strange notation for MVN covariance matrix

[partev]

In Section 3.2.1 formula (3.16) you have the following notation:

Only $\sigma_1^2$ and $\sigma_2^2$ are always positive $\sigma_{12}^2$ can be negative.

It is strange because it implies that $\sigma_{12}^2$ is positive while in reality if the correlation $\rho$ is negative then $\sigma_{12}^2$ is also negative implying an imaginary value for $\sigma_{12}$.

So the notation seems unhelpful.