A$.$ when you have associated certain features with a concept, you are engaging in Follow$-$back reasoning but when you start with a concept but remain fiexible regarding the features, you are engaging in Followforward reasoning, which is more powerful.

B$.$ When you have a concept, such as "bicycle", and you see other objects that roughly fit it$,$ you try to assimilate these new examples into your concept but also have to stretch your concept to make it accommodate examples that do not fit.

C$.$ any process, no matter how reliable, always produces variation and with enough examples, that variation takes the shape of a so$-$called "normal distribution", with most example clustering near a mean and some outliers on either side with decreasing frequency.

D$.$ you can learn complex concepts such as "bicycle" more readily by observing examples of it than by being reinforced for getting guesses correct.