# Why do $\omega$-models of subsystems of $\mathsf{Z}_2$ satisfy full induction?

Richard Shore, in his 2010 paper in the Bulletin of Symbolic Logic, 'Reverse Mathematics: The Playground of Logic', writes that

Obviously, if an $\omega$-model $\mathcal{M}$ (those with $M = \mathbb{N}$) is a model of one of the systems above, such as $\Pi^1_1\text{-}\mathsf{CA}_0$, then it is also a model of the analogous system, such as $\Pi^1_1\text{-}\mathsf{CA}$.

Unfortunately, it is not obvious to me, although I'm sure I'll kick myself once someone explains.

-

An $\omega$-model $M$ has for its first-order part the standard natural numbers from the metatheory (think: some model of ZFC containing $M$ as a set). Since the standard numbers satisfy induction for all properties definable in the metatheory, in particular they satisfy induction for all properties definable in the model $M$, because any property definable in $M$ by an $L_2(M)$ formula is definable in the metatheory by a formula of set theory with parameters.
A similar thing happens even if we start with a model $N$ of $\mathsf{ACA}_0$ which may not have satisfy the full induction scheme. If $M$ is a countable coded $\omega$-submodel of $N$, meaning it has the same numbers as $N$, then $N$ will think that $M$ satisfies full induction. This is because $M$ is coded in $N$ using a single sequence of sets, so set quantification over $M$ can be simulated by arithmetical quantification in $N$ using $M$ as a set parameter. Because $\mathsf{ACA}_0$ includes the induction scheme for arithmetical formulas with parameters, $N$ believes that $M$ satisfies full induction.
One thing to watch out for is that some people, including Shore, use $\mathbb{N}$ to refer to the set $\omega = \{0, 1, 2, \ldots\}$ from the metatheory, while others, including Simpson, use $\mathbb{N}$ to refer to the first order part of any given model of arithmetic, which may not be $\omega$.
Thanks Carl, that clarifies things substantially. The point about countable coded $\omega$-submodels is also really interesting. – Benedict Eastaugh Feb 24 '12 at 13:52
Thanks. The two situations are completely parallel: even if $V$ is a nonstandard, non-well-founded model of set theory, if it satisfies ZFC then it will think of itself as well founded, and so it will think that any $\omega$-model it contains must satisfy full induction. – Carl Mummert Feb 24 '12 at 14:23