Say there is a superkey \(K\) that identifies an attribute each tuple in a relation \(R\) distinctly. For any item \(i\) with attributes \(\{A_j\}\) in the table, we can say that \(\forall A_j, K(i) \to A_j\)
A key is a superkey such that there \(\nexists k \subset K\) such that \(k\) is a valid superkey of \(R\) as well
A key may either be a primary key or a candidate key