Perceived Causality Can Alter the Perceived Trajectory of Apparent Motion

Abstract

When objects collide, human observers perceive not only motion but also causal relations, such as which object caused the other to move. In the present experiments, we investigated whether such causal interpretations can actually influence the perceived path of apparent motion. Displays contained two alternately flashing motion targets positioned at either end of a semicircular occluder. Two additional “context objects” moved in such a way that the motion targets appeared to collide with and launch them. The collision was manipulated so that it was consistent with apparent motion either along the straight path between the targets or along a curved path passing behind the occluder. Subjects almost exclusively perceived motion consistent with the implied launch, which suggests that causally coherent interpretations can influence basic perceptual processes.

Keywords

motion perception causality visual perception cognition

Comprehension of physical events in terms of cause and effect is fundamental for making sense of the dynamic physical world. Causal interpretation may involve cognitive inference based on previous experience and real-world knowledge, but at least some situations give rise to more perceptual causal impressions that seem immediate, automatic, and irresistible. When a moving object suddenly stops adjacent to a stationary object and then the stationary object starts to move along the same path that the previously moving object was following, most observers tend to see the motion of the second object as caused by the impact of the first one (the launching effect; Michotte, 1946/1963). Ever since Michotte’s seminal work, most studies of causal perception have focused on determining the factors that mediate such phenomenological experiences of causality (e.g., Boyle, 1960; Choi & Scholl, 2004, Guski & Troje, 2003; Schlottmann & Anderson, 1993; Scholl & Nakayama, 2004; for a review, see Scholl & Tremoulet, 2000). In such studies, the perception of causality is implicitly considered a global interpretation or “Gestalt” imposed on a visual event. In the experiment reported here, we investigated whether the perception of causality can play a more substantial role in influencing basic perceptual processes.

To assess how the perception of causality influences motion interpretation, we capitalized on the well-known phenomenon of apparent motion: Alternate presentations of two spatially distinct objects produce the percept of intervening continuous motion (Kolers, 1972; Wertheimer, 1912/1961). Apparent motion involves an inherent ambiguity of the motion path because an infinite number of potential paths connect the two locations. Observers typically perceive the shortest possible path, generally a straight line. In some circumstances, this shortest-path bias can be overcome to yield a perception of a longer, curved path (Anstis & Ramachandran, 1985; Farrrell & Shepard, 1981; Kim, Feldman, & Singh, 2012; Shepard & Zare, 1983; Shiffrar & Freyd, 1990). The motion-path ambiguity provides an ideal context in which to investigate the influence of perceived causality on motion perception. In the present experiments, we tested whether perceived causality can push apparent motion away from the default shortest path and toward a longer, curved one. As in our previous studies (e.g., Kim et al., 2012), we employed a curved occluder, which establishes a potential “detour” for the motion path. Moving objects are often occluded by other surfaces that they pass behind. But the visual system generally maintains a percept of continuous existence behind occluders, a phenomenon termed amodal completion (Michotte, Thinès, & Crabbé, 1964/1991). In the present experiments, we tested whether the perception of momentum elicited by the collision of an apparently moving object with a second object can reciprocally interact with the amodal completion of the first object’s motion behind a curved occluder to induce a perceived motion path consistent with the direction in which the second object is launched.

In one type of apparent-motion display (see Fig. 1a), two rectangles of the same color (context objects; shown in green) appear immediately above two rectangles of a different color (motion targets; shown in red). The two red rectangles flash in alternation, with a variable interstimulus interval (ISI) between the disappearance of one rectangle and the appearance of the other. As soon as a red rectangle appears, the green rectangle above it moves upward and then reverts to its original position during the ISI. This display is typically interpreted by naive observers in one of two ways. Some observers perceive a red rectangle apparently moving along a straight path between the two positions, with the green rectangles moving up when the red rectangles appear beneath each of them and then moving back down when the red rectangles disappear, without an impression of collision. Other observers see two red rectangles alternately emerging into existence to “push” the green ones upward and then disappearing, without a strong impression of apparent motion. Hence, this display does not seem to simultaneously support both apparent motion (of the red rectangle) and a causal interaction between the red and green objects. However, if a semicircular occluder is included in every frame of the same motion sequence (vertical-launch display; Fig. 1b), observers tend to see the red rectangle moving back and forth behind the occluder and colliding with the green rectangles at each end of the occluder to launch them upward. The red rectangle then appears to reverse direction behind the occluder while the green rectangle “falls” passively down to its original position apparently because of gravity.

