The survey research center at the University of Michigan¿s institute for social research has begun the dissemination of OSIRIS 4, the latest version in Michigan¿s series of statistical systems. The system consists of 16 procedures for data analysis, 10 for general data management, and 6 for the management of structured files. The data analysis procedures include the usual routines for crosstabulation, correlation, regression, and analysis of variance. The system also contains search, a new version of aid and related programs. A sampling error analyzer and a repeated samples program are also included in the system. Several nonmetric scaling routines will shortly be added. In the file-building procedure, the user constructs a dictionary describing the raw data. The user can then treat the raw data file as if it were an OSIRIS internal-format file described by the dictionary. The user has the option of processing from the raw data, as well as the ability to generate a clean internal-format life. The most important advance in the system is its ability to generate and handle structured files. Osiris 4 is a tape-oriented system which handles its files sequentially. It is able to process hierarchical and penal study files directly, without the need for extract or work files. Structure definitions are discussed. The structured file facility is advised to be more difficult to use than any statistical system that does not handle such files. An example is provided to show what is required to produce an entry.