Demos: Effects of Grouping and Attention on the Perception of Causality
This page contains demonstrations discussed in the following paper:
Choi, H., & Scholl, B. J. (2004). Effects of grouping and attention on the perception of causality. Perception & Psychophysics, 66(6), 926-942.
These demonstrations are provided as Quicktime movies, which can be downloaded or viewed directly in most web-browsers. In some cases the spatiotemporal parameters may be slightly different than those used in the experiments, but they should be sufficient to illustrate the basic effects.
In each case, you should judge the causal status of the specified event: does it look like a causal launch (wherein one item collides with another, causing its motion) or a non-causal pass (wherein one moving item simply passes across the entire display, passing over a stationary item -- and possibly swapping colors with it)?
The overall goal of these studies is to demonstrate and explore several new aspects of the perception of causality. In particular, we demonstrate here for the first time that several types of perceptual grouping can influence causal perception. We further suggest that such grouping effects are mediated by attention, and we directly demonstrate that causal perception can be strengthened or attenuated based on where observers are attending.
Basic Launching and Ambiguous Full-Overlap Displays
Experiment 1: Dumbbells
This is a standard launching display, a la Michotte. Beyond the kinematics involved in the event, you see causality: one object collides with another, causing its motion.
Ambiguous Full-Overlap Event
This event -- first used in Scholl & Nakayama's causal capture demonstrations -- serves as a baseline condition for many of our studies. It is identical to the basic launching event, except that the two discs fully overlap before the second motion. While this ambiguous event may sometimes still be seen as causal launching, many observers can perceive it as non-causal passing (as described above). (If you find it difficult to perceive passing, put the event in the periphery. Note that you can see passing here more easily, even when you can still discriminate the different colors.)
Several of our studies also involve adaptations of Scholl & Nakayama's causal capture effect, illustrated here. In this movie you should still be judging the causal status of the top event -- which is in fact identical to the ambiguous full-overlap event. Now, though, in the presence of the lower 'context' event which is perceived as an unambiguous launch, the causal status of the top event is 'captured' so that it too is seen as robustly causal (in contrast the noncausal 'pass' percept which is easily seen without the context). (Observers who tend to see the full-overlap event as causal even when it is presented alone may notice that they can at least force themselves to see it as a non-causal pass. In contrast, doing so is nearly impossible here, even with the identical stimulus.)
Baseline Unconnected Condition
Experiment 2A: Basic Grouping Effects
Here you should judge the causal status of the upper full-overlap event. This is a variant of the causal capture display, but it has been weakened in two ways: (1) The spatial separation between the two events is increased; and (2) the grouping by common motion is weakened by having the two events move in opposite directions. Causal capture is much weaker here, and a majority of observers can perceive non-causal passing.
This display is identical, except for the added single-pixel line -- which makes a tremendous difference, promoting the perception of causal launching in the top event. (See the paper for a description of additional conditions not shown here, manipulating the connectedness and the temporal dynamics of the added line.) This initial condition demonstrates that the perception of causality can be affected by perhaps the strongest possible grouping cue -- physical connectedness.
Baseline Full-Overlap Event
Experiment 2B: Varying the Number of Contextual Discs
This is the same event referred to above: the ambiguous full-overlap event which is often seen as non-causal passing.
Same-Motion Grouping Condition
Here you should judge the causal status of the bottom pair of discs, which is the same full-overlap event. The added discs, however, group with the initially-central disc in this event via proximity, good continuation, and common motion. The result is that observers are much more likely to perceive causal launching (for reasons discussed at length in the paper). This demonstrates that in some cases the underlying units of causal perception can be groups as well as objects.
No-Motion Grouping Condition
Here the additional discs simply remain stationary at the center of the screen. This changes the grouping over time, and observers are still easily able to perceive non-causal passing. (See the paper for a discussion of additional conditions not shown here, manipulating the direction of motion of the additional discs.)
