# Understanding models and valuations

I'm thinking about part (a) of the following exercise in Just/Weese page 77:

Here is the definition of valuation:

For example, say we have a model of the language of group theory, $( \mathbb Z/ 2 \mathbb Z, +, 0)$. Let $\varphi = \forall v_0,v_1: v_0 + v_1 = v_1 + v_0$ and let $s: \omega \to \mathbb Z / 2 \mathbb Z$ be the map $s(n) = 0$ for all $n$. Then we should have $( \mathbb Z / 2 \mathbb Z, +, 0) \models_s \varphi$ but I am confused about what happens to variables under a given valuation. For the valuation I defined above the formula becomes $\varphi = \forall 0,0: 0 + 0 = 0 + 0$. Which is true but what is "$\forall 0,0:$" supposed to mean? Am I misunderstanding what a valuation is? If yes, would someone correct my example? Thanks for your help.

-

You’ve not actually given the definition of valuation, but I suspect that you’ve misunderstood what it says about bound variables. If $\varphi$ has no free variables, then $\mathfrak A\vDash_s\forall v_i\varphi(v_i)$ if $\mathfrak A\vDash_s\varphi[a_i]$ for all $a_i\in A$ according to the definition that you did give. (This may not be exactly how they do it, but it should be similar.)

-
Thanks, I corrected it! – Rudy the Reindeer Nov 14 '12 at 10:31
@MattN.: And it’s essentially what I thought, though you have to go through $\lnot\exists\lnot$ instead of having it defined directly for $\forall$. – Brian M. Scott Nov 14 '12 at 10:45
I'm still confused about the exercise though: shouldn't $s^\ast$ be related to $s$ somehow? Otherwise the following would provide a counter example: $\varphi = \exists x(x = 1)$ with the model $(\mathbb Z, +)$ with $s$ sending everything to $1$ and $s^\ast$ sending everything to $0$. Then $\mathfrak A \not\models_{s^\ast} \varphi$ and $\mathfrak A \models_{s} \varphi$. – Rudy the Reindeer Nov 15 '12 at 9:13
@MattN.: (Note that you need $1$ in your language even to have that $\varphi$.) Look at the final clause of the definition of a valuation, the one that deals with the existential quantifier: your $\mathfrak A\vDash_{s^*}\varphi$, because $s^*$ doesn’t determine the value of the bound variable $x$. – Brian M. Scott Nov 15 '12 at 21:06
Yes, the $s^\ast$ in that last clause assigns same values as $s$ except for $k$. But in the exercise $s^\ast$ seems to be an arbitrary valuation. Perhaps not? – Rudy the Reindeer Nov 15 '12 at 21:14

The main thing is that you should not substitute the given values in a variable which is bound by any of the quantifiers $\forall,\ \exists$:

Note that the definition of $\mathfrak A\models_s \exists v_i \varphi$ doesn't use at all $s(v_i)$, it's defined as there exists $s^*$ which is a modification of $s$, possibly $s(v_i)$ is replaced to any other value, such that $\mathfrak A\models_{s^*}\varphi$. This wants to mean exactly that 'there is a value for $v_i$ such that $\varphi$ with that value becomes true' -- fixing the rest of the evaluation $s$.

-

You should read the definitions properly when you read them. For example, the case $\exists$ tells you that $\mathfrak A \models_s \exists v_i \varphi$ if and only if there exists a valuation $s'$ such that $s'$ assigns the same values as $s$ to all other variables and assigns a value to $v_i$ such that $\varphi$ holds.

Similarly, you can get the $\forall$ case: $\mathfrak A \models_s \forall v_i \varphi$ if and only if for every valuation $s'$ that assigns the same values to all variables except $v_i$ you have that $\varphi$ holds in $\mathfrak A$.

In your example with $\varphi = \forall v_1, v_2 (v_1 + v_2 = v_2 + v_1)$ with the valuation $s(n) = 0$ for all $n$ it certainly is true that $0 + 0 = 0 + 0$ (as the rules tell you, you don't replace the variables in the quantifier, it makes no sense as you note in the question).

-