Two decades of research conducted to date has examined selective visual attention to threat and reward stimuli as a function of individual differences in anxiety using the dot-probe task. The present study tests a connectionist neural-network model of meta-analytic and key individual-study results derived from this literature. Attentional bias for threatening and reward-related stimuli is accounted for by connectionist model implementation of the following clinical psychology and affective neuroscience principles: 1) affective learning and temperament, 2) state and trait anxiety, 3) intensity appraisal, 4) affective chronometry, 5) attentional control, and 6) selective attention training. Theoretical implications for the study of mood and anxiety disorders are discussed.