IN ORDER TO TEST THE EFFICACY OF METHODS USED BY VARIOUS RESEARCHERS FOR GENERATING SPONTANEOUS SPEECH FROM UNDERPRIVILEGED CHILDREN, 20 THREE- TO FOUR-YEAR-OLD GIRLS WERE SELECTED AT RANDOM FROM A HEADSTART-TYPE PROGRAM AT HOWARD UNIVERSITY FOR TESTING. ALL WERE NEGROES AND FROM A TARGET AREA FOR POVERTY PROGRAMS. THE CHILDREN WERE TESTED INDIVIDUALLY OR TWO AT A TIME BY TWO FEMALE EXAMINERS, ONE WHITE AND ONE NEGRO. TEST SESSIONS WERE SHORT AND CHILDREN UNWILLING TO PARTICIPATE OR SUSPICIOUS OF THE TEST WERE EXCLUDED. TECHNIQUES FOR GENERATING SPONTANEOUS SPEECH INCLUDED USE OF A DOLL AND CRIB, TOYS, COLORING BOOKS, A MALE PUPPET, AND A TELEPHONE. PLATES FROM THE AMMONS FULL RANGE PICTURE VOCABULARY TEST WERE ALSO USED. IT WAS FOUND THAT MOST OF THE CHILDREN GAVE LITTLE MORE THAN ONE WORD RESPONSES. LONGER RESPONSES WERE USUALLY ELICITED RATHER THAN SPONTANEOUS. THEREFORE, MANY OF THE METHODS USED TRADITIONALLY TO GENERATE SPONTANEOUS SPEECH IN THE UNDERPRIVILEGED CHILD WERE JUDGED UNSUCCESSFUL OR OF QUESTIONABLE VALUE. THIS ARTICLE APPEARED IN THE FALL 1967 ISSUE OF "THE JOURNAL OF NEGRO EDUCATION." SINGLE COPIES ARE AVAILABLE UPON REQUEST FROM THE OFFICE OF THE DEAN, THE COLLEGE OF LIBERAL ARTS, HOWARD UNIVERSITY, WASHINGTON, D.C. 20001. (JD)