This page has demonstrations of three tasks using the apparent motion technique: 1. color-motion feature binding, 2. crowding, and 3. sequential binocular rivalry. You will need a Quicktime plugin in your browser for these movies to play. If you do not have one, you can download one from <click here> Apple. Even with the right plugin, the movies do not match the quality of the experimental displays and, depending on many factors, may be severely degraded for your display. So if these do not play well on your browser or computer, my apologies. Another computer may play them better. Even if they do not play well, they should help show the arrangement of the displays even if they do not show exactly how they look. For each movie, click on the start button at the lower left of the movie and click it again to stop before going on to the next movie.
In each patch, color alternates between red and green and motion alternates between inward and outward. In half the trials in the experiment, red is moving inward (and green outward) and in the other half, green is moving inward (and red outward). In the demonstrations here, only one of the color-motion pairings is shown. The task is to report the direction of the red dots while fixating the central cross. In the static attention condition, one location is attended (we asked subjects to pick the bottom location, but if they noticed any other, that would be OK). In the moving condition, they followed the guide ring with attention (still fixating the central cross), and reported the motion within the ring.
In this first example, alternating at a rate of 4 Hz, it is difficult to report the pairing when fixating the central dot and attending to a single alternating patch. Click on the play button on the lower left to start the movie.
If we now add a guide that jumps from location to location in the same stimulus, it appears that one item is moving within the ring. The guide steps are synchronized to the feature alternations, so that the ring always encircles the same value, allowing moving attention to sample from the same feature pairing on successive frames. Start this next movie, fixate again and follow the ring, attending to the feature values inside the ring. For all subjects, the guide substantially improved their accuracy of report. For some subjects, this 4 Hz rate is too fast. Additional pairs of the guided and unguided movies at 2 Hz and 3 Hz are available by clicking here: 2 Hz, 3Hz. At the slowest rate, it is already easy to report the color and motion pairing with attention to a fixed location, so the ring does not add anything to the performance at that speed.
Does the advantage of moving attention come from simply removing the alternation of colors and motions in the attended stream or does it represent an accumulation of partial information from successive samples? To test this, we presented the moving guide crossing over 1 to 5 items and found the performance increased from near chance with only one item (even thought there were no preceding or following colors or motions in the attended stream to interfere with the identification of the single item sampled). Performance then rose to asymptote with 3 or 4 items. Here is the display with 3 items at 4Hz. Again fixate the central dot. Displays with 1 and 5 items can be seen by clicking here.
Perhaps the ring itself provides the 200 to 300% performance increase (in maximum rate for 75% correct) because it is an exogenous cue and it would help no matter where it was presented. However, the following controls suggest that the ring on its own has litle effect, whether at a single location or at many. The important factor for the guide appears to be a predictable path that samples the same value on each successive frame.
1. A single guide ring on one location, every second frame, 4Hz. Subjects little better than chance at 4 Hz. Click here.
2. Guide rings on all locations, always encircling red-inward, 4 Hz. This is also no help except for some observers who find they can see all the guide circles moving around the center together, either clockwise or counterclockwise. In this case, the pairing of color and motion is again easier to pick out. Click here.
3. Random guide locations, 4Hz. No improvement in performance. Without a predictable trajectory, attention may not get to the cued locations in time to pick up the encircled stimulus. Click here.
In each of sixteen radial arms, a single target letter (3rd position) alternates with 4 flanking distractors (positions 1, 2, 4, and 5). While fixating the center dot and attending to one radial arm, the flickering target is flanked by flickering distractors. The task is to report whether the letter is normally oriented or left-right reversed. The target in each arm is different (randomly) but they all have the same orientation. In the fixed attention condition, the subject is asked to report the target orientation from the bottom position, but if any other target is noticed first, that is OK.
In the next display, a light sector encloses one radial arm at a time to guide attention from one arm to the next in synchrony with the target-distractor alternation. As a result, only the target is seen on each arm as it is selected in turn. If crowding is retinotopic, the target should be as degraded here as it is with fixed attention. If crowding is specific to the moving attention window, there should be less crowding since the moving window avoids the distractors. To make a fair comparison to the fixed attention case, the target array rotates with the guide so that the same target can be sampled successively during the trial (as is the case in the fixed attention). Further controls will test the effect of moving the target array. Fixate in the center and track the moving sector with attention.
The eight colored patches are shown first to one eye and then to the other at 3 Hz. The two eyes see different colors and orientations. At frequencies below about 3 Hz (O'Shea & Crassini, 1984), subjects report veridical alternation of the two colors. At higher frequencies, subjects often report rivalry: a flickering green stimulus is seen for a while followed by a flickering red stimulus. The red - green alternation is then slower than the veridical rate. In this display at 3 Hz (here for free-fusers, in our experiments we use mirrors to superimpose the left and right images) while fixating in the center, most but not all subjects see red-green rivalry when attending to one location (it may take a few seconds to appear). A few subjects require a higher frequency.
In the next display, a thin blue line points at successive locations in synchrony with the alternations. This indexed item here is always red in one eye and blank in the other. If attention follows the pointer around the display, it samples the current state of rivalry at each location. If rivalry occurs in unattended locations and is undisturbed by the arrival of attention, the apparent color at the end of the pointer should alternate between red and green. To test this, try following the blue pointer (while fixating the center) around the display in the next movie.
Subjects so far report strong, often complete dominance of the red (if the pointer is set on red, green if set on green) with little or no evidence of rivalry. We will test the possible influence of moving attention on the persistence that underlies the sequential rivalry (persistence links successive reds into a flickering red and blocks the visibility of the green of the other eye during the blank between the reds). If the absence of attention affects the persistence at unattended locations, sequential rivalry would fail for "technical" reasons. So far we find no effect of the moving guide on persistence itself. In a subsequent control, we will test whether attention's arrival at each location might reset rivalry. Finally, recall that we have concentrated on sequential rivalry because regular rivalry is unsuited to this moving attention paradigm. The sequence of stimuli in the attended stream (say, red versus green) will the same whether attention is moving or fixed. Only the sequential rivalry task allows us to dissociate the stimulus stream in the moving and fixed cases (as we did with crowding and feature binding).