The estimation processes used by fifth through eighth grade students as they responded to computational estimation test items were examined. Interview-based process descriptions were cross-validated using large group test data from an open-ended test and a multiple choice test. Five question formats were used to test different estimation processes: standard multiple choice; operation in foils; benchmark; and order of magnitude or operation in stems (for fractions). The mental processes tested were: (1) rounding by the usual rules to the closest power of ten or to the closest whole number; (2) front-ending, or rounding down to the power of ten of the leading digit or to the whole number of a mixed numeral; (3) other rounding, including all numbers up or some numbers up and others down; (4) using compatible numbers or numbers relatively close to the given numbers; and (5) compensating, or adjusting an estimate to reflect variations that might result from rounding or the use of some other adjustment process. Students had a strong mental set to round numbers to the nearest leading power of ten even when the items required other estimation processes. Performance differed by item format, types of numbers and operations in the items, and grade level of students. (The appendices include some test items). (Author/JGL)