I am looking at the Yoneda lemma trying to see where the assumption of "locally small" really comes in. Obviously in order to define a functor to the category sets using $Hom$-spaces we need our $Hom$-spaces to be sets. However if we consider a enriched-category, enriched over some non-locally small monoidal category M, then for any element of the category, our $Hom$-sets give us a functor $Hom(A,-)$ to M.

In particlar, in the statement $$ Hom(Hom(-,A),F) \simeq F(A), $$ where $F$ is a set-valued functor, where does the assumption of "smallness" play a role.

In the answer to this question, it is stated that the category of sets can be replaced by any Grothendieck universe $U$. However, the definition of a Grothendieck universe assumes that $U$ is a set. Moreover, the enriched Yoneda lemma again assumes "smallness". In these answers, is smallness a necessary assumption?

  • 3
    $\begingroup$ Yes, there is an enriched Yoneda lemma: see this question, or this nLab page. $\endgroup$
    – varkor
    Nov 26, 2020 at 15:01
  • $\begingroup$ But in the nLab version they require that the monoidal category is locally small. This seems like a more special version rather than a generalisation. $\endgroup$ Nov 26, 2020 at 15:14
  • 3
    $\begingroup$ I would need to check more carefully, but I think you are right: I think the only size assumption are in order for the enriched ends defining $V^C(F,G)$ to exists. But assuming that C has a set of objects, or rather that $V$ has limits of size the set of objects (of isomorphism class actually) of C seem to be enough for this. It does not seems usefull to assume that $C$ or $V$ are locally small. $\endgroup$ Nov 26, 2020 at 16:08
  • 3
    $\begingroup$ I don't understand the question. Is it "why does the nLab page on the enriched Yoneda lemma assume that the monoidal category is locally small?" If so then I think Simon is right that it's not necessary; probably whoever wrote the page was just throwing in a sufficient set of assumptions rather than thinking about which were actually necessary. $\endgroup$ Nov 26, 2020 at 17:26
  • 2
    $\begingroup$ @Mike: Yes, this is precisely the question. I want to know if the assumption is necessary or not. If it is not, I would be much happier :) $\endgroup$ Nov 26, 2020 at 19:01

1 Answer 1


The Yoneda lemma is a purely formal result that does not require any size assumptions. For any closed symmetric monoidal category $\mathbf{V}$, any $\mathbf{V}$-category $C$, any object $A\in C$, and any functor $F:C\to \mathbf{V}$, there is an isomorphism

$$ [C,\mathbf{V}](よ^A,F) \cong F(A). $$

Here $よ^A$ denotes the hom-functor $C(A,-)$ and $[C,\mathbf{V}]$ denotes the $\mathbf{V}$-enriched hom-category. It is true that one needs $\mathbf{V}$ to have limits of the size of $C$ in order for $[C,\mathbf{V}]$ to exist as a $\mathbf{V}$-category, but even if this fails, the statement is true and provable in the following sense: if we write down the diagram whose limit would, if it existed, be the LHS, then the RHS is a limit of that diagram.

It is even possible to formulate and prove versions of the Yoneda lemma that do not require $\mathbf{V}$ to be closed or symmetric, and even that allow it to be a multicategory rather than a monoidal category. See, for instance, Lemma 5.29 of my paper enriched indexed categories, or proposition 8.2 of my paper with Richard Garner, enriched categories as a free cocompletion for the bicategorical case.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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