A set $A$ is $X$-definable in an $L$-structure $\mathcal{M}$ with a domain $M$ iff there are an $L$ formula $\phi(a,\bar x)$ and $\bar x \in X^n$ such that $A=\{a \in M^m | \mathcal{M} \models \phi(a,\bar x)\}$.
Often people say that a set $A$ is just definable. Do they mean that $A$ is $\emptyset$-definable or that $A$ is $M$-definable?
A set $D$ is $A$-definable in $M$ if there is parameter-free formula $\varphi(z,x)$ and a tuple $a\in A^{|z|}$ such that $D=\{b\in M^{|x|}\ :\ M\models\varphi(a,b)\}$. (Were $|\cdot|$ denotes the length of the tuple.)
The following is useful to know (it justifies the notation used above):
[i] My writing $a\in A^{|x|}$ is overscrupulous. Usually it is agreed that $a\in A$ stands for $a\in A^{|a|}$ (it has become very common to use the same notations for tuple and for elements).
[ii] Definable stands for $M$-definable unless the author explicitly says the contrary.
[iii] Sometimes $0$-definable is used for $\varnothing$-definable.
[iv] In 90% of the scientific articles, a structure is denoted with the same symbol used for its domain.