Phenomenal Causality and Sensory Realism

Kristof Meding, Sebastian A. Bruijns, Bernhard Schölkopf, Philipp Berens, Felix A. Wichmann

TL;DR

This study investigates how visual realism and physical accuracy in simulated events affect our perception of causality. Surprisingly, simply making abstract shapes look like realistic billiard balls decreased perceived causality, but adding realistic physics (momentum, friction) restored and even enhanced it, suggesting a complex interplay between visual fidelity and physical plausibility.

What Problem Does This Paper Address?

Humans are remarkably adept at inferring cause and effect from everyday events, a skill crucial for navigating the world. Albert Michotte pioneered the study of "phenomenal causality" using simple "launching events" – where one object appears to cause another to move upon collision. While these abstract stimuli effectively evoke a sense of causality, it's unclear how realistic visual and physical properties influence this perception. This paper aims to bridge this gap by systematically varying the realism of launching events, from abstract shapes to photorealistic billiard balls with accurate physics, to understand how these factors shape our causal judgments.

Methods

The researchers employed a controlled experimental design using a within-subjects approach. They presented participants with three types of launching stimuli:

  1. Michotte: The classic abstract, featureless disks.
  2. Rendered: Photorealistic billiard balls with realistic visual textures but simplified physics.
  3. Physical: Photorealistic billiard balls with realistic visual textures and physically accurate motion dynamics, including momentum and friction.

These stimuli were presented with varying temporal gaps between the initial object's motion and the second object's movement, a key manipulation in Michotte's original work. Participants rated the perceived causality of each event on a scale of 1 to 9. The study also explored the effect of 2D versus 3D presentation, though this factor was found to be non-significant. Statistical analysis involved Generalized Additive Mixed Models (GAMMs) to account for individual differences and the complex relationships between variables.

Datasets

Not discussed in the paper. The stimuli were computer-generated.

Key Insights and Ideas

  • The "Uncanny Valley" of Causality: Simply increasing visual realism (from abstract disks to photorealistic balls) without corresponding physical accuracy paradoxically reduced perceived causality. This suggests that visual fidelity alone isn't enough; it can even create a sense of "unnaturalness" if not matched by plausible physics.
  • Physics Trumps Visuals (to a point): When realistic visuals were combined with accurate physics (momentum, friction), causal ratings not only recovered but surpassed those of the original abstract Michotte stimuli. This highlights the critical role of intuitive physics in our causal judgments.
  • Core vs. Peripheral Cues: While fundamental aspects like the temporal gap and object interaction (core features) are crucial for causality, peripheral sensory cues (like surface textures and motion dynamics) can significantly influence the strength of the perceived causal link.
  • Cue Conflict Matters: The "Rendered" condition, with realistic visuals but simplified physics, likely created a "cue conflict" where visual expectations clashed with motion dynamics, leading to lower causal ratings. The "Physical" condition, with consistent realism, minimized this conflict.

Key Results

  • The Physical condition (realistic visuals + realistic physics) consistently yielded the highest causal ratings.
  • The Michotte condition (abstract disks) generally received higher ratings than the Rendered condition (realistic visuals, simplified physics).
  • Increasing visual realism alone (Michotte to Rendered) led to a dip in causal ratings.
  • Adding realistic physics to realistic visuals (Rendered to Physical) significantly boosted causal ratings, often exceeding the Michotte baseline.
  • The effect of temporal gaps was consistent across conditions, with longer gaps reducing perceived causality.
  • Stereoscopic (3D) presentation did not significantly affect causal ratings compared to 2D.

Limitations and Future Work

The authors acknowledge that using rating scales to quantify subjective experience has inherent limitations, and performance-based methods might offer more precise measurements. However, they argue that rating scales were appropriate for capturing the subjective "feel" of causality and allowed for direct comparison with previous studies.

Future work could explore:

  • Creating stimuli with Michotte visuals but realistic motion dynamics to further disentangle visual and physical realism.
  • Investigating the precise mechanisms by which peripheral sensory cues influence causal perception.
  • Examining how different levels of realism interact with other factors, such as object properties and environmental context.
  • Further exploring the "uncanny valley" effect in causality perception, particularly how partial realism can lead to unease or reduced causal attribution.

Summary updated

Abstract

One of the most important tasks for humans is the attribution of causes and effects in all wakes of life. The first systematical study of visual perception of causality-often referred to as <i>phenomenal causality</i>-was done by Albert Michotte using his now well-known <i>launching events</i> paradigm. Launching events are the seeming collision and seeming transfer of movement between two objects-abstract, featureless stimuli ("objects") in Michotte's original experiments. Here, we study the relation between causal ratings for launching events in Michotte's setting and launching collisions in a photorealistically computer-rendered setting. We presented launching events with differing temporal gaps, the same launching processes with photorealistic billiard balls, as well as photorealistic billiard balls with realistic motion dynamics, that is, an initial rebound of the first ball after collision and a short sliding phase of the second ball due to momentum and friction. We found that providing the normal launching stimulus with realistic visuals led to <i>lower</i> causal ratings, but realistic visuals together with realistic motion dynamics evoked higher ratings. Two-dimensional versus three-dimensional presentation, on the other hand, did not affect phenomenal causality. We discuss our results in terms of intuitive physics as well as cue conflict.
BibTeX
@inproceedings{meding2020phenomenal,
  author = {Kristof Meding and Sebastian A. Bruijns and Bernhard Schölkopf and Philipp Berens and Felix A. Wichmann},
  title = {{Phenomenal Causality and Sensory Realism}},
  booktitle = {i-Perception},
  year = {2020},
  doi = {10.1177/2041669520927038},
}
Figure