The theory of constructive operators was used as a framework to design two versions of a paradigm (color matching task, CMT) in which items are parametrically ordered in difficulty, and differ only contextually. Items in CMT-Balloon are facilitating, whereas items in CMT-Clown contain misleading cues. Participants of ages 7-14 years and adults (N = 149) were studied. We found significant model-predicted graded differences in performance between the facilitating and misleading tasks, across and within age groups, expressing age versus items' demand interactions. Younger children were differentially affected by contextual cues. Even though both task versions were highly correlated with a well-established developmental measure of attentional capacity, CMT-Clown, which contained misleading cues, was a better measure of working memory capacity. These results show a need to estimate degree of misleadingness whenever performance levels in working memory or mental attention tasks are compared and interpreted. Developmental profiles of both tasks are discussed in terms of contextual differences and neoPiagetian stages of development. (Contains 2 figures and 3 tables.)