“If I get angry, I don’t usually lose my control. I get very controlled…like I’m uptight, talking very slowly, very direct, little animation.”

Examples of silent anger are illustrated by these two quotes:

“I go downstairs, find a stone to throw in the field.”

“It was more non-verbal anger.”

Classifying Outcomes of Anger Expression. Since there is little theoretical synthesis in the literature on anger outcomes (see similar argument in Averill 1982, Glomb and Hulin 1997, Fitness 2000), we initially utilized a grounded theory approach (Glaser and Strauss 1967, Charmaz 2006) to create categories for perceived outcomes of anger episodes. We started with no a priori categories and used the data to generate our categories inductively. Using a subset of the 158 episodes, two of the authors developed and refined a coding guide of 68 categories using many “in vivo” codes from respondents’ own words (Strauss and Corbin 1990: 69). Rater’s reliability on coding the 554 reported outcomes into 68 outcome categories was 92% for unitized (Guetzkow 1950) and 74% for non-unitized reliability with a significant Cohen’s kappa (Tb=11.217, sig. < .001), indicating an acceptable level of agreement.

To reduce this large number of categories to theoretically meaningful groupings, we next used multidimensional scaling (MDS) to confirm whether a smaller number of categories could be used to organize our data. Multidimensional scaling is the set of mathematical techniques that enables researchers to uncover the structure of their data (Kruskal and Wish, 1978:5). Hence it is an ideal technique for creating classifications and determining how many dimensions best fit the data. Our preliminary coding of the data suggested two dimensions for classifying anger outcomes. These independent dimensions are found in the extant literature. First, anger outcomes can be divided by a valence dimension: positive and negative (Averill 1982, Glomb and Hulin 1997, Lang 1995, Watson and Clark 1984). This distinction relies on a functional approach to emotions (see Keltner and Gross 1999), suggesting that we can characterize the outcome of an anger episode by assessing whether respondents perceive that the episode served adaptive or beneficial purposes (e.g., helped them achieve goals or solve problems) versus