In [GH], Griffiths and Harris asked whether a projective complex submanifold of codimension two is determined by the moduli of its second fundamental forms. More precisely, given a nonsingular subvariety $X^n subset {mathbb P}^{n+2}$, the second fundamental form $II_{X,x}$ at a point $x in X$ is a pencil of quadrics on $T_x(X)$, defining a rational map $mu^X$ from $X$ to a suitable moduli space of pencils of quadrics on a complex vector space of dimension $n$. The question raised by Griffiths and Harris was whether the image of $mu^X$ determines $X$. We study this question when $X^n subset {mathbb P}^{n+2}$ is a nonsingular intersection of two quadric hypersurfaces of dimension $n >4$. In this case, the second fundamental form $II_{X,x}$ at a general point $x in X$ is a nonsingular pencil of quadrics. Firstly, we prove that the moduli map $mu^X$ is dominant over the moduli of nonsingular pencils of quadrics. This gives a negative answer to Griffiths-Harriss question. To remedy the situation, we consider a refined version $widetildemu^X$ of the moduli map $mu^X$, which takes into account the infinitesimal information of $mu^X$. Our main result is an affirmative answer in terms of the refined moduli map: we prove that the image of $widetildemu^X$ determines $X$, among nonsingular intersections of two quadrics.