To decide whether a $sink$ exists from an adjacency matrix, we need to traverse the whole matrix in the worst case. So the best algorithm requires $O(V^2)$ comparisons.
On the other hand, to find if a $universal\space sink$ exists from an adjacency matrix, the best algorithm requires $O(2V) = O(V)$ operations only, as we can skip each row and column on one comparison.
So, the answer is, $a$) $O(V^2).$