Background. Attention bias modification (ABM) is a new treatment for affective disorders. A meta-analysis of ABM for anxiety disorders showed that the effect size may be large but the number of studies is low. The working mechanism is still unclear, and little is known about the optimal treatment parameters. ABM for depression is much less studied. A few studies claimed positive effects but the sample sizes are low. Furthermore, the treatment parameters varied widely and differed from the anxiety literature. Aim. To select the most promising version of ABM for depression for further evaluation in clinical trials. Methods. Multiple case series design. We tested six versions of ABM that varied on stimulus duration and training direction. Thirty students with mild to moderate symptoms of depression underwent four sessions of ABM. Change of attentional bias was measured during each session. Generalization of treatment effects and the role of awareness of receiving training were also investigated. Results. None of the investigated versions of ABM had a consistent effect on attentional bias. Changes of attentional bias in individual participants the effects did not generalize to untrained stimuli. Conclusion. It is unlikely that any of these ABM versions will have a specific effect on symptoms in controlled studies.