Because the main difference between combinatory weak equality and $\lambda\beta$-equality is that the rule \begin{equation*}\tag{\xi} X = Y \vdash \lambda x.X = \lambda x.Y\end{equation*} is valid for the latter but not the former, it is easy to assume that another way of defining combinatory $\beta$-equality is to add rule $(\xi)$ to the postulates for weak equality. However, to make this true, one must choose the definition of combinatory abstraction in $(\xi)$ very carefully. If one tries to use one of the more common abstraction algorithms, the result will be an equality, $=_\xi$, that is either equivalent to $\beta\eta$-equality (and so strictly stronger than $\beta$-equality) or else strictly weaker than $\beta$-equality. This paper will study the relations $=_\xi$ for several commonly used abstraction algorithms, distinguish between them, and axiomatize them.