Experiment 1

In Experiment 1, we tested the effect of perceived causality on apparent motion using several types of context events, including the vertical-launch display (Fig. 2). In bounce displays, the context objects were stationary across the entire motion sequence. In launch displays, as soon as a motion target appeared, the context object above or next to it respectively moved upward (vertical-launch display) or laterally (horizontal-launch display). Horizontal displacement of the context objects is consistent with straight horizontal apparent motion, which appears to launch the context objects horizontally. Similarly, vertical displacement of the context objects is consistent with a curved motion path behind (or through) the occluder that results in vertical motion at the end of the occluder, which appears to launch the context objects vertically. Thus, the question was whether the perceptual interpretation of collision could bias the perceived path of target motion.

Method

Participants

Thirteen Rutgers University undergraduates participated in Experiment 1 for course credit.

Materials and design

The experiment was controlled by a program written with MATLAB (The MathWorks, Natick, MA) using the Psychophysics Toolbox extensions (Brainard, 1997; Pelli, 1997). Participants were seated in a dark room approximately 60 cm in front of a computer monitor.

Stimuli were computer-generated apparent-motion sequences of two alternately presented motion targets (Fig. 2), each of which was separated in time by two frames of ISI. Motion targets were textured rectangles composed of red-and-white dot patterns (1.00° × 0.61° of visual angle), which appeared on top of the center of each end of an upturned semicircular tube (width = 2.2°) that had a radially white-to-black-gradient color. The diameter of the semicircular tube (equivalent to the separation between the two motion targets) was 5.73°. There were four levels of ISI (100, 233, 367, and 500 ms). Target duration was 150 ms. The semicircular tube remained on screen during the whole motion sequence. The entire stimulus display appeared against a black-and-white background of random dots.

There were five display types (Fig. 2; see Videos S1–S5 in the Supplemental Material available online for animated demonstrations of each display type). In no-context displays, motion targets appeared without context objects (textured rectangles composed of green-and-white dot patterns). In the other four displays, two additional context objects appeared on either side or above the locations of motion targets. In vertical-bounce displays, context objects were placed immediately above each motion target; in horizontal-bounce displays, context objects were placed on the outer side of each motion target. In both types of bounce display, the context objects remained visible across the entire motion sequence without any location change. In vertical-launch displays, context objects appeared immediately above the two motion targets, and 50 ms after target onset, the context object was displaced upward by 1.8°, and then returned to its original position 50 ms before the next target onset. In horizontal-launch displays, context objects appeared on the outer side of the two motion targets and were displaced laterally by 1.8° for the same duration as in vertical-launch displays.

Procedure

Each trial started with a fixation cross for 500 ms, followed by three cycles of the eight-frame motion sequence, then a blank screen. Participants pressed one of two keys to indicate whether the motion target appeared to be moving horizontally or along a curved path behind (or through) the semicircular tube (see Fig. 3a). Participants were additionally asked to rate how strongly and clearly they perceived apparent motion on a 3-point scale. The main dependent measure was the proportion of “curved-path” responses, with a secondary analysis of the motion-quality ratings. Each participant completed a total of 200 experimental trials (5 display types × 4 ISIs × 10 repetitions) presented in random order over four blocks, following 26 practice trials.

