There is growing evidence to support attentional bias modification (ABM) techniques such as the modified dot-probe task within the pain literature. Such techniques can help to inform theoretical models of pain by identifying the causal role of attentional bias constructs. The aim of this research was to explore the effects of dot-probe ABM that trains individuals towards (+) or away from (-) sensory (S) and affective (A) pain words, on attentional biases, interpretation biases, and pain outcomes. Healthy undergraduate students (N = 106) completed questionnaires, an attentional bias dot-probe task, and an interpretation bias task before and after ABM, one of four ABM versions that differed in training direction (S+A+, S-A+, S+A-, S-A-), and pain outcomes using the cold pressor task. Those trained towards affective pain words were found to have a greater pain threshold but also greater distress at tolerance. However, mechanisms of change could not be established, as ABM did not affect attentional or interpretation bias, even though changes in attentional bias were associated with pain outcomes. These findings provide partial support for the threat interpretation model and highlight the utility of affective pain ABM, although further investigation of causal mechanisms is warranted.