Under normal viewing conditions, due to e.g. the motion of objects, the retinotopic representation of the environment constantly changes. Yet we perceive the world as stable and we easily keep track of moving objects, indicating the presence of non-retinotopic representations in the brain. We investigated the neural correlates of such non-retinotopic processing in the human visual cortex by means of high-resolution fMRI. To this end, we used a Ternus-Pikler display: Three horizontally aligned dark/light gray checker-boards moved back and fourth in a left-right apparent motion while participants kept their gaze fixated above the stimulus, resulting in a non-retinotopic integration of the elements across frames. In a second condition, only two checker-boards were presented at the same location in each frame, thus causing the elements to be integrated retinotopically. Crucially, the checker-boards could be either flickering, i.e. inverting polarity with each frame, or non-flickering. This retinotopic flickering or non-flickering is perceived in the 2-element conditions. In the 3-element conditions, however, because of the non-retinotopic correspondence, the percept is inverted such that the non-flickering checker-boards are perceived to flicker and vice versa. In V1, we found a higher BOLD response for retinotopically flickering stimuli, as expected, but the activity did not depend on the percept of flickering vs. non-flickering. In hMT+ we found a stronger response for the 3-element conditions, which could be due to the apparent motion as well as the larger size of the stimuli. More interestingly, there was also an interaction effect, namely higher activity for perceived flickering than perceived non-flickering. Thus, apart from the retinotopic properties of the stimuli, also the endogenous percept of non-retinotopic origin is reflected in the hMT+ activity.