Results and discussion

Results of Experiment 1 are depicted in Figure 3b. A 5 × 4 repeated measures analysis of variance (ANOVA) was conducted to determine the effects of display type and ISI, respectively, on motion path. The main effect of display type was significant, F(4, 48) = 5.84, p = .006; observers reported curved-path motion far more when viewing the vertical-launch display (M = .98) and least when viewing the horizontal-launch display (M = .07) compared with the other three types of displays. Planned comparisons confirmed that the proportion of “curved-path” responses to either launch display significantly differed from the proportion of “curved-path” responses to any other displays, ps < .001. That is, apparent motion was perceived far more often in the direction of the launch. The main effect of ISI was also significant, F(3, 36) = 17.51, p < .001; “curved-path” responses increased with ISI. In the displays in which context objects were absent or stationary, the probability of perceiving curved-path motion increased with ISI—ISI × Display Type interaction: F(12, 144) = 7.82, p < .001—which suggests that apparent motion can deviate from the shortest path if amodal completion of the motion trajectory behind the curved occluder can “explain” the invisibility of the object during a long ISI (see Kim et al., 2012). However, the identical ambiguous apparent-motion display was perceived unambiguously as either straight-path motion (when context objects moved laterally) or curved-path motion (when context objects moved upward), regardless of the ISI.

The rated quality of apparent motion was higher with stationary context objects (M = 1.77) than without context objects (M = 1.50) and highest with moving context objects (M = 2.49), F(4, 48) = 36.32, p < .001. That is, the causal context substantially enhanced the percept of apparent motion.

These results suggest that observers attributed the displacement of the context object to a collision with the motion target (Michotte, 1946/1963), with the direction of the momentum implied by this collision decisively influencing the perceived direction of apparent motion. In the vertical-launch displays, a potentially ambiguous contact between the motion target and the context object was interpreted so as to be consistent with the momentum conveyed by the perceived motion behind the curved occluder. This inferred momentum then served as an “explanation” of the motion of the context object.

Experiment 2

In Experiment 2, we wished to rule out a potential alternative explanation of these results based on motion priming. Motion priming typically results from motion seen prior to or concurrently with the target motion (e.g., Anstis & Ramachandran, 1987; Pantle, Gallogly, & Piehler, 2000). In contrast, because of the nature of our displays, the choice of motion path was affected by a context motion that occurred after the target motion. Nevertheless, in Experiment 2, we tested three new displays similar to the vertical-launch display used in Experiment 1 but with different spatial or temporal patterns of context events (Fig. 4). These displays were designed to weaken the causal percept but not to substantially diminish motion priming. If the percept of curved motion in the vertical-launch display was simply the result of priming, these displays should elicit it at least as frequently as the vertical-launch display did.

Method

Participants

Nine new Rutgers University undergraduates participated for course credit.

Design and procedure

Diameter (i.e., target separation) and width of the semicircular tube were 4.5° and 1.9° of visual angle, respectively, from a viewing distance of 100 cm. Motion targets subtended 0.77° × 0.47°. The target duration was 220 ms, and there were three ISIs (150, 300, and 450 ms).

The four displays used in Experiment 2 are depicted in Figure 4 (see Videos S6–S9 in the Supplemental Material for animated demonstrations of each display). The vertical-launch display was identical to that used in Experiment 1, except that context objects were displaced upward by 1.55° 70 ms after each target onset. In Experiment 2, we employed three new displays. In the both-object-launch display, both context objects were displaced upward and reverted to their original positions together in synchrony at every target onset. In the early-launch display, each context object moved upward (ISI − 50 ms) before each target onset (i.e., 50 ms after previous target offset) and then continued to float until the next context object moved upward. In the late-launch display, the context object above each motion target was displaced upward 50 ms after each target offset and reverted to its original position at the next target onset. Procedures were the same as in Experiment 1.

