Appraisal theory is a cognitive approach to emotion recognition intended to be behaviourally inclined and context-aware. While popular in cognitive psychology, dimensional approaches have rarely been used in conversation emotion recognition. In text analysis, crowd-\textsc{EnVent} is the benchmark corpus and model for an appraisal-theoretic analysis of emotion \citep{troiano-etal-2023-dimensional}. In this paper, we extend the use of this corpus and extend their model to analyse three prominent emotion-annotated dialogue corpora. First, we showcase a dynamic, attention-based, in-situ method for context compression to account for dialogue context when evaluating individual utterances. Next, we use this method on dialogue corpora to extract appraisal information which we use to perform categorical emotion detection. Our analysis shows that while maintaining parity with existing baselines, an appraisal-based model provides relevant insights into conversation behaviour. Finally, we examine the change in appraisal dimensions throughout a conversation to get qualitative insights into emotion dynamics which may be pertinent for downstream tasks like empathetic dialogue generation.