How the elements of a visual scene are grouped into objects is one of the most fundamental but still poorly understood questions in visual neuroscience. Most investigations of perceptual grouping focus on static stimuli, neglecting temporal aspects. Using a masking paradigm, we show that the neural mechanisms underlying grouping seem to be both fast and complex. For example, a vernier target was followed by, first, a briefly presented grating and, then, a long-lasting, extended grating. Under these conditions, the briefly presented grating is hardly visible. Still, vernier discrimination strongly changed with the number of elements of the briefly displayed grating being worst for small gratings. In accordance with a neural network model of masking, we propose that the edges of the briefly presented grating and the vernier interfere in spite of the short presentation time. We suggest that this fast edge processing is a first step for unconscious grouping processes.