not capture one through a neat definition?
This lesson makes the strongest version of the contrast clear. The classical realist treats successful concepts as tracking mind-independent sameness in reality. The family-resemblance, practice-centred alternative treats at least many categories as sustained by learned public use, salient similarities, and human purposes. Crucially, these positions do not map onto each other quite as neatly as the slogan suggests.
Three questions that must not be conflated
When philosophers argue about concepts, they often slide between three distinct questions:
| Question | Classical realist tendency | Practice-centred tendency |
|---|---|---|
| Ontology: What exists? | Real shared natures, properties, or universals may exist in things. | Perhaps only particular things, their relations, and our classificatory practices exist. |
| Semantics: What makes applying a term correct? | The term is correct when it tracks the relevant objective nature. | Correctness depends on established public standards of use, shaped by purposes and examples. |
| Cognition: How do people form and use concepts? | The mind abstracts an objective commonality from instances. | We learn to apply words through examples, correction, and participation in practices. |
The first question is strictly metaphysical; the second concerns meaning and normativity; the third is psychological. They are connected, but they are not interchangeable.
For present purposes, take a universal to be one and the same qualitative nature that can be instantiated in multiple distinct objects. On a realist picture, two red apples do not merely happen to resemble each other: each instantiates redness, or shares a real basis for being classified as red. The word “red” is not what makes them alike; it succeeds because it answers to their likeness.
Classical Theory of Concepts, The
The Internet Encyclopedia of Philosophy’s “Classical Theory of Concepts” gives a compact map of the metaphysical options hidden behind ordinary talk of “having the same concept.”
In subsection “b. Concepts as Universals,” begin at “Concepts are also generally thought to be universals.” Read the metaphysical map. Focus on the distinction between realism, nominalism, and conceptualism: a concept can be treated as distinct from its instances, reducible to instances or linguistic items, or dependent on minds.
One immediate clarification: conceptualism is not the same as conventionalism. A conceptualist says concepts are mental entities or capacities. A conventionalist account says that the standards governing some classifications depend fundamentally on shared social practices. A person could accept both views, but neither entails the other.
The realist-classical package: concepts discover real sameness
The classical conception is most naturally paired with a moderate form of realism, associated with Aristotle rather than necessarily with Plato.
A Platonic realist says that universals exist independently of their concrete instances: the Form of Dog, for example, does not depend on any actual dogs. An Aristotelian or in re realist makes a more modest claim: the common nature exists in the individual things. Chihuahua and Great Dane are both dogs because each has canine nature, even though their accidental features differ dramatically.
Moderate Realism (Aquinas 101)
“Moderate Realism” from the Thomistic Institute’s Aquinas 101 offers a short historical orientation to the three poles: Platonic realism, nominalism, and the Aristotelian-Thomistic middle position. Treat it as a clear conceptual map rather than as a neutral verdict on the debate.
Watch the core question, which introduces the puzzle of recognising very different dogs as members of one kind. Then watch the two extremes for the contrast between independently existing Forms and mere naming. Finish with moderate realism, focusing on the claim that a form is in individual things and is abstracted by the mind.
On this package, a sound classical definition is not merely a convenient rule we invent. It aims to articulate the structure that makes the instances the kind of thing they are. Necessary and sufficient conditions matter because they are supposed to identify a real boundary in the world, not merely impose an administrative boundary on it.
For a classical realist, the explanatory direction is:
- Particulars possess a shared nature or structure.
- That shared nature grounds their similarity and their membership in a kind.
- Our concept succeeds when it identifies that nature.
- A definition, if available, expresses what has been discovered.
This is the background assumption behind the stronger reading of Rand’s measurement omission. The mind omits measurements such as size, colour, or weight; it does not manufacture the units’ identity. The concept is mental, and the linguistic label is conventional, but its objective basis is not. There must be some real commonality that makes these units appropriate to group.
That last qualification matters. A realist need not claim that every ordinary word has a useful or exact definition. The stronger classical ambition is that, where a concept identifies a genuine kind, there is a fact about what its instances have in common that does not depend on our voting, preferences, or current linguistic habits.
This makes disagreement potentially a matter of discovery. If two biologists disagree about whether a specimen belongs to a species, the realist hopes there is a truth of the matter grounded in the organism’s nature, lineage, or structure, even if neither scientist yet knows it.
The family-resemblance package: classification through practice
Wittgenstein’s challenge begins with a refusal to assume that the explanatory direction above always applies. We competently call chess, poker, tag, solitaire, and children’s games “games”; yet the attempt to identify a single common essence may repeatedly fail.
In that setting, the fact that one word applies across cases does not establish that all cases instantiate one universal. What holds the category together may instead be a network of salient similarities, learned through examples and maintained through shared practices.
The chair panel illustrates an important subtlety. Perhaps all the pictured objects afford sitting; but that broad feature may be too inclusive to distinguish chairs from benches, toilets, car seats, saddles, or rocks used as seats. Conversely, features that better distinguish some chairs exclude others. The point is not that chairs share literally nothing. It is that no single feature may perform the explanatory work demanded by an essential definition.
The family-resemblance position therefore relocates the source of unity:
- We learn paradigmatic cases.
- We learn which comparisons are relevant and which are not.
- We correct mistaken extensions of a word.
- We adjust precision when a particular purpose requires it.
On this view, the correct use of “game” is not an individual’s private preference. It is constrained by a public practice. Calling a pencil a game is simply a misuse; calling a structured form of digital play a game may be a harder case, resolved by precedent, purpose, analogy, and perhaps explicit stipulation.
That is the sense in which this approach is conventionalist. For many categories, convention is not just a name pasted onto a fully articulated natural division. Shared human activity partly establishes which resemblances matter and how sharply a boundary needs to be drawn.
