Embedding a target within contextual elements can influence performance in visual tasks. For example, when a vernier is flanked by two lines, discrimination performance deteriorates strongly compared to unflanked presentation. This contextual modulation is usually explained by local spatial interactions. In a number of experiments, we have shown that this explanation is inadequate. We propose instead that flanks interfere most strongly with the vernier when they are grouped with the vernier i.e., when they share attributes such as size or color. This interference is lessened when the flanks form their own group and thus allow the vernier target to stand out. Here, we extend these results to flank grouping by geometrical properties. For example, vernier discrimination thresholds are significantly lower when the vernier is flanked by two cuboids compared to being flanked by two lines even though these lines are contained in the cuboids. Our results show that global figural aspects are crucial in contextual modulation and may even point to an anticipatory re-wiring of early visual cortex