Language acquisition has long been thought of as a process of imitation and reinforcement. Children learn to speak, in the popular view, by copying the utterances heard around them, and by having their response strengthened by the repetitions, corrections, and other reactions that adults provide. In recent years, it has become clear that this principle will not explain all the facts of language development. Children do imitate a great deal, especially in learning sounds and vocabulary; but little of their grammatical ability can be explained in this way. Two kinds of evidence are commonly used in support of this – one based on the kind of language children produce, the other on what they do not produce. The first piece of evidence derives from the way children handle irregular grammatical patterns. When they encounter such irregular past-tense forms as went and took or such plural forms as mice and sheep, there is a stage when they replace these by forms based on the regular patterns of the language. They say such things as wented, taked, mices, mouses, and sheeps. Evidently, children assume that grammatical usage is regular, and try to work out for themselves what the forms 'ought' to be – a reasoning process known as analogy. They could not have learned these forms by a process of imitation. The other kind of evidence is based on the way children seem unable to imitate adult grammatical constructions exactly, even when invited to do so.