Ethnomethodological studies of work attempt to examine ordinary activities for the ways in which they exhibit observably and accountably competent work practice as viewed by practitioners. Because it is the analyst's task to describe activities as viewed by practitioners, qua practitioners, two methodical problems must be solved. The first problem is how the analyst can know and describe the members' point of view. Because members display their point of view to each other, the problem can be formulated as the question of how the analyst can identify members' displays. This is the problem of observability. The second problem is how the analyst can make a distinction between activities done by practitioners, qua practitioners, and other activities. How can the analyst for instance, distinguish 'talk as work' from 'talk at work'? This is the problem of specificity. This paper's aim is to critically examine the ways in which ethnomethodology manages these two problems. First, the concept of the 'occasioned corpus of setting features' is discussed. Next, two examples of ethnomethodological studies of institutional work are examined. It is argued that these studies fail to provide an adequate description of the specificity of the work studied because both disregard practitioners' orientation to the requirements of institutional accounting and reporting. Finally, it is discussed how practitioners' orientations to institutional criteria can be made observable in the analysis.