VSS 2003 - Vision Science Society 2003 Conference Abstracts

Kushnier, A. & Pylyshyn, Z. (2003) Can flashing objects grab visual indexes in multiple object tracking? Vision Sciences 2003, Sarasota, FL.

Visual Indexing Theory (Pylyshyn, 1989, 2001) assumes that events such as flashes cause indexes to be automatically assigned, or "grabbed," by the flashed objects. It is also assumed that once assigned, indexes stay assigned to the same individual objects as these objects move around in the visual field, and even when they briefly disappear, thus accounting for the high performance observed in multiple object tracking (MOT) studies. In the present study we ask whether already assigned indexes can be draw away by flashing nontargets during tracking. Method: Observers tracked 4 targets among a total of 8 identical randomly-moving objects. Approximately midway through the 6 s tracking trial, 4 objects flash on and off for 300 ms: two of these were targets being tracked and the other two were nontargets. We examined only those trials (37% of the total trials in our case) in which exactly one target was lost and replaced by one nontarget, since in those cases it was unambiguous which target had been lost and which nontarget erroneously selected as its replacement. For these trials we asked: (1) was the dropped target more often one that had flashed than one that had not? and (2) was the erroneously substituted nontarget more often one of the distractor objects that had flashed than one that had not. Results: The results showed that (1) the target object that was lost was no more likely to be one that had flashed than one that had not flashed, and (2) the nontarget that was erroneously selected was significantly more often one that had flashed. We interpreted these results to suggest that once objects were indexed for tracking, flashing them did not tend to cause the index to be dropped, but flashing nontargets tended to draw indexes away from tracked targets.

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP

Keane, B. & Pylyshyn, Z. (2003) Does tracking disappearing objects in MOT involve predicting the locus of reappearance? Vision Sciences 2003, Sarasota, FL.

Purpose. In Multiple Object Tracking (MOT), subjects follow a flashed subset of identical visual objects that move independently about a display. It is known that under some conditions it is possible to track objects even when they completely disappear from view (e.g. Scholl and Pylyshyn (1998)). Our primary aim in this study is to examine whether tracking mechanisms are predictive, and, in particular, whether subjects track better when objects reappear at a location predicted by their trajectories rather than at some other location, such as at the location at which they disappeared. A secondary aim is to see for how long objects can disappear without significantly interfering with tracking. Methods. Four of eight objects momentarily flash, and subjects are asked to track the flashed objects for 5 seconds. Midway into each trial, all objects on the screen disappear and reappear either a) at point of disappearance ("non-move" condition); or b) at a position consistent with their trajectory ("move" condition). Three disappearance intervals were used: 150ms, 300ms, and 450ms. Results. Subjects tracked better in the non-move condition than in the move condition for all disappearance durations. For the move condition subjects tracked best in the shortest disappearance interval and worst in the longest. There was no significant difference between performance at the different disappearance durations for the non-move condition. Conclusion: The main conclusion suggested by this experiment is that objects are not tracked predictively for disappearance durations we have examined (up to 450 ms). Subjects appear to only keep track of where objects disappear and not where they are likely to reappear.

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP

Pylyshyn, Z. & Leonard, C. (2003) Inhibition of Nontargets During Multiple-Object Tracking. Vision Sciences 2003, Sarasota, FL.

