Marginally perceptible prototypes as primes lead to slowed reactions to related category exemplars as compared to unrelated ones. This at first glance counterintuitive finding has been interpreted as evidence for a particular mechanism of lateral inhibition, namely the centre surround inhibition mechanism. We investigated the semantic surround of category labels by experimentally manipulating the prototypicality of stimuli. Participants first learned two new categories of fantasy creatures in a 5-day-long learning phase before they worked through a semantic priming task with the category prototypes as primes and category exemplars as targets. For high-prototypical targets we observed benefit effects from related primes, whereas for low-prototypical targets we observed cost effects. The results define when the centre surround inhibition mechanism is applied, and furthermore might explain why previous studies with word stimuli (i.e., material that prevents experimental manipulation of prototypicality) observed mixed results concerning the prototypicality of targets.