By combining the paradigms of motion induction (presentation of an inducing stimulus, followed after a short delay by the presentation of an elongated bar next to it) and visual search (many-item displays with or without a pop-out target), it was possible to demonstrate the existence of two separate contributions to the motion induction effect. Illusory motion in the test bar could be produced either preattentively or by facilitation due to attentional capture. The former effect is fast, independent of the delay between the inducers and the test bar and operating simultaneously at all locations across the visual display, the latter is slower (full strength in 200-300 msec) and confined to the vicinity of the pop-out inducer. The two possibly also differ in their spatial extent, the attentional capture effect extending over a larger area around the inducer. We conclude that the motion induction effect can be used to show the existence of several effects due to the sudden presentation of a visual stimulus.

Motion induction is the illusory motion within an elongated stimulus, such as a bar or a line, when it is preceded by a priming stimulus next to one of its ends. Motion is away from this primer. The presentation of two priming spots at both ends of a stimulus bar results in motion away from both spots with a collision in the center of the bar. With a sufficiently long delay between the spots, motion will be seen only as away from the second spot. Similarly, in a bar with a luminance gradient an illusory motion is perceived as away from the high-luminance end, presumably due to the known dependence of neural processing speed on luminance. In the present study, these two illusory motions were made to oppose each other. The particular luminance gradient which would just cancel the motion induction effect when motion is seen optimally as away from the second spot (cancellation gradient) was determined, resulting again in a collision near the center of the bar. Furthermore, the luminance dependence of the reaction time to stimulus detection was measured in a separate experiment. Thus for each observer, the processing time difference associated with the cancellation gradient was established. This delta t then gives the amount of time by which processing is speeded up in motion induction due to the priming spot. In a simple model of motion processing it can also be identified as the built-in delay delta t of a typical Reichardt-type motion detector. With the present conditions, it varied between 14 and 19 msec for different observers for a bar length of 5.3 deg. In this way, we show not only that the priming effect in motion induction can be understood as a speed-up of neural processing, but also provide a way of measuring the times involved. In additional experiments, we examined the effect of bar length and luminance profile. These results allow us to estimate the gradients of the attentional fields.

Many parameters have been investigated as to their effect on the way in which the visual system is able to integrate different motion directions at the same visual location. Of special interest have been parameters that determine the depth relationship between surfaces, such as disparity, relative contrast, and occlusion versus transparency. The preferred stimulus for this research has been the 'plaid', usually constructed from two linear gratings. The present study concentrated not on these Cartesian plaids, but on polar plaids, made from a combination of concentric circles and radial gratings. These kinds of plaids also have a special theoretical significance: within the Lie Transformation Group approach to visual pattern processing, Cartesian and polar stimuli represent different invariances in the visual world. This study compared Cartesian, polar and hybrid plaids as to their propensity to be perceived as coherently moving stimuli. Cartesian and polar plaids were similar in terms of the effects of intersection luminance and relative contrast on coherence, polar plaids being consistently less coherent. Hybrid plaids did not usually cohere at all. Adaptation to an unambiguously coherent plaid decreased perceived coherence when tested with a bistable plaid from the same, and not from the other Lie group, i.e. there was within-group adaptation but no betweengroup adaptation. Polar plaids also offer the possibility of studying the influence of another depth parameter on motion integration: expansion or contraction of circular gratings, which represent motion-in-depth toward or away from the observer. This motion-in-depth was tested for interaction with disparity or relative contrast in the determination of motion integration. The results were negative under the present conditions. Thus not all depth parameters contribute equally to the determination of the stimulus depth relations affecting the motion integration process.

Many moving plaid stimuli are ambiguous, and perception switches between a coherent plaid pattern and two transparent gratings. Here, experiments are reported that examined the effect of stereodepth between the two gratings of the moving plaid stimulus on the perception of coherence or motion transparency. Increasing disparity increased the percentage of time that two independently drifting transparent gratings were perceived. This was studied for plaids with various levels of intersection luminance. Using intersection luminances beyond conditions of physical transparency increased the percentage of time that one coherent plaid was seen. These two opposing influences could be pitted against each other to achieve constant levels of coherence. An adaptation paradigm was also used in which observers adapted to a stationary stimulus with either zero, crossed or uncrossed disparity between the gratings, and then indicated the occurrence of coherence and motion transparency in test stimuli of drifting plaids with zero, crossed or uncrossed disparity. Adaptation to crossed and uncrossed stereodepth increased relative perceived coherence equally, especially for zero test disparity. An analysis of the length of the episodes of coherence and motion transparency indicated that the effect of adaptation was to decrease the length of motion transparency episodes, while the length of coherence episodes did not change. It is concluded that mechanisms involved in the processing of stereodepth must have an input to the integration stage of the motion channel and that pattern and component motion mechanisms can operate quite independently.

Two experiments examined the effects of pattern vs component adaptation on motion integration in stimuli with or without disparity. Observers adapted to either downward pattern motion, downward component motion, or a grey screen, and were then tested with plaids containing either crossed, uncrossed or zero binocular disparity, moving downward. In a second experiment, the same test conditions were employed following adaptation to upward pattern motion. The total amount of time that coherence or transparent sliding was perceived was measured. Adaptation to component motion increased the amount of perceived coherent motion whereas adaptation to pattern motion decreased it. The second experiment revealed that adaptation to the upward-moving pattern had no effect on perceived coherence. The results demonstrate the complex nature of the interaction between depth and motion mechanisms in motion integration. Under prolonged inspection, a moving squarewave plaid 1 yields bistable percepts: Perception switches between separate gratings sliding over one another in their respective directions, and a coherent plaid moving in a third direction. The "separate" percept implies that each grating is processed by an appropriate direction-sensitive mechanism and that both are operating simultaneously. This is also referred to as the stage of component motion. The "coherent" percept implies a further stage of processing, whereby the two component gratings are integrated into a new pattern, a plaid, which moves in a composite direction, different from the two original directions. This stage has been referred to as pattern motion, and models have been developed to describe the perceived pattern direction (Adelson & Movshon, 1982; 1 Plaid stimuli are constructed by superimposing two grating stimuli of different orientations that may also differ in terms of their spatial frequency and contrast and other parameters. Each individual grating moves in a direction perpendicular to its orientation. The areas where the grating lines cross each other are called intersections and have a luminance that may differ from those of the peaks and troughs of the gratings, thus giving the plaid-like appearance of the compound stimulus.