Utilization of the card catalog in the main library (Sterling Memorial Library) of Yale University was studied over a period of more than a year. Traffic flow in the catalog was observed, and was used as the basis for scheduling interviews with a representative sample of catalog users at the moment of catalog use. More than 2000 interviews were completed. Data were collected on user objectives and starting clues. Follow-up studies were done on the matches among user clues, catalog card data, and information available in the front matter of cataloged documents. Reasons for search failures were determined. In terms of immediate intent, 73 percent of searches are document ("known item") searches and 16 percent are subject searches; in terms of underlying interest, 56 percent are document searches and 33 percent are subject searches. Remaining searches are 6 percent author searches (to find out what is on hand from a known author or institution) and 5 percent bibliographic searches (to complete or verify a reference on the basis of catalog-card data). The importance of secondary search clues in achieving retrieval despite incomplete or inaccurate primary clues is discussed. (Author)