A long standing conjecture of Bagemihl (1956) states that there can be at most eight tetrahedra in 3-space, such that every two of them meet in a two-dimensional set. We settle this conjecture affirmatively. We get some information on families of similar nature, consisting of eight tetrahedra. We present a joint result, showing that there can be at most fourteen tetrahedra in 3-space, such that for every two of them there is a plane which separates them and contains a facet of each one of them.