This program is intended as a flexible, user-oriented data screening aid prior to utilizing statistical software which lack provisions for exclusion of missing values. Individual data records are checked for the extent of omitted multivariate observations (i.e., questionnaire items, predictor variables, rating scale item components, etc.). Individual and/or group information is printed. The user can bypass this overview and directly refine the data set by creating another data set which excludes all data records having more than a user-supplied number of missing responses. (Where no information is available about the fullness of the data, the user may desire to first obtain the overview, then refine the data accordingly.) Multiple data sets can be handled by the program in one job submittal. Tape, disk, or card variable-format input is feasible on whatever units are available at the user installation. Input. A title card, parameter card, A-type variable format cards (for input, and output if options for generating a refined data set are utilized). Output. Regardless of the option selected, (I) number of cases, (2) number of items, (3) input format used. For group statistics only, (I) number of cases without any blanks, (2) average number of blanks, (3) average percentage of blanks. For group statistics and individual statistics, those listed directly above and, for each case, (I) number of blanks, (2) percentage of blanks, (3) item numbers of blank items. For creation of a refined data set, (I) output format selected, (2) case acceptance criterion-snurnber of blank items permitted by user in new data set, (3) item numbers of blank items for those cases entering refined data set and for those cases screened out, (4) number of blank items for those cases screened out (i.e., exceeding case acceptance criterion), (5) number of cases entering refined data set. Computer and Language. This program was written in FORTRAN IV for the IBM 360/Model 91 computer and requires 98K. Program BLANKS is not installation-specific since it contains its own FORTRAN subroutine to check for blanks. Restrictions. Up to 500 observations per data record, 10,000 records, and 10 input variable-format cards.