Previous studies have demonstrated that context manipulations by semantic blocking and category priming can, under particular design conditions, give rise to semantic facilitation effects. The interpretation of semantic facilitation effects is controversial in the word production literature; perceptual accounts propose that contextually facilitated object recognition may underlie facilitation effects. The present study tested this notion. We investigated the difficulty of object recognition in a semantic blocking and a category priming task. We presented all pictures in gradually de-blurring image sequences and measured the de-blurring level that first allowed for correct object naming as an indicator of the perceptual demands of object recognition. Based on object recognition models assuming a temporal progression from coarse- to fine-grained visual processing, we reasoned that the lower the required level of detail, the more efficient the recognition processes. The results demonstrate that categorically related contexts reduce the level of visual detail required for object naming compared to unrelated contexts, with this effect being most pronounced for shape-distinctive objects and in contexts providing explicit category cues. We propose a top-down explanation based on target predictability of the observed effects. Implications of the recognition effects based on target predictability for the interpretation of context effects observed in latencies are discussed.