Results and discussion

Results of Experiment 2 are presented in Figure 5. Participants reported curved-path motion more often when viewing the vertical-launch displays (M = .94) than when viewing the early-launch displays (M = .64), late-launch displays (M = .56), and both-object-launch displays (M = .50), F(3, 24) = 3.81, p = .023. Planned comparisons showed that the vertical-launch display produced significantly more “curved-path” responses than did all other displays, ps ≤ .02, but no other pairwise difference was significant, ps > .05. As in Experiment 1, the main effect of ISI was significant, F(2, 16) = 7.68, p = .005. The ISI × Display Type interaction was marginally significant, F(6, 48) = 1.97, p = .089. Motion-quality ratings were significantly higher for the vertical-launch display (M = 2.55) than for the early-launch display (M = 2.06), late-launch display (M = 1.85), and both-object-launch display (M = 2.18), F(3, 24) = 7.44, p = .001.

For the both-object-launch display, motion priming should be stronger because both context objects prime the target motion in the vertical direction. However, the causal percept should be weakened because two context objects’ common motion cannot be explained by a collision with a motion target. In addition, early or late launch should selectively weaken the causal impression because causality perception is sensitive to the timing of the launched object’s motion onset (Michotte, 1946/1963), but priming is presumably relatively insensitive to the exact timing of the priming motion. Therefore, the reduced curved-motion percept in the new displays cannot be explained by motion priming. Curved-path motion was still almost exclusively associated with a causal launch.

General Discussion

The path of apparent motion can be influenced by physical constraints, such as object permanence (Gerbino, 1984), occlusion (Anstis & Ramachandran, 1985; Kim et al., 2012), rigidity (Kolers & Pomerantz, 1971), and biomechanics (Shiffrar & Freyd, 1990). The present finding adds another regularity of nature to the list: mechanical causality. Human observers can experience an impression of causality even when the objects involved are not real, and no interaction between them is actually taking place (Michotte, 1946/1963; White, 2006). In spite of this illusory character, perceived causality seems to reflect a reliable mapping between kinematic properties of the stimulus and a physically plausible configuration of forces (Hecht, 1996; Runeson & Frykholm, 1983). Hence, phenomenological causality seems to incorporate real-world regularities to disambiguate impoverished sensory data.

It is interesting that the influence of causal interpretation on perceived motion in our displays was retrospective; that is, the context objects’ displacement occurred after the target appeared. In a physical collision, the motion of the launching object necessarily precedes the motion of the launched object. But, in our displays, the perceived path of the launching object was affected by the direction of the subsequent motion of the context object (the launched object). The visual system appears to reconstruct the most likely causal antecedent (i.e., the most plausible motion path) to explain the observed consequent (i.e., the displacement of the context object) ex post facto. This result is consistent with findings of previous studies suggesting postdictive processes in the perception of both apparent motion (Eagleman & Sejnowski, 2003; Kolers & von Grünau, 1976) and causal collisions (Buehner & Humphreys, 2010; Choi & Scholl, 2006). This retrospective character is broadly reminiscent of Bayesian inference, often referred to as “inverse probability” because it involves estimating causes retrospectively from their effects.

The phenomenon uncovered in these experiments is unusual in demonstrating that causal perception can interact with the dynamics of visual motion, both influencing and being influenced by the motion percept and the amodal motion path, and making an integral contribution to the phenomenology of visual motion. Causality is conventionally regarded as a high-level property that emerges later than low-level percepts such as motion contiguity in space-time (cf. Buehner & Humphreys, 2010; Scholl & Nakayama, 2004). But our results suggest that perceived causal relations among visual items are not merely a summary interpretation imposed on motions already determined by perceptual processes, but rather may make a potentially fundamental contribution to the disambiguation of the underlying sensory signal itself.

Footnotes

Acknowledgements

We wish to thank Thomas Papathomas, Brian Scholl, Philippe G. Schyns, and two anonymous reviewers for helpful comments.

Declaration of Conflicting Interests

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

Funding

This work was supported in part by National Institutes of Health Grant (NEI)EY021494.

Supplemental Material

Additional supporting information may be found at

References

Anstis

Ramachandran

V. S.

(1985). Kinetic occlusion by apparent movement. Perception, 14, 145–149.

