The present study investigated the temporal dynamics of the object-scene congruity during a categorization task of objects embedded in a scene. Participants (n = 28) categorized objects in scenes as natural or man-made while event-related brain potentials (ERPs) were recorded. The object-scene associations were either congruous (e.g., a tent in a field) or incongruous (e.g., a fridge in a desert). The results confirmed that contextual congruity affects item processing in the 300-500 ms time window with larger N300/N400 complex in the incongruous than in the congruous condition. However, unlike previous work which found an effect of congruity starting at similar to 250 ms poststimulus on fronto-central regions, the earliest sign of a reliable context congruity effect arose at similar to 170 ms at left centro-parietal regions in the present study. The present results are in line with those of previous studies showing that object and context are processed in parallel with continuous interactions from 150 to 500 ms, possibly through feed-forward co-activation of populations of neurons selective to the processing of the object and its context. The present finding provides novel evidence suggesting that online context violations might affect earlier visual processes and routines of matching between possible scene-congruent activated schemas and the upcoming information about the item to process.