Abstract
The development of APPRAISAL theory has opened up areas of interpersonal meanings that had been relatively neglected within Systemic Functional Linguistics and other approaches to the analysis of discourse. The model is comprehensive and discourse-based, and in many respects it works well in practice. However, text analysis using the model has thrown up a number of problematic aspects; and in this paper I explore some of the problems in the attitudinal systems. In particular, I question whether all representations of feeling (through mental processes or agnate attributes) should be categorized as AFFECT, and I propose a more restricted definition of AFFECT. I also examine critically the criteria for distinguishing between JUDGEMENT and APPRECIATION, particularly in the area of behavior. Finally, I discuss what I call the ‘Chinese box’ issue: the possibility of seeing APPRAISAL choices as layered, with a choice in one system functioning as a token of a choice in a different system. This appears to be necessary in order to capture how APPRAISAL works in discourse, but it can become dangerously impressionistic. I argue for an approach to analysis which, while taking such layering into account, consistently stays as close to the wording of the text as possible.