Anstis

Ramachandran

V. S.

(1987). Visual inertia in apparent motion. Vision Research, 27, 755–764.

Boyle

(1960). A contribution to the study of phenomenal causation. Quarterly Journal of Experimental Psychology, 12, 171–179.

Brainard

D. H.

(1997). The psychophysics toolbox. Spatial Vision, 10, 433–436.

Buehner

Humphreys

(2010). Causal contraction: Spatial binding in the perception of collision events. Psychological Science, 21, 44–48.

Choi

Scholl

B. J.

(2004). Effects of grouping and attention on the perception of causality. Perception & Psychophysics, 66, 926–942.

Choi

Scholl

B. J.

(2006). Perceiving causality after the fact: Postdiction in the temporal dynamics of causal perception. Perception, 35, 385–399.

Eagleman

D. M.

Sejnowski

T. J.

(2003). The line-motion illusion can be reversed by motion signals after the line disappears. Perception, 32, 963–968.

Farrell

J. E.

Shepard

R. N.

(1981). Shape, orientation, and apparent rotational motion. Journal of Experimental Psychology: Human Perception and Performance, 7, 477–486.

10.

Gerbino

(1984). Low-level and high-level processes in the perceptual organization of three-dimensional apparent motion. Perception, 13, 417–428.

11.

Guski

Troje

(2003). Audiovisual phenomenal causality. Perception & Psychophysics, 65, 789–800.

12.

Hecht

(1996). Heuristics and invariants in dynamic event perception: Immunized concepts or non-statements. Psychonomic Bulletin & Review, 3, 61–70.

13.

Kim

Feldman

Singh

(2012). Curved apparent motion induced by amodal completion. Attention, Perception, & Psychophysics, 74, 350–364.

14.

Kolers

P. A.

(1972). Aspects of motion perception. New York, NY: Pergamon Press.

15.

Kolers

P. A.

Pomerantz

J. R.

(1971). Figural change in apparent motion. Journal of Experimental Psychology, 87, 99–108.

16.

Kolers

P. A.

von Grünau

(1976). Shape and color in apparent motion. Vision Research, 16, 329–335.

17.

Michotte

A. E.

(1963). The perception of causality. New York, NY: Basic Books. (Original work published 1946)

18.

Michotte

Thinés

Crabbé

(1991). Amodal completion of perceptual structures. In Thinés

Costall

Butterworth

(Eds.), Michotte’s experimental phenomenology of perception (pp. 140–167). Hillsdale, NJ: Erlbaum. (Original work published 1964)

19.

Pantle

A. J.

Gallogly

D. P.

Piehler

O. C.

(2000). Direction biasing by brief apparent motion stimuli. Vision Research, 40, 1979–1991.

20.

Pelli

D. G.

(1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437–442.

21.

Runeson

Frykholm

(1983). Kinematic specification of dynamics as an informational basis for person and action perception: Expectation, gender recognition, and deceptive intention. Journal of Experimental Psychology: General, 112, 585–615.

22.

Schlottmann

Anderson

N. H.

(1993). An information integration approach to phenomenal causality. Memory & Cognition, 21, 785–801.

23.

Scholl

B. J.

Nakayama

(2004). Illusory causal crescents: Misperceived spatial relations due to perceived causality. Perception, 33, 455–469.

24.

Scholl

B. J.

Tremoulet

P. D.

(2000). Perceptual causality and animacy. Trends in Cognitive Sciences, 4, 299–309.

25.

Shepard

R. N.

Zare

S. L.

(1983). Path-guided apparent motion. Science, 220, 632–634.

26.

Shiffrar

Freyd

(1990). Apparent motion of the human body. Psychological Science, 1, 257–264.

27.

Wertheimer

(1961). Experimental studies on the seeing of motion. In Shipley

(Ed.), Classics in psychology (pp. 1032–1088). New York, NY: Philosophical Library. (Original work published 1912)

28.

White

P. A.

(2006). The causal asymmetry. Psychological Review, 113, 132–147.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.18 MB

0.00 MB