The development of a new scheme for classifying the more than 7,000 documents on reading in the ERIC/CRIER data collection is described. An original scheme, developed in the early days of the clearinghouse, provided a means for listing documents in any of 45 broad subject categories in the field of reading and for identification by grade level. The system was found to be inadequate in that it did not permit sufficient precision for effective retrieval. The newly developed classification scheme proposes 27 classes into which documents can be divided. Each of these uses a numerical schedule which includes subclasses and provides means of adding more as needed. Further classification by grade level, source of material, type of material, and date of publication is also provided through a faceting technique. To facilitate use of the classification scheme, an index was developed which alphabetically lists terms included in the scheme. Testing of the scheme by ERIC/CRIER personnel was carried out on a 200-document sample, and comparisons were made of their notations. Overall agreement of better than a designated standard of 60 percent was determined for 77 percent of the documents, with higher agreement noted in individual parts of the notation. Illustrations of aspects of the scheme are included. (MS)