The frequency with which a child's parents use a given linguistic form has been considered influential in language development. This hypothesis has been challenged, however, notably by Ervin (1964) and Brown (1973). The frequency hypothesis makes the assumptions that: (1) children are not selective in what they attend to, (2) they listen to most of what they hear, and (3) what they do listen to is an unbiased sample of what is said. Using data from two Japanese girls recorded between the ages of 2 and 4, the present study shows that these assumptions are not wholly correct. As an index of whether the children were attending to adult speech or not, adult influences in what the children said were looked for, specifically in the form of imitations, answers to questions, replies and comments by the child, rejection of adult remarks, and miscellaneous indications, such as compliance to requests. These data were compared with the characteristics of children's speech when it is not under the influence of adult speech, including repetition, sound play, and unsystematic jumping from topic to topic. It is suggested that children do not attend to parental speech when they are engaged in some forms of internal activity, linguistic or otherwise. (Author/CLK)