How Many $N=4$ Strings Exist ?

Possible ways of constructing extended fermionic strings with $N=4$ world-sheet supersymmetry are reviewed. String theory constraints form, in general, a non-linear quasi(super)conformal algebra, and can have conformal dimensions $\geq 1$. When $N=4$, the most general $N=4$ quasi-superconformal algebra to consider for string theory building is $\hat{D}(1,2;\a)$, whose linearisation is the so-called `large' $N=4$ superconformal algebra. The $\hat{D}(1,2;\a)$ algebra has $\Hat{su(2)}_{k^+}\oplus \Hat{su(2)}_{k^-}\oplus\Hat{u(1)}$ Ka\v{c}-Moody component, and $\a=k^-/k^+$. We check the Jacobi identities and construct a BRST charge for the $\hat{D}(1,2;\a)$ algebra. The quantum BRST operator can be made nilpotent only when $k^+=k^-=-2$. The $\hat{D}(1,2;1)$ algebra is actually isomorphic to the $SO(4)$-based Bershadsky-Knizhnik non-linear quasi-superconformal algebra. We argue about the existence of a string theory associated with the latter, and propose the (non-covariant) hamiltonian action for this new $N=4$ string theory. Our results imply the existence of two different $N=4$ fermionic string theories: the old one based on the `small' linear $N=4$ superconformal algebra and having the total ghost central charge $c_{\rm gh}=+12$, and the new one with non-linearly realised $N=4$ supersymmetry, based on the $SO(4)$ quasi-superconformal algebra and having $c_{\rm gh}=+6$. Both critical string theories have negative `critical dimensions' and do not admit unitary matter representations.