The focus of the two tutorials on how to cope with varying input conditions for a speech recognition system is quite complementary between the two papers. FURUI gives examples of both supervised and unsupervised adaptation of the recognizer models by means of stochastic pattern matching to cope with signal variation due to channel or speaker variability. HERMANSKY describes a method (known as RASTA) to include auditory knowledge into the recognition algorithm rather than "blind implementation of scattered accidental knowledge".