Exercise 2: The HAT unit is slightly slower to activate and asymptotes at a lower level (.82) because there are two other words (CAT and BAT) that become quickly active and inhibit the HAT unit.
Exercise 3: The netwrok takes about 30 cycles to settle on a value of 0.53 for the N unit. As the SUN unit becomes active it feeds activation down to the N unit (in the third position) but not the T unit. That is, the word context is selecting the N letter unit.
Exercise 4: The netowrk does demonstrate the word superiority effect because the activation of the middle O unit rises more quickly when ZOO is presented than when AOE is presented. This occurs because there is feedback from the ZOO unit that supports the activation of the O unit, whereas the AOE pattern only partially activates a number of word units one of which is YUM which is inconsistent with a middle O.
Exercise 5: The effect for the psuedoword ZOG is actually somewhat stronger than for ZOO (consistent with observations of some pseudowords in the human data). The reason is that ZOG strongly activates both FOG and DOG both of which support the O in the middle position.
Exercise 6: The network does misperceive the O as an A in order to complete the word BAT.