This study evaluated Johnson’s (1965) measure of phrase independence in the recall of sentences. Twenty-five Ss were presented with groups of three sentences that differed in structure and were immediately asked to recall the sentences. Fewer errors were made on subject, verb, and object words for all sentence types, but the probability of a transitional error (TEP) at the phrase boundary was not always greater than at other transitions in the sentence. The results indicate that the TEP is a misleading measure and that sentence meaning is more appropriate as a unit of recall than is the sentence phrase. The number of underlying clauses in the sentence proved to be a valuable metric in predicting errors in recall.