The relationship between risk-taking behavior and substance dependence has proven to be complex, particularly when examining across participants expressing a range of substance use problem severity. While main indices of risk-taking in the Balloon Analogue Risk Task (BART) positively associate with problematic alcohol use in adolescent populations (e.g., MacPherson, Magidson, Reynolds, Kahler, & Lejuez, 2010), several studies have observed a negative relationship when examining behavior within adult substance using populations (Ashenhurst, Jentsch, & Ray, 2011; Campbell, Samartgis, & Crowe, 2013). To examine potential mechanisms that underlie this negative relationship, we implemented multilevel regression models on trial-by-trial BART data gathered from 295 adult problem drinkers. These models accounted for participant behavior on trials following balloon bursts or cash outs as indices of loss and reward reactivity, respectively, and included control variables including age, IQ, and individual delay discounting rate. Results revealed that individual trial pumping was significantly predicted by trial number, and by whether or not the previous trial was a big burst or a big cash out (i.e., large magnitude of potential gains) in a manner consistent with a "near-miss" effect. Furthermore, severity of alcohol problems moderated the effect of a previous trial big burst, but not of a big cash out, on subsequent trial behavior such that those with greater severity demonstrated relative insensitivity to this "near-miss" effect. These results extend previous studies suggesting that alcohol abusers are less risky on the BART by specifying a mechanism underlying this pattern, namely, diminished reactivity to large magnitude losses.
