Proposition 1.6.3. A commutative square of categories is a pullback square if and only if applying \(\Map(T,-)\) gives a pullback square for every category \(T\).
Proof
We first show that applying \(\Map(T,-)\) to the chosen pullback \(C\times_E D\) results in a pullback square. Postcomposition with the two projections and whiskering with \(\tau_{\univ}\) gives a comparison functor \[ \Map(T,C\times_E D)\longrightarrow \Map(T,C)\times_{\Map(T,E)}\Map(T,D). \] For an inverse, let \(P\) be the anima on the right. Uncurrying the projections to \(\Map(T,C)\) and \(\Map(T,D)\) gives maps from \(P \times T\) to \(C\) and \(D\). Moreover, the identification in the commutative square for \(C \times_E D\) results in an identification of two functors \(P \times T \to E\). The pullback axiom thus supplies a functor from \(P \times T\) to \(C\times_E D\), with the projection comparisons and their compatibility with the squares. Currying gives a candidate inverse functor \(P \to \Map(T,C \times_D E)\).
To compare the composite on the right to the identity, we first do so on its two projections, where it follows by currying the identifications we just constructed. To show that the resulting two commutative squares are identified, we may again do so after uncurrying, where it follows from unwinding definitions.
For the composite on the left, we need to show its uncurrying agrees with the evaluation map. This can in turn be tested after both projections and comparing the commutativity identifications.
The claim now also follows for an arbitrary pullback square by considering the equivalence from its cone point to \(C\times_E D\).