If the union of two theories is not consistent there exists a sentence $P$ for which one of them implies $P$ and the other one implies $\lnot P$.

Can someone please check if this is true or not?

I used to prove this by contradiction. I assumed for every sentence $P$, if $T_1$ implies $P$, then $T_2$ satisfies $P$ or $T_2$ doesn't imply $P$ but never $T_2$ implies $\lnot P$. So i think there exists a model of $T_2$ as a substructure of a model of $T_1$( i'm not sure if this is true... I just say it intuitively ), so I take that model of $T_1$ as a model of $T_1 \cup T_2$. Since the union doesn't have a model so there exists a sentence $P$ such that $T_1$ implies $P$ and $T_2$ implies the $\lnot P$.

  • 1
    $\begingroup$ Your terminology is not standard. A theory is typically a collection of assumed statements or axioms. Rather than saying a theory satisfies a proposition, we might say a theory implies (or proves) a proposition. We do speak of a model (interpretation) satisfying a proposition. $\endgroup$
    – hardmath
    Feb 20, 2018 at 22:41
  • $\begingroup$ Thanks i edited it. $\endgroup$
    – user470412
    Feb 20, 2018 at 22:51
  • $\begingroup$ Is "implies" referring to "formally provable from the theory", or is it in the sense of entailment that "every model satisfying $T$ also satisfies $P$"? $\endgroup$ Feb 20, 2018 at 23:39
  • $\begingroup$ I mean every model satisfing T also satisfies P $\endgroup$
    – user470412
    Feb 20, 2018 at 23:56

1 Answer 1


If the language of both theories is a first-order language, so that the compactness theorem applies, then the statement is true.

So, suppose $T_1 \cup T_2$ is inconsistent. Then, by compactness, there is a finite set of axioms from $T_1 \cup T_2$ which is inconsistent. Let this finite set of axioms be $\phi_1, \ldots, \phi_m \in T_1$ and $\psi_1, \ldots, \psi_n \in T_2$. Then if the contradiction entailed by $\phi_i, \psi_j$ is $Q \wedge \lnot Q$, then we get both $T_1 \vdash (\psi_1 \wedge \cdots \wedge \psi_n \rightarrow Q)$ and $T_1 \vdash (\psi_1 \wedge \cdots \wedge \psi_n \rightarrow \lnot Q)$. It follows that $T_1 \vdash \lnot (\psi_1 \wedge \cdots \wedge \psi_n)$, whereas of course $T_2 \vdash \psi_1 \wedge \cdots \wedge \psi_n$.


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.