Arguably, situational, social, and cultural context play an important role in structuring our emotional responses and states. Psychological research has focused on stimulus and response, largely devoid of contextual factors where the linguistics tradition has focused on how lexical categories structure the conceptualization of an emotion. Using novel methods in corpus linguistics, this study seeks to integrate controlled contextual dimensions into the description of emotions. The case study examines ANGER in three typologically distinct languages: Czech; English; Portuguese. The primary hypothesis is that we will observe significant and systematic contextual effects on the stimulus (cause) and response (state) of ANGER. The aim is to model quantitatively these contextually sensitive patterns. The study involves three stages. The first is the collection of data. Great care is taken to ensure homogeneity in style and register across the three languages. Personal diaries are used not only because does these data control for stylistic variation but because they are narrational and rich in natural context, with descriptions of emotional causes and responses. Keyword searches are performed on the dataset to extract 600 occurrences of ANGER events. In order to control for part of speech effects, the most frequent lexemes in each langauge are used a keywords. The second stage is the sentiment analysis. This method, developed in Cognitive Linguistics (Geeraerts et al. 1994, Gries 2003), has recently be adopted in computational linguistics (Wiebe et al. 2005, Pang & Lee 2008). Based on the linguistic (Lakoff 1987) and psychological (Fontaine et al. 2014) literature, an annotation schema is devised focusing on characteristics of each of the possible participants in the ANGER event. This annotation is systematically applied to the full set of occurrences. Subjective categories are annotated by a secondary rater and a Kappa score of inter-rater reliability determined. This produces a large set of meta-data on the contextualised use of the lexemes designating ANGER. The third stage is the quantitative investigation of the meta-data. Due to their high dimensionality, exploratory multivariate analysis is used to identify usage patterns. A combination of multiple correspondence analysis and k-medoid cluster analysis are used to identify correlations and underlying structures. Preliminary results indicate that type of cause, norm violation, and social engagement in response are amongst the crucial factors structuring the emotion. Confirmatory modelling, in the form of mixed-effects multinomial logistic regression, confirm statistically significant contextual effects and permit some degree of predictive modelling with regard to the three languages concerned. Geeraerts, D. et al. 1994. The Structure of Lexical Variation. Berlin: Mouton. Gries, St. 2003. Multifactorial Analysis in Corpus Linguistics. Continuum: London. Weibe, J. et al. 2005. Annotating Expressions of Opinions and Emotions in Language. Language Resources and Evaluation 39: 165–210 Pang, B. & Lee, L. 2008. Opinion mining and sentiment analysis. Foundations and Trends in Information Retrieval 2: 1–135 Fontaine, J. et al. 2014. Components of Emotional Meaning. OUP: Oxford. Lakoff 1987. Women, Fire, and Dangerous Things. UCP: Chicago.