LISTS OF HIGH ASSOCIATION AND LOW ASSOCIATION SENTENCES WERE PREPARED FROM CONTROLLED ASSOCIATION NORMS. AN INTRA-LIST INTERFERENCE CONDITION WAS CREATED FOR HIGH ASSOCIATION SENTENCES BY PAIRING THE NOUN PHRASE OF ONE SENTENCE WITH THE VERB PHRASE OF ANOTHER SENTENCE IN THE SAME LIST. A SIMILAR PROCEDURE WAS FOLLOWED WITH THE LOW ASSOCIATION SENTENCES. EACH OF THE FOUR EXPERIMENTAL LISTS WAS PRESENTED TO A DIFFERENT GROUP OF 24 UNDERGRADUATE STUDENTS AS A SENTENCE RECALL TASK. THE PERFORMANCE OF THE HIGH ASSOCIATION GROUPS WAS SUPERIOR TO THE PERFORMANCE OF THE LOW ASSOCIATION GROUPS, AND THE PERFORMANCE OF THE NONINTERFERENCE GROUPS WAS SUPERIOR TO THE PERFORMANCE OF THE INTERFERENCE GROUPS ON SEVERAL MEASURES OF RECALL. TRANSITIONAL ERROR PROBABILITIES WERE COMPUTED FOR THE BETWEEN-PHRASE AND THE WITHIN-PHRASE TRANSITIONS OF THE SENTENCES IN EACH OF THE LISTS, AND THE RESULTS OF THE ANALYSIS OF THESE DATA WERE CONSISTENT WITH THE NOTION THAT ASSOCIATIVELY-INTEGRATED SENTENCES ARE PROCESSED IN UNITS LARGER THAN THE PHRASE, WHILE LOW ASSOCIATION SENTENCES ARE PROCESSED IN PHRASE UNITS. THIS REPORT APPEARED IN "STUDIES IN LANGUAGE AND LANGUAGE BEHAVIOR, PROGRESS REPORT IV," 1967, PUBLISHED BY THE CENTER ON LANGUAGE AND LANGUAGE BEHAVIOR, UNIVERSITY OF MICHIGAN, 220 EAST HURON STREET, ANN ARBOR, MICHIGAN 48108. (AUTHOR/JD)