The Component Process Model posits that attention is appraisal-driven rather than stimulus-driven and that the appraisal of relevance is of critical importance in such a mechanism. This means that any stimulus can attract attention or not depending on how relevant it is appraised. This hypothesis was tested in an implicit border similarity judgement task, in which thirsty participants were presented with bottles and vases that were respectively very relevant and weakly relevant to their goal to quench their thirst. These stimuli were also presented to quenched participants for whom none of the stimuli was relevant. The findings support the idea that our attention is more likely to be appraisal-driven than stimulus-driven, since bottles produced an attentional interference in thirsty participants only. It was also observed that, even if vases were judged weakly relevant by thirsty participants, they produced an attentional interference compared to empty stimuli, which was not the case in the quenched participants group. The concept of goal relatedness was proposed as an explanation for this result, and methodological implications were also discussed.