A slightly modified version of data used in French (1992). These patterns are used to
show that even if a system learns new data very slowly, in the absence of interleaving of
previously learned patterns (or approximations of these patterns) serious interference can
still occur;