Evaluation is a pervasive element in spoken and written language but its identification poses serious problems to linguistic researchers, especially when they are dealing with larger amounts of text which require the application of computer-assisted analytic techniques. This article explores ways of identifying items of evaluative meaning in a three million word corpus of linguistic book reviews, a text type that is particularly rich in expressions of positive and negative evaluation. It discusses whether it is at all possible to capture evaluation in a corpus in a systematic way and which analytic strategies may be most promising in the search for a larger set of meaningful patterns. The paper ends with a discussion of some unresolved issues in the area of evaluation research and sketches tasks for future activities in the field.