Sorry for my bad English.
In Mumford's "Abelian variety" section 16, there is vanishing theorem as follow;
Let $X$ be an abelian variety of dimension $g$, and $L$ be an ample line bundle. Then there is a unique integer $i=i(L), 0\leq i\leq g$, such that $H^p(X, L)=0 $ for $p\neq i$, and $H^i(X,L)\neq 0$.
The last of the proof is as follow;
if $h^i(L):=\dim_k H^i(X, L)$, we have $\Sigma_{i=0}^q h^i(L)h^{q-i}(L^{-1})= 0, (0\le q< g),$ and $\Sigma_{i=0}^g h^i(L)h^{g-i}(L^{-1})=\deg \phi_L$. Now $h^i(L)\ge 0$ so we get this proposition.
But I can't understand why this equation deduce this proposition. In fact, for example $g=1, h^0(L)>0, h^1(L)>0, h^0(L^{-1})=0, h^1(L^{-1})>0$ is counterexample.
Please help me, thanks.