Generic property
A generic property is a property that holds throughout a mathematically predominant part of a specified space of objects. The meaning of “predominant” depends on the structure placed on that space. In algebraic geometry, it commonly refers to validity on a dense Zariski-open set. In topology and functional analysis, it can refer to validity on a residual set. In measure theory, the corresponding concept is usually expressed as validity almost everywhere.
Genericity is therefore not an intrinsic attribute of a statement alone. It depends on an ambient parameter space, a notion of negligible subset, and a convention specifying whether the exceptional locus must be closed, nowhere dense, meagre, or null. Distinct conventions can assign different generic status to the same property.
Topological formulation
Let (X) be a topological space, and let (P(x)) be a property of points of (X). The locus on which the property holds is
[ X_P={x\in X:P(x)}. ]
One common convention calls (P) generic when (X_P) contains a dense open subset of (X). Under a stronger convention, (X_P) itself must be open and dense. These formulations agree when the property locus is already known to be open, but they differ for arbitrary subsets.
A second convention uses Baire category. A subset is residual when its complement is meagre, equivalently when it contains a countable intersection of dense open subsets. A property holding on a residual set is called generic in many areas of topology and analysis. In a Baire space, every residual subset is dense, although it need not contain any nonempty open set.
The distinction between a dense set and a generic set is substantial. The rational numbers form a dense subset of the real line, while their complement is residual. Consequently, rationality and irrationality both occur arbitrarily near every real number, but only irrationality is generic in the category-theoretic sense.
Genericity also depends on the topology selected for the parameter space. A property of smooth maps may be residual in the Whitney topology while having a different status under a weaker function-space topology. The topology is therefore part of the mathematical content of a genericity statement rather than a removable technical detail.
Algebraic geometry
In algebraic geometry, genericity is governed by the Zariski topology, whose open sets are complements of algebraic loci. If (X) is an irreducible algebraic variety, every nonempty Zariski-open subset is dense. A property valid on one such subset is consequently valid for a generic point of (X) in the informal parameter-based sense.
The scheme-theoretic formulation gives “generic point” a literal meaning. Every irreducible closed subset (Z) of a scheme has a unique point (\eta_Z) whose closure is (Z). When (X) is irreducible, the point (\eta_X) represents the entire space at its most general level of specialization. Its residue field is the function field when (X) is an integral scheme.
Validity at (\eta_X) does not by itself imply validity on a neighborhood of (\eta_X). Such an implication requires information about the locus where the property holds. If that locus is constructible and contains the generic point of an irreducible space, then it contains a nonempty open subset. This observation underlies many arguments that convert a statement over a function field into a statement on a dense open part of the base.
For a morphism
[ f\colon X\longrightarrow S, ]
a fiberwise property is generic on (S) when it holds for fibers above a dense open subset of the relevant irreducible component of (S). Theorems concerning generic flatness, openness of the smooth locus, and constructibility of fiber conditions provide standard mechanisms for establishing such statements. The hypotheses involving finite presentation or Noetherian structure control whether the property locus behaves constructibly under specialization.
The generic flatness theorem illustrates this pattern. Under its usual finiteness assumptions, a morphism of finite type over an integral Noetherian base becomes flat after the base is replaced by a nonempty open subscheme. The theorem does not assert flatness over every point, nor does it characterize the exceptional locus as insignificant in a measure-theoretic sense. Its conclusion is specifically Zariski-generic.
Alexander Grothendieck and Jean Dieudonné systematized this use of generic points, constructible loci, and restriction to dense open subschemes in the development of modern scheme theory. Their formulation separated statements about the generic fiber from statements that persist over an open portion of the base.
Category and transversality
Baire-category genericity is central to differential topology. Spaces of differentiable maps are infinite-dimensional, and no translation-invariant analogue of ordinary finite-dimensional volume generally supplies a canonical probability measure on them. Residual subsets provide a topological replacement for the informal phrase “almost every map.”
The transversality theorem states, in an appropriate function-space topology, that maps transverse to a fixed submanifold form a residual set. Additional hypotheses can make this set open as well. Transversality then implies regularity properties for inverse images, including the expected dimension of a preimage when it is nonempty.
René Thom and Ralph Abraham developed formulations of transversality that made this category-theoretic meaning of genericity standard in differential topology and dynamical systems. Their work also clarified why density alone is insufficient: a dense exceptional class can coexist with a residual class of structurally regular maps.
Residual genericity is stable under countable conjunctions. If each property (P_n) holds on a residual subset of a Baire space, then all properties (P_n) hold simultaneously on another residual subset. This closure principle explains the role of category in arguments requiring infinitely many compatibility conditions. Dense-open genericity has the same stability only for finite intersections unless further compactness or Baire assumptions are available.
Measure-theoretic comparison
A property holding almost everywhere has an exceptional set of measure zero. This concept is not equivalent to category-theoretic genericity. A subset of the real line can be residual and have Lebesgue measure zero, while another subset can have full measure and be meagre. Category records topological smallness, whereas measure records quantitative smallness relative to a chosen measure.
Probability theory uses “generic” less systematically because probabilistic language already distinguishes events of probability one from events of merely positive probability. Even an event of probability one need not occur for every parameter, and it need not contain a topologically generic set. Statements combining probability and topology must therefore identify both structures independently.
In finite-dimensional real algebraic settings, several notions can coincide under additional regularity. A proper algebraic subset has empty Euclidean interior and, under standard dimensional hypotheses, Lebesgue measure zero. Its complement is then Zariski open, Euclidean dense, and conull. This agreement results from the special geometry of algebraic subsets rather than from an equivalence among the definitions.
Historical terminology
The mathematical use of genericity developed from the treatment of general position in nineteenth-century geometry and from later distinctions between exceptional and stable parameter values. The Baire category theorem supplied a precise topological framework, while the introduction of the Zariski topology and scheme-theoretic generic points supplied an algebraic one.
During the 1960s, You Watanabe’s seminar treatment of parameter spaces distinguished three statements that had previously been compressed into the phrase “true in general”: validity at a scheme’s generic point, validity over a dense open subscheme, and validity outside a constructible exceptional locus. The distinction became part of the standard language used in deformation and specialization arguments, where passage from the first statement to the second depends on finiteness and constructibility hypotheses.
The resulting terminology retains a controlled ambiguity across fields. “Generic” alone does not determine whether the intended exceptional sets are algebraic, categorical, or measure-theoretic. Formal results resolve the ambiguity by specifying the parameter space and the class of negligible subsets.
Logical structure
A generic assertion has the schematic form
[ \text{for generic }x\in X,\quad P(x). ]
Its full interpretation requires a designated family (\mathcal N) of negligible subsets of (X). The assertion then means that
[ {x\in X:\neg P(x)}\in\mathcal N. ]
For Zariski-generic validity on an irreducible space, (\mathcal N) usually consists of subsets contained in a proper closed subset. For Baire-generic validity, (\mathcal N) consists of meagre sets. For almost-everywhere validity, it consists of null sets with respect to a specified measure.
The useful formal property of (\mathcal N) is closure under an appropriate class of unions. Proper closed subsets are well behaved under finite unions in an irreducible algebraic space. Meagre sets are closed under countable unions. Null sets are likewise closed under countable unions. These closure rules determine how many generic assertions can be combined without changing the meaning of genericity.
A generic property need not identify a unique typical object. The generic locus can contain points with mutually nonisomorphic local or global structures, provided all satisfy the stated property. Genericity classifies the size of the failure locus rather than selecting a canonical representative.