1-Disc Grouping Condition
Experiment 2C: Varying the Proximity of Contextual Discs
Observers are more likely to perceive causal launching because of such grouping effects even in this case, where only a single disc has been added to the display. (Cover this disc from view with your arm, and it should be easy to see non-causal passing. Watch the whole display, and you should more readily perceive causal launching.) This demonstrates the power of grouping in this context: the addition of even a single additional object in a display can dramatically affect the perception of causality.
Experiment 3: Reverse Causal Capture
Here the grouping effect is attenuated by increasing the distance between the grouped items, demonstrating a role for proximity in addition to common motion.
Partial-Overlap Launching Condition
Experiment 4a: Effect of Attention on Perceived Causality
In this event the two discs only overlap partially, but observers still tend to perceive robust causal launching when it is presented in isolation.
Reverse Causal Capture
Here the same event is presented (on the bottom, with the blue disc), but an additional full-overlap event is added to the display. This full-overlap event is likely to be seen as non-causal passing in isolation, and we find that in its presence, the partial-overlap event can now also be seen as non-causal passing. This is the first demonstration that contextual information can actually attenuate the perception of causality in an event which is reliably seen as causal launching in isolation. As such, we dub this the 'reverse causal capture' effect in the paper. (See the paper for several additional studies of reverse causal capture, exploring the role of proximity.)
Attending to the Test Event
Experiment 5: Effects of Attention to Competing Context Events
We suggest that all of the grouping effects reported in the paper may be explained in terms of the operation of visual attention, and in this experiment we directly demonstrate that causal perception can be influenced by the allocation of attention within a scene. Using a 'reverse-causal capture' display (as described above), we use a precue to tell observers where to attend. (Observers always report their perception of the causal status of the event containing the blue disc, regardless of whether it appears as the top event or the bottom event.) Here that precue is a flashing yellow arrow that tells subjects to attend to the partial-overlap test event, and observers often perceive causal launching.
Attending to the Context Event
In this example the cue tells subjects to attend to the full-overlap context event, and subjects now perceive the test event as noncausal passing on a majority of trials (illustrating 'reverse-causal capture'). Thus we find that the degree of perceived causal launching in the very same display can modulated by more than 50% based only on where subjects attend. This result demonstrates that the 'input' to the processes which mediate causal perception is filtered through attention.
Attending Between the Two Events
In this example the cue tells subjects to attend to the space in between the two events, resulting in an intermediate degree of perceived causal launching in the test event. (See the paper for additional intriguing results related to anisotropies in attentional resolution.) These massive effects of attention help to unify the experiments presented in the paper: Experiment 4 demonstrated an effect on causal perception of voluntary attentional allocation (via the task instructions and the cued row), while the other experiments demonstrated effects of automatic attentional spread, induced by the connectedness (in Experiment 1) and other grouping cues (in Experiment 2). Thus the seemingly disparate experiments reported here may in fact all reflect the same underlying process: the voluntary or automatic spread of attention
Attending to Competing Contexts
Here, in another demonstration of the influence of attention on causal perception, we show that attention can also select among competing context events. Observers always fixate and judge the causal status (launching vs. passing) of the central ambiguous full-overlap event. This test event is flanked by two other context events, which compete to influence the central event, and subjects are induced to attend to only one of them (while continuing to fixate and judge the central event). When the 'Unambiguous passing' context event (here presented below the test event) is attended, observers tend to see the test event as passing. When the 'Unambiguous launching' context event (here presented above the test event) is attended, however, observers are more likely to see the test event as causal launching. This effect illustrates the influence of attention on causal perception even when both the fixation position and the display itself remains constant.
Like Michotte, we find that the perception of causality is mediated by strict visual 'rules'. Beyond Michotte, we find that these rules operate not only over discrete objects, but also over perceptual groups, constrained by the allocation of attention. This work demonstrates that the perception of causality interacts richly with other well-known visual processes.