The Universe of Discourse

Fri, 25 Jan 2008

Nonstandard adjectives in mathematics
Ranjit Bhatnagar once propounded the notion of a "nonstandard" adjective. This is best explained by an example. "Red" is not usually a nonstandard adjective, because a red boat is still a boat, a red hat is still a hat, and a red flag is still a flag. But "fake" is typically nonstandard, because a fake diamond is not a diamond, a fake Gucci handbag is not a Gucci handbag.

The property is not really attached to the adjective itself. Red emeralds are not emeralds, so "red" is nonstandard when applied to emeralds. Fake expressions of sympathy are still expressions of sympathy, however insincere. "Toy" often goes both ways: a toy fire engine is not a fire engine, but a toy ball is a ball and a toy dog is a dog.

Adjectives in mathematics are rarely nonstandard. An Abelian group is a group, a second-countable topology is a topology, an odd integer is an integer, a partial derivative is a derivative, a well-founded order is an order, an open set is a set, and a limit ordinal is an ordinal.

When mathematicians want to express that a certain kind of entity is similar to some other kind of entity, but is not actually some other entity, they tend to use compound words. For example, a pseudometric is not (in general) a metric. The phrase "pseudo metric" would be misleading, because a "pseudo metric" sounds like some new kind of metric. But there is no such term.

But there is one glaring exception. A partial function is not (in general) a function. The containment is in the other direction: all functions are partial functions, but not all partial functions are functions. The terminology makes more sense if one imagines that "function" is shorthand for "total function", but that is not usually what people say.

If I were more quixotic, I would propose that partial functions be called "partialfunctions" instead. Or perhaps "pseudofunctions". Or one could go the other way and call them "normal relations", where "normal" can be replaced by whatever adjective you prefer—ejective relations, anyone?

I was about to write "any of these would be preferable to the current confusion", but actually I think it probably doesn't matter very much.

[ Addendum 20080201: Another example, and more discussion of "partial". ]

[ Addendum 20081205: A contravariant functor is not a functor. ]

[ Addendum 20090121: A hom-set is not a set. ]

[ Addendum 20110905: A skew field is not a field. The Wikipedia article about division rings observes that this use of "skew" is counter to the usual behavior of adjectives in mathematics. ]

[ Addendum 20120819: A snub cube is not a cube. Several people have informed me that a quantum group is not a group. ]

[ Addendum 20140708: nLab refers to the red herring principle, that “in mathematics, a ‘red herring’ need not, in general, be either red or a herring”. ]

[ Addendum 20160505: The gaussian integers contain the integers, not vice versa, so a gaussian integer is not in general an integer. ]

[ Addendum 20190503: Timon Salar Gutleb points out that affine spaces were at one time called “affine vector spaces” so that every vector space was an affine vector space, but not vice versa. ]

[ Addendum 20221106: the standard term for this appears to be “privative adjective”. ]

[ Addendum 20240508: [Robin Houston asks if there is such a thing as an incorrect proof]( [Simon Tatham points out]( that a “manifold with boundary” is not a manifold. ]

[Other articles in category /math] permanent link