Individual is presented with words with some letters missing and told to complete the word.

Adapted from the "Angry Cognitions Scale" (Martin and Dahlen 2007). Volunteers read hypothetical scenarios where another person acted aggressively but with unclear intent (e.g., "You are driving through a residential area when someone backs their car out of a driveway and nearly hits you."). Volunteers respond to items (from "Very Unlikely" to "Very Likely") to indicate how they would think about the situation (e.g., "He/she did that just so I'd have to stop. He/she was trying to scare me."). Volunteers responded to one set (4 scenarios) at time point one and a second set (5 scenarios) at time point two. Reponses were summed within each scenario and averaged across scenarios to indicate level of hostile attribution bias. Individual scores at each time point could range from 0 (no hostile attribution bias) to 24 (high hostile attribution bias).

State Aggression Survey: This survey is adapted from several others in the literature to measure variance along the normal spectrum of aggressive behaviors in daily life that the average person might display (Álvarez-García, et al., 2016; Deffenbacher, et al., 2001; Deffenbacher, J. et al., 2002). The survey specifically asks about driving behaviors (e.g., yelling at other drivers). Scoring is count of aggressive behaviors during reporting period.

Seven items adapted from the Cyber-Aggression Questionnaire for Adolescents by Álvarez-García et al. (2016). Volunteers reported at time point two how often (during preceding 24 h) they engaged in various aggressive online behaviors (e.g., posted rude comments about someone on a social network). Response choices ranged from 1 = never to 4 = always. Scores were recorded dichotomously to represent whether a volunteer reported any online aggression during the reporting period (i.e. "never" was recoded as "0" to indicate no aggression, and all other responses were recoded as "1" to indicate at least some aggression. A higher percentage of volunteers in a study condition reporting usage of aggression on social media indicates a worse outcome for that study condition.

Trait Anger Scale: Brief measure of trait anger validated by Wilk et al., (2015). Scale (1-5, strongly disagree to strongly agree); Two questions regarding individual's perspective on their anger. Higher values indicate greater anger.

HBMT: Individual is presented with words with some letters missing and told to complete the word.

Other Training: Alternative to HBMT training

Only volunteers who completed all measures were analyzed.

Adapted from the "Angry Cognitions Scale" (Martin and Dahlen 2007). Volunteers read hypothetical scenarios where another person acted aggressively but with unclear intent (e.g., "You are driving through a residential area when someone backs their car out of a driveway and nearly hits you."). Volunteers respond to items (from "Very Unlikely" to "Very Likely") to indicate how they would think about the situation (e.g., "He/she did that just so I'd have to stop. He/she was trying to scare me."). Volunteers responded to one set (4 scenarios) at time point one and a second set (5 scenarios) at time point two. Reponses were summed within each scenario and averaged across scenarios to indicate level of hostile attribution bias. Individual scores at each time point could range from 0 (no hostile attribution bias) to 24 (high hostile attribution bias).

HBMT: Individual is presented with words with some letters missing and told to complete the word for form non-hostile words.