According to Schlesinger (1973) there are two errors in our research on language development: (a) we attribute too much underlying structure to child utterances; (b) we interpret child language in terms of adult language. In response, we restate the types of evidence we have used to derive semantic representations for child utterances, and emphasize that only such representations can make sense of the way the child uses his language. One of the reasons that child sentences are brief in comparison with the structure they communicate is that the child communicates by using the immediate situation. He constructs parts of the semantic structure not conceptually but at the sensory–motor level and communicates them deictically. Nevertheless the entire structure is under the child's control. We further clarify our notion of adverbial, which has permitted us to conclude that adverbials are absent in very early speech.