Let $S$ be a K3 surface over the complex numbers $\mathbb{C}$. If $C\subset S$ is a smooth rational curve, the normal bundle $N_{C/S}$ is isomorphic to $\mathbb{O}(2)$ and thus $C$ is rigid. What about a curve $C'\subset S$ whose normalization is a rational curve? I think it may admit deformation in higher genus family, but can it admit deformation in a family of (possibly singular) rational curves?

4$\begingroup$ Not over $\mathbb{C}$, but this can occur in positive characteristic. $\endgroup$ – Jason Starr Dec 18 '13 at 15:02
To develop what Jason says: if your curve deforms in a family of rational curves, it means that you can find a dominant rational map from a ruled surface onto your K3. This is forbidden (over $\mathbb{C}$): e.g. because the nonzero 2form of the K3 would lift to a nonzero 2form on the ruled surface.

$\begingroup$ Thank you for the answer. To clarify your answer, the map is dominant because the curve has a singularity and the pullback 2form extends because the singularity of the ruled surface is of codimension 2. $\endgroup$ – Hiro Dec 18 '13 at 22:30
Here is another proof using deformation theory:
Let $f:\mathbb P^1\to S$ be the composite of inclusion $C'\subseteq S$ and the normalization of $C'$. Then there is an exact sequence
$$0\to T_{\mathbb P^1}\to f^*T_{S}\to N_f\to 0,$$
where $N_f$ is the normal sheaf of $f$ and $H^0(\mathbb P^1,N_f)$ is the first order deformation space of the morphism $f:\mathbb P^1\to S$ with the target fixed.
Since $S$ is a K3 surface, $c_1(T_S)=0$, on the other hand $c_1(T_{\mathbb P^1})=2$, so by exact sequence we have $c_1(N_f)=2$. In particular $H^0(\mathbb P^1,N_f)=0$. So $C'$ does not deform in a family of singular rational curves.