I have long been confused about this notion. I know that for a statement within a single category, forming the dual statement is just reversing every arrows. But what about a statement concerning several categories and functors between them?
Mac Lane suggests reversing all arrows in all categories and leaving the functors invariant in his book "Categories for the working mathematician" (page 32), but it seems that Mac Lane himself doesn't always follow this discipline (when he states "the dual of Yoneda Theorem", he doesn't reverse the arrows in $\mathbf{Set}$ !) Can anyone give me a formal definition of "dualize a statement in category theory"?
Best Answer
Some of the categories in a statement are "variables" (e.g. "$C$"), and you should take the opposite of those categories but not any categories which are "constants" (e.g. $\text{Set}$). The point is that any category which is a "variable" is being quantified over, so you can always replace that category with an opposite category, in the same way that you can substitute anything you want for a variable in an identity.
For example, if a functor $F : C \to D$ is a left adjoint, then it preserves colimits. The dual of this statement is obtained by replacing $C$ and $D$ with opposites, since they are both "variables," and you get that if $F : C^{op} \to D^{op}$ is a left adjoint, then it preserves colimits. But this is equivalent to a statement about $C$ and $D$ themselves, which is that if $F^{op} : C \to D$ is a right adjoint, then it preserves limits.