A useful practical analogy comes from institutional concepts. Whether a transfer is “settled” or “final” is not settled by the physical properties of a database entry alone. It depends on the relevant payment-system rules, legal framework, operational procedures, and the purpose of the classification. Those rules are not arbitrary, and the underlying transactions are not unreal; nevertheless, the status is partly constituted by an institution.
The conventionalist claim about concepts is analogous: the world constrains our practices, but it does not always supply one uniquely privileged taxonomy independently of those practices.
Family resemblance does not automatically prove nominalism
It would be too quick to infer:
This concept lacks a classical definition; therefore universals do not exist.
Wittgenstein’s example of game undercuts a universal claim about the form of all concepts. It does not by itself establish a general theory of reality. A person could consistently hold all of the following:
- some categories are family-resemblance categories;
- some other categories track genuine natural structures;
- a concept’s everyday use can be fuzzy even if science later identifies a more exact explanatory kind.
Likewise, nominalism is more varied than “there is no real similarity, only words.” Some nominalists reject universals while retaining objective particular qualities, objective causal patterns, or primitive resemblance relations. They deny that one numerically identical universal must be shared by many things; they need not deny that things can be genuinely alike.
Nominalism in Metaphysics (Stanford Encyclopedia of Philosophy)
The Stanford Encyclopedia of Philosophy’s “Nominalism in Metaphysics” sharpens the ontological issue. It is particularly useful for separating rejection of universals from the far stronger claim that there are no objective qualitative similarities at all.
In Section “1.1 Nominalism as the Rejection of Universals,” begin with the red apple and green avocado example. Read the account of universals. Focus on the question of whether one qualitative nature can be present in more than one distinct object. Note the intermediate option of tropes: particularised properties that may be objectively similar without being one shared universal.
This yields the most important correction to the course’s shorthand contrast:
- Classical realism says that successful concepts often answer to mind-independent common natures and that classical definitions aim to identify them.
- Family resemblance says that some concept applications may instead be governed by overlapping similarities within a practice, without a common essence.
- Conventionalism is one way to develop the latter view: it treats shared human practices as fundamental to determining relevance, membership, and boundary precision for at least some categories.
- But family resemblance alone is not a proof of global conventionalism or a proof that the world lacks real structure.
The image of a family helps here. A family lineage may be objective, while resemblance among relatives is heterogeneous. Similarly, objective relations and similarities may exist even when our category does not correspond to one common property.
What, exactly, is at stake?
The disagreement can now be stated at its strongest.
| Issue | Classical realist answer | Family-resemblance conventionalist answer |
|---|---|---|
| Why are several things grouped under one concept? | They share a real nature, essence, or objectively privileged structure. | They stand in overlapping relevant similarities sustained by a practice of use. |
| What makes a classification correct? | Correspondence to the nature of the things classified. | Competent application within a public practice, constrained by cases, purposes, and the world. |
| What is a definition for? | To articulate the essence or real membership conditions of a kind. | Often to guide use; its precision should match the task, not satisfy an a priori demand for essence. |
| What explains a difficult borderline case? | Ignorance of the nature, imperfect measurement, or perhaps vagueness in our language. | The practice may not have fixed a determinate answer at that level of detail. |
| How can categories change? | Better inquiry may correct an earlier misunderstanding of the kind. | Practices can evolve, and institutions can refine or revise criteria as needs change. |
Consider the question: Is a new activity involving competition, virtual environments, rules, and monetary stakes a game?
The classical realist asks: what feature makes activities games rather than merely adjacent phenomena? If there is an essence of gamehood, the answer should eventually be answerable by discovering whether the activity has that feature.
The family-resemblance conventionalist asks: how is “game” used in the relevant context? Which established cases is this activity appropriately compared with? Is the question about leisure, regulation, education, gambling law, or software design? Different purposes can legitimately call for different precisifications.
Neither answer amounts to “anything goes.” The realist is constrained by the object’s nature; the conventionalist is constrained by public usage, established paradigms, institutional consequences, and practical success. Their dispute concerns the ultimate source of the norm governing the classification.
Two symmetrical dangers
The realist faces a burden that Wittgenstein makes vivid: one must not simply assume that every familiar general term conceals an essence. An enormous disjunctive list of cases, or a vacuous commonality such as “things we call games,” does not qualify as a discovered nature.
The conventionalist faces an equally serious burden: why are some similarities relevant? If a chess match resembles poker in having rules, poker resembles roulette in involving chance, and roulette resembles a weather forecast in involving uncertainty, why does the chain stop at games rather than spread without limit?
The conventionalist answer appeals to training, purposes, and use. We inherit a practice in which some comparisons support the extension of a term and others do not. But that answer can feel unsatisfying to a realist because it appears to explain stability by reference to what a community does rather than to what the objects are.
This is the real pressure point. The disagreement is not simply about whether definitions can be neat. It is about whether stable conceptual unity must be grounded ultimately in mind-independent identity, or whether it can be grounded in socially organised patterns of use without any hidden common core.
Conclusion
The core contrast is now sharper:
- A classical realist holds that at least successful kind-concepts track objective common natures. Definitions aim to discover and state those natures.
- A family-resemblance account denies that common-word use always requires such a shared essence. Some categories work through overlapping similarities, paradigms, and learned practices.
- A conventionalist development of the family-resemblance view says that shared practices partly determine which similarities count and how precise a boundary must be.
- Yet family resemblance does not entail that all classification is arbitrary, nor does it refute realism about every kind. It may be entirely appropriate for institutional and functional categories while being inadequate for categories that track stable natural structures.
Next, we turn from metaphysical and semantic claims to empirical psychology: Rosch’s prototype theory and the evidence that human categorisation often exhibits graded membership and typicality effects.