Consider sets of Vitali's type in models of $\mathsf{ZF}+\mathsf{GCH}$ where $V \neq L$. Are there sets of Vitali's type in both $L$ and $V \backslash L$? If so, is there any way one can distinguish the constructible sets of Vitali's type from the nonconstructible sets of Vitali's type?

By a set of Vitali's type it is meant a subset of $\mathbb{R}$ containing exactly one element of every equivalence class of the relation $x - y \in G$, where $G$ is some fixed countable subgroup of $( \mathbb{R} , + )$.

  • $\begingroup$ Could you clarify what is meant by a set of Vitali type? For example, I think if $G$ is any countable subgroup of $(\mathbb{R}, +)$, then choosing representatives under the equivalence relation $x - y \in G$ will give a "Vitali set relative to $G$". The rationals are one instance. $\endgroup$
    – Avshalom
    Dec 30, 2014 at 18:52
  • $\begingroup$ @Avshalom: Exactly so. What you wrote is the definition of sets of Vitali's type, relative to $G$. As you point out, $\mathbb Q$ is the usual subgroup of $($$\mathbb R$,$+$$)$ used to define "Vitali sets" . My question to you is, does any choice for $G$ make the set of representatives under the equivalence relation $x$$-$$y$$\in$$G$ nonmeasurable with respect to the Lebesgue measure? The phrase "sets of Vitali's type" is used as a synonym for "Vitali sets". $\endgroup$ Dec 30, 2014 at 22:04
  • $\begingroup$ (With respect to $\mathbb Q$, that is....) $\endgroup$ Dec 30, 2014 at 22:11
  • $\begingroup$ Let $G$ be a countable subgroup. If $x_E$ is the equivalence class of $x \in \mathbb{R}$, then $\mathbb{R} = \bigcup_{g \in G}(x_E + g)$; now apply $\sigma$-additivity and translation invariance of Lebesgue measure. Will that work? The argument will apply more generally. $\endgroup$
    – Avshalom
    Dec 30, 2014 at 22:26
  • $\begingroup$ If so, then the countable subgroup $\langle G, \tau \rangle$, where $\tau$ is a generic real, should give rise to new Vitali sets that are not in the ground model too. $\endgroup$
    – Avshalom
    Dec 31, 2014 at 0:26

1 Answer 1


There are several things to say that may answer your question.

  • It is consistent with ZFC$\pm$GCH that the set of all real numbers in $L$ is countable in $V$, in which case every set of reals in $L$ has measure zero.

  • It can happen that $V$ and $L$ have precisely the same sets of reals, yet $V\neq L$, because they disagree only very high up. In this case, of course there can be no set of reals in $V-L$.

  • Meanwhile, if by a set of "Vitali type" you mean a set that selects one member of each equivalence class under the relation of translation by rationals, then if $L$ has such a set, then it must have all the reals (since every real differs by a rational from a member of it), and so $\mathbb{R}^V=\mathbb{R}^L$.

  • Also, one can force to add a new Vitali set without adding reals, and so one can have ZFC+GCH+$V\neq L$ with Vitali sets in both $L$ and in $V-L$.

  • $\begingroup$ @Prof. Hamkins: Thank you for that very nice answer--it is very helpful. The case that interests me most is the last one, that is, the case where one forces to add a new Vitali set without adding reals so that one can have $ZFC+GCH$+$V$$\neq$$L$ with Vitali sets in both $L$ and $V-L$. In this case, is there a way to distinguish the Vitali sets in $L$ from those in $V-L$? I ask this question to see if it makes sense to speak of an 'evolving continuum'--that is, a continuum that grows larger to meet the needs of evolving mathematical practice. $\endgroup$ Dec 30, 2014 at 15:34
  • $\begingroup$ Actually, I should really say that evolving mathematical practice might force one to construct a new Vitali set not in $L$--sorry. As for the 'evolving continuum' one could always form new Vitali sets by adding reals so as not to violate CH, right? $\endgroup$ Dec 30, 2014 at 15:56

Your Answer

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

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