A series of five runs comprised of 20 searches each were carried out in a three-year span of the INSPEC data base. The same queries were processed manually in the corresponding volumes of Science Abstracts, and comparison of the output of each search method exposed reasons for the failures. Emphasis is laid on the fact that a query comprises other factors in addition to its subject matter and means of satisfying these criteria automatically are discussed. The more flexible nature of the manual searches is examined in detail with a view to improving the flexibility of the computer searches. (Author/DGC)