Abstract: PURPOSE: Previous research demonstrated that implicit or explicit contingency information associated with facial cues can be acquired, via attention training, and used to predict the appearance of targets. However, little is known about the effect of attention training without awareness. The current research therefore examined whether an unconscious attention training using positive and neutral faces as cues may result in contingency acquisition and corresponding attention modification.
METHODS: A variant of dot-probe task was employed in combination with continuous flash suppression (CFS). On each trial, a positive-neutral face pair and a colored CFS mask were presented to the non-dominant and the dominance eyes respectively, thereby suppressing the awareness of the faces. Subsequently, a gabor patch with an orientation of one degree appeared in the location previously occupied by one of the faces. Participants discriminated the gabor orientation as accurately as possible. There were three contingency conditions: 1) gabor patches always replaced positive faces; 2) gabor patches always replaced neutral faces; 3) gabor patches replaced either the positive or the neutral faces with equal probability. This task contained 200 critical trials. Awareness check was performed after attention training.
RESULTS: Accuracy rates of the first and second blocks of the attention training were calculated using the critical trials for each participant. Repeated measures ANOVA was performed. The 2 block (first block vs. second block) and 3 contingency condition interaction effect reached significance when excluding participants failed to pass the awareness check. Separate paired-sample t-tests indicated significant increase in accuracy under the two high contingency conditions rather than the random contingency condition.
CONCLUSIONS: Unconscious attention training could facilitate contingency learning. The mechanisms underlying the current training effect await further elucidation.

Key words: unconscious attention training, contingency learning, continues flash suppression