In various categories one encounters what are referred to as 'test-objects'. For example, singleton set serves as test-object (in testing the equality of functions) in the category of sets. In the category of graphs, we need two test objects: dot D and arrow A to separate graph maps (Conceptual Mathematics, page 215; Exercise 1, page 250; page 340). I am wondering if it makes much sense to seek a definition of Test-Objects that works across various categories.

Defining Test-Objects seemed like an interesting exercise given that objects in categories such as sets and graphs are determined entirely by their test-objects (for example, D-figures and A-figures [along with incidence relations] in the category of graphs; see Conceptual Mathematics, page 245; Sets for Mathematics, pp. 154-5; http://at.yorku.ca/t/o/p/d/65.htm).

I eagerly look forward to your corrections and clarifications!


This is called a separating family or (regrettably) generating family for the category. Every category has a tautological separating family, namely the whole category itself; this gives the notion of generalised element.

However, not every category has a small separating family. For example, let $\mathcal{C}$ be a category such that there exists an object with a proper class of isomorphism classes of subobjects; then $[\mathcal{C}^\textrm{op}, \textbf{Set}]$ does not admit a small separating family. (If it did, then it would satisfy Giraud's axioms for a sheaf topos, but it is known that [$\mathcal{C}^\textrm{op}, \textbf{Set}]$ is not a topos when $\mathcal{C}$ is "too big".)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.