In Multiple Object Tracking (MOT), observers keep track of a set of (about 4) targets which move randomly and independently among a set of identical nontargets. Earlier we reported that keeping track of the identity of targets (as measured by recall of their previously assigned labels) tended to be poorer that keeping track of their status as targets, and that ID errors tended to increase more rapidly with trial length. We provided evidence that this was likely due to the fact that targets tended to be more frequently confused with other targets than with nontargets.
A possible reason for this asymmetry is that nontargets might be actively inhibited in MOT, thus resulting in their being less readily confused with targets. In the present study we investigated this hypothesis using the dot-probe method first adopted by (Watson & Humphreys, 1997). Observers viewed an MOT display during which they had to detect the brief appearance of a small spot that occurred on 50% of the trials at one of 3 location types: on a target, a nontarget, or somewhere in between. At the end of each trial, observers selected the targets using a mouse, and then indicated whether or not there had been a probe during that trial.
We found that accuracy for detecting a probe was significantly worse when the probe was on a nontarget. No other pairwise contrast was statistically reliable. Tracking performance also appeared to be the same for all probe locations as well as for the no-probe condition.
These results are consistent with the hypothesis that nontargets are inhibited during MOT, thus accounting for the asymmetry between target-nontarget and target-target confusions. They are also consistent with the hypothesis that MOT does not consume the same visual-attention resources as monitoring a visual event. These results are discussed in relation to the FINST Visual Indexing Hypothesis.

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP

Annan, V. & Pylyshyn, Z. (2003) Voluntary Indexing requires serial visitation. Vision Sciences 2003, Sarasota, FL.

Last year, we showed that targets in Multiple Object Tracking (MOT) can be indexed voluntarily as well as involuntarily. However, whereas involuntary indexing is automatic and simultaneous, voluntary indexing requires that each target object be visited serially. This year we present further evidence to support this claim.
In experiment 1, the observer was asked to track 3 out of 8 gray, disk-shaped, objects on a black screen. The targets were either the objects on which a vertical line flashed (flash condition), or the objects on which the vertical line did not flash (nonflash condition). We varied the amount of time available for indexing by having the vertical line flash once (300 ms) or three times (900 ms). We hypothesized that if serial visitation was necessary, then we should see better performance with the longer time indexing interval. As expected, for the flash condition, the results showed no difference in tracking performance across indexing intervals. For the nonflash condition, tracking improved with the longer indexing intervals. This is consistent with the serial visitation hypothesis. However, a similar result could also be obtained if the automatically assigned indexes first had to be detached from the flashed objects and re-applied to the nonflashed target objects, a process which could itself explain the additional time it takes to index nonflashed items. Experiment 2 controlled for automatic indexing by having horizontal and vertical lines flash on all 8 objects simultaneously, instead of requiring observers to track nonflashed items. Observers had to track the 3 objects that were cued with either horizontal lines or vertical lines. The results of this experiment show the same pattern of performance as the nonflash condition, suggesting serial visitation. This study provides further evidence in support of a serial visitation explanation of voluntary indexing.

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP

Leonard, C. & Pylyshyn, Z. (2003) Measuring the attentional demand of multiple object tracking (MOT). Vision Sciences 2003, Sarasota, FL.

Recently we reported on an attempt to measure the attentional demand of the Multiple Object Tracking (MOT) task, which many people have claimed requires considerable attentional resources. The experiment we reported appeared to support this contention since tracking performance was worse when carried out simultaneously with a color-change monitoring task. However the monitoring task we used required an immediate response, so it is possible that this decrement in tracking was not due to the attentional requirement of visual monitoring, but to the requirement of responding while tracking. Method: In the present studies we compared performance on MOT while monitoring for a brief change in the color of targets, nontargets or an area-matched fixation point. We also measured baseline tracking performance without monitoring, or on trials when a monitored event failed to occur. We measured dual task performance under both an immediate-response (forced choice) condition and a delayed response condition, in which the observer had to indicate whether a color change had taken place after completing a tracking trial. Results: We found that tracking performance was worse when monitoring than in a baseline control, but only when the response had to be made immediately, during tracking. When the response to the monitoring task was made at the end of each trial no decrement in tracking performance was observed and all monitored locations led to the same tracking performance. This finding is compatible with other findings (e.g., Alvarez, Horowitz & Wolfe, 2000) showing that observers can track without decrement while engaged in a secondary task such as search. It is also compatible with the view (Pylyshyn, 2001) that tracking involves a preconceptual mechanism. These findings suggest that apparently attentional tasks such as MOT may not draw on the same resource as other apparently attentional tasks such as monitoring or search.

