The definitions of Morley rank & degree I use are

Morley rank definition

Morley degree definition

I understand these definitions, but I am having a hard time to use them concretely in exercises. For example,

Let $L$ be a countable language and $A$ an $\omega$-saturated $L$-structure.

(a) let $p$ be a ranked type over $A$. Show that $p$ has Morley degree 1.

(b) Let $\varphi(x_1, \dots, x_n)$ be a ranked $L_A$-formula, show that

$$ \text{dM}(\varphi(\bar{x})) = |\{p \in S_n(\text{Th}_{L_A}(A)) : \varphi \in p \text{ and } \text{RM}(p) = \text{RM}(\varphi) \}| $$

I don't know how to connect $\omega$-saturation with Morley rank & degree. I thought to use that $\text{Th}(A)$ has an $\omega$-saturated model iff its type spaces are countable iff there is no binary tree of consistent formulas, in the hope that $\text{dM}(p) \geq 2$ would allow me to construct such a binary tree, but I cannot get it to work.

  • $\begingroup$ (1) "I don't know how to connect $\omega$-saturation with Morley rank & degree". The connection is that $\omega$-saturation is assumed in your definitions of Morley rank and degree. So you can't even talk about Morley rank and degree (using your definition) over models that aren't $\omega$-saturated. (2) "I thought to use that $\text{Th}(A)$ has an $\omega$-saturated model iff its type spaces are countable..." This is not true! Every theory has $\omega$-saturated models. The theorem is: $T$ has a countable $\omega$-saturated model if and only if its type spaces are countable. $\endgroup$ Jun 17, 2019 at 16:33
  • $\begingroup$ @AlexKruckman (1) Yes, but I am not sure why the $\omega$-saturated condition is even required, unless you want to talk about the Morley rank of a theory (RM of $x=x$). (2) Oh, yes of course. And we were not given that $A$ is countable. So then I am even more stuck. $\endgroup$
    – Ibrahim
    Jun 17, 2019 at 20:46
  • $\begingroup$ The reason for the $\omega$-saturated condition is to ensure the desirable property the Morley rank of a formula only depends on the theory, not the model $A$. Indeed, you can prove that if $M\preceq N$ are both $\omega$-saturated models, and $\varphi(x)$ is an $L_M$-formula, then $\text{RM}_x(M,\varphi(x)) = \text{RM}_x(N,\varphi(x))$. If you evaluate Morley rank using your definition in a non-$\omega$-saturated model, you may get a smaller value. $\endgroup$ Jun 18, 2019 at 13:33

1 Answer 1


I'm assuming your source defines Morley degree to be the $k$ from Lemma 5.10.

For (a), Lemma 5.10 ensures that every $L_A$-formula $\varphi(x)$ can be decomposed into $k$ disjoint pieces of full Morley rank, where $k = \text{dM}(\varphi(x))$. The main observation is that the formulas $\varphi_j(x)$ in this maximal decomposition all have Morley degree $1$. Indeed, if $\varphi_j(x)$ had Morley degree $d>1$, then applying the lemma to this formula, we could split $\varphi_j(x)$ into $d$ disjoint pieces of full Morley rank. Replacing $\varphi_j(x)$ with these formulas in the original decomposition of $\varphi(x)$ would decompose $\varphi(x)$ into $k-1+d > k$ disjoint pieces of full Morley rank.

Now for any complete type $p(x)$, let $\varphi(x)$ be a formula in $p(x)$ of minimal Morley rank and degree. Suppose for contradiction that $\text{dM}(\varphi(x)) = k > 1$. Then $\varphi(x) \leftrightarrow \bigvee_{j=1}^k \varphi_j(x)$, and each $\varphi_j(x)$ has Morley rank equal to $\text{RM}(\varphi(x))$ and Morley degree $1$. Since $p(x)$ is a complete type, $\varphi_j(x)\in p(x)$ for some $1\leq j\leq k$, contradicting minimality of $\varphi(x)$.

For (b), I'll give you a hint. Let $k = \text{dM}(\varphi(x))$, and let $n$ be the number of complete types of full Morley rank containing $\varphi(x)$. You can prove $k\leq n$ by decomposing $\varphi(x)$ by Lemma 5.10 and extending each formula in the decomposition to a complete type of maximum Morley rank. And you can prove $n\leq k$ by picking disjoint formulas containing each of the types of full Morley rank, and applying the bound from Lemma 5.10.

  • $\begingroup$ I understand your answer for (a), but what do you mean by "full Morley rank"? $\endgroup$
    – Ibrahim
    Jun 19, 2019 at 11:00
  • $\begingroup$ Equal to the formula in question (not less). $\endgroup$ Jun 19, 2019 at 13:16
  • $\begingroup$ I think I have (b): for $k \leq n$ you use that $\text{RM}(\varphi(x) \lor \psi(x)) = \max(\text{RM}(\varphi(x)), \text{RM}(\psi(x)))$ and $\text{RM}(x=x) = \text{RM}(\varphi(x) \lor \neg\varphi(x)) \geq \text{RM}(\chi(x))$ for any $\chi(x)$, to extend $\varphi, \varphi_i$ to a complete type $p$ such that $\text{RM}(p) = \text{RM}(\varphi)$. These types must be different because over $\text{Th}_A(A)$ the $\varphi_i$ are incompatible. $\endgroup$
    – Ibrahim
    Jun 20, 2019 at 10:40
  • $\begingroup$ For $n \leq k$: assume WLOG that $n < \infty$, there are $\psi_i$ such that $[\psi_i \land \varphi] \cap \{p_j\}_{j \leq n} = \{p_i\}$. Also $\text{RM}(\psi_i \land \varphi) = \text{RM}(\varphi)$, so the formulas $\psi_i \land \varphi$ witness that $k = dM(\varphi) \geq n$. $\endgroup$
    – Ibrahim
    Jun 20, 2019 at 10:52

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.