This thesis is about the cortical mechanisms underlying low-level motion processing in the retino-geniculo-cortical pathway of the primate visual system. The findings in this thesis provide new insights into the mechanisms underlying local motion detection and its integration into a global motion percept.
Detection of local motion requires correlation in time and space. To study the spatiotemporal characteristics of normal (phi) motion, we have developed a new stimulus paradigm. The stimulus consists of a random dot pattern in which each dot was presented in two frames only, separated by a specified interval. On each frame, half of the dots were refreshed and the other half was a displaced reincarnation of the pattern generated one or more frames earlier. By flipping the contrast of the displaced dots, we could also study the spatio-temporal characteristics of reverse-phi motion . We also developed a no-phi stimulus. In this stimulus the contrast of the displaced dots was set to the background luminance.
A remarkable finding was that the spatiotemporal properties of phi and reverse phi motion are very similar. With the no-phi stimulus, we demonstrated that the lack of motion information is insufficient to trigger a reverse-phi percept and explain its spatio-temporal sensitivity. This strongly suggests that efficient detection of correlation across ON and OFF channels occurs.
This postulation naturally led to the question how these signals from the ON and OFF channels are combined. Mo and Koch (2003) suggested that correlations between opposite contrast polarities lead to activation of motion detectors sensitive to motion in the opposite direction. However, an alternative explanation is possible, namely that correlations with opposite contrast polarities lead to a reduced activity of motion detectors sensitive to movement in one direction than the physical movement. We show that our alternative hypothesis gives a better explanation. The proof is based on experiments showing that reverse-phi percepts in many respects behave like motion after-effects , and not, as suggested by Mo and Koch (2003), as normal movement. Correlations between equal contrast polarities increase the activity and are used as `positive' evidence that a movement component. Correlations between opposite polarities decrease the activity and provide 'negative' evidence that a movement component is absent. Such a mechanism is advantageous for the visual system, because the weighing of positive and negative evidence of movement at the lowest level components of the motion signal to noise ratio increases and thus leads to a more efficient system for motion detection.
The findings of this thesis are incorporated into a model and discussed in a broader context. The thesis ends with concluding remarks and suggestions for further research and some preliminary results that support our conclusions.