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP

Dennis, J. & Annan, V. (2003) Can visual objects be accessed in rapid counting without their positions being encoded? Vision Sciences 2003, Sarasota, FL.

Previous research has shown that "subitizing", i.e., the process of enumeration when there are fewer than 4 items, is accurate and relatively rapid (70 ms/item) whereas "counting", i.e., the process of enumeration when there are more than 4 items, is slow (300 ms/item), and error-prone. Trick & Pylyshyn (1994) have hypothesized that subitizing requires preattentive individuation and access by a visual index, as assumed to be provided by the FINST indexing mechanism (Pylyshyn 1989). Their assumption is that subitizing relies on a process of enumerating indexes without the need to encode object locations. Slower counting, on the other hand, relies on scanning attention to each of the objects in the display, and thus may result in the encoding of the objects' locations. To explore this proposal we presented observers a total of 10 objects (small rectangles). A subset of 3, 4, or 5 of these (the "targets") were a different color from the rest of the objects (they were either red or green), and all objects were present for 300 ms before they changed to a neutral gray. Subjects were instructed to count the number of objects of the specified color as fast as possible. After indicating how many targets there were (by pressing keys marked 3, 4 or 5), observers selected the targets by clicking on the objects that had been the target color. If the rapid counting involved in this procedure (135 - 300 items/sec) does not require accessing the objects by scanning the display, then observers may be able to enumerate objects without the opportunity to encode their locations. Hence they may be able to count objects without correctly locating them or to locate objects without having counted them. Errors in counting need not be accompanied by errors in locating objects. We found that when subjects made counting errors there was only a 0.19 probability that they would also make a location error. The large difference between expected and observed errors will be discussed.

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP

Bullot, N. (a,b,c), Droulex, J. (a), & Pylyshyn, Z. (c) (2003) Keeping track of objects while exploring an informationally impoverished environment: Local deictic versus global spatial strategies. Vision Sciences 2003, Sarasota, FL.

(a) Laboratoire de Physiologie de la Perception et de l'Action, Coll├Ęge de France, 11 place Marcelin Berthelot, 75005 Paris, France.
(b) Institut Jean Nicod, CNRS-EHESS-ENS, 1 bis avenue Lowendal, 75007 Paris, France.
(c) Rutgers Center for Cognitive Science, Rutgers University, USA.

This study investigates a new experimental paradigm called the Modified Traveling Salesman Problem (Bullot & Droulez, submitted). This task requires subjects to visit once and only once n invisible targets in a 2D display, using a virtual vehicle controlled by the subject. Subjects can only see the directions of the targets from the current location of the vehicle, displayed by a set of oriented segments that can be viewed inside a circular window surrounding the vehicle. Two conditions were compared. In the "allocentric" condition, subjects see the vehicle move across the screen and change orientation under their command. The "egocentric" condition is similar except for how the information is provided: the position and orientation of the vehicle icon remains fixed at the center of the screen and only target directions, as indicated by the oriented segments, change as the subject "moves" the vehicle. The unexpected finding was that this task can be performed, in either condition, for up to 10 targets. We consider two possible strategies that might be used, a location-based strategy and a segment strategy. The location-based strategy relies on spatial memory and attempts to infer the locations of all the targets. The segment strategy is more local and focuses on the directional segments themselves, keeping track of the ones that represent already-visited targets. A number of observations suggest that the segment strategy was used, at least for larger numbers of targets. According to our hypothesis, keeping track of the segments requires one to use indexical reference for associating the segments with their status in the task - given by current status predicates Visited(x) or Not-visited(x) -, perhaps using visual indexes (Pylyshyn, 2001), deictic pointers (Ballard et al., 1997), or object files (Kahneman et al, 1992).

Credit: This research was supported by NIMH Grant R01-MH 60924 to ZP


Contact RuCCS

Psychology Building Addition
152 Frelinghuysen Road
Piscataway, NJ 08854-8020


  • 848-445-1625
  • 848-445-6660
  • 848-445-0635


  • 732-445-6715