It is well known how to check stability of a linear system $\dot x=Ax$ when the components $a_{ij}$ of a matrix $A=|a_{ij}|$ are known precisely (and do not depend on time $t$). In real life, we usually do not know the exact values of the coefficients $a_{ij}$; instead, we know intervals ${\bf a}_{ij}$ of possible values of $a_{ij}$.

In some real-life
situations, we know that the system is * static*, i.e., that the
value $a_{ij}(t)\in {\bf a}_{ij}$ does not depend on time $t$. In some
other cases, the values $a_{ij}(t)\in {\bf a}_{ij}$ may actually
depend on time $t$. In the paper under review, a new necessary
condition is described that guarantees stability of such * time-varying
interval systems*. It is shown that this new condition is applicable
to some systems for which previously known stability conditions do
not work.