4

Full question: Suppose that $\models \phi \supset \psi$, and that $\phi$ is not a contradiction nor $\psi$ a tautology. Show that there is a propositional variable variable $p$ that occurs in both $\phi$ and $\psi$.

My initial thought was just to chase definitions around, but after doing so I'm not seeing where to assert that there is a propositional variable. What should I appeal to? Recursion theorem?

Just a small push, please.

Noah Schweber
  • 245,398

1 Answers1

2

By contradiction: Suppose $\phi$ and $\psi$ have no variable in common, but still $\models \phi \supset \psi$. Let $V_{\phi}, V_{\psi}$ be the variables of $\phi$, $\psi$ respectively.

Because $\phi$ is not a contradiction, there is some assignment of truth values $val_{\phi}\colon V_{\phi} \to \{0,1\}$ to the variables of $\phi$ such that the value $\phi^{val_{\phi}}$ of $\phi$ under $val_{\phi}$ is 1.

Similarly, because $\psi$ is not a tautology, there is some assignment of truth values $val_{\psi}\colon V_{\psi} \to \{0,1\}$ to the variables of $\psi$ such that the value $\psi^{val_{\psi}}$ of $\psi$ under $val_{\psi}$ is 0.

Since $V_{\phi}$ and $V_{\psi}$ are disjoint, the valuations $val_{\phi}$ and $val_{\psi}$ can be extended to a common valuation $$val_{(\phi \supset \psi)} \colon V_{(\phi \supset \psi)} \to \{0,1\} $$ where $V_{(\phi \supset \psi)} = V_{\phi} \cup V_{\psi}$ is the set of variables of $(\phi \supset \psi)$, and where $$ val_{(\phi \supset \psi)} \restriction V_{\phi} = val_{\phi} \\ val_{(\phi \supset \psi)} \restriction V_{\psi} = val_{\psi} $$

So the value of $\phi \supset \psi$ under the valuation $val_{(\phi \supset \psi)}$ is $0$. But supposedly $\models \phi \supset \psi$, so the value of $\phi \supset \psi$ under any valuation is 1. Contradiction.

BrianO
  • 16,579
  • Thanks, but next time could you just provide a hint, not the entire proof. –  Oct 31 '15 at 23:54
  • Sorry about that :) I forgot you had asked for just a nudge as I was working it out. You can recast the argument to minimize the "by contradiction" aspect. Essentially: there are valuations that satisfy $\phi$, and there are valuations that falsify $\psi$. No pair of those can have a common extension to the variables of $\phi \supset \psi$, as such an extension would falsify $\phi \supset \psi$. If the variables of $\phi$ and $\psi$ were disjoint, this wouldn't be the case, as any pair of valuations could then be so extended. – BrianO Nov 01 '15 at 00:16
  • Cool. Thank you for your help. –  Nov 01 '15 at 00:38
  • You're welcome. – BrianO Nov 01 '15 at 00:42