These rules tell us that a sentence for the child can be: (a) a single word from the open class (e.g., "milk"); (b) a pivot class word followed by an open class word (e.g., "see boy," "bye-bye mommy"); or (c) two open class words (e.g., "mommy sleep"). Rules 4 and 5 list some of the members of the pivot and open classes; these serve as the final replacement elements to be substituted for the P and 0 in Rules 1 to 3. There also turned out to be a very small class of pivot-like words that occurred in the final position of the children's two-word utterances. Therefore the pivot class was subdivided into PI and P2, and Rule 2 was replaced by two rules.
Recent work has uncovered some significant weaknesses in the pivot and open classes description of two-word speech (this evidence against pivot-open analysis is reviewed by Brown, 1973, pp. 97-111). Bloom (1970) has shown that pivot and open classes are not found in the speech of all children learning English. Now that more data have been collected against which to test them, the rules simply do not fit the data; pivots occur alone and in combination with other pivots, sentences longer than two words occur often and cannot be accounted for, and there is distributional evidence that suggests that children at this stage are using more than two word classes. Also, much data from children learning languages other than English cannot be characterized in terms of two classes of words (Brown, 1973). The most important criticism is that the pivot and open characterization of child speech is a superficial one which underestimates the child's knowledge (Bloom, 1970). That is, it does not consider what the child is attempting to communicate, but merely deals with the order of the words he produces. We will now turn to some recent work which does consider what the child is attempting to communicate.
The Child's Representation of Underlying Conceptualizations. The more recent work which includes studying the conceptualizations underlying early speech is discussed in detail by Brown (1973). These studies apparently indicate the direction of inquiry in the field for at least the next few years; so far, however, they have dealt only with very early syntactic forms, primarily those of children in what Brown has designated as Stage I. These stages are defined quantitatively by the mean-length-of-utterance (MLU), measured in morphemes. The maximum MLU of Stage I is 2.00. The "ideal value" for the MLU at Stage I is 1.75, and for Stage II, 2.25. It should be noted that although Brown's stages are not "true" stages in the sense of being necessarily qualitatively different from each other, they are useful for avoiding the confusions resulting from grouping children by age, since children of the same age may be in very different stages of language acquisition.