Mechanisms Underlying Directional Motion Processing and Form-Motion Integration Assessed with Visual Perceptual Learning

Rita Donato, Andrea Pavan, Giovanni Cavallin, Lamberto Ballan, Luca Betteto, Massimo Nucci, Gianluca CampanaVision 2022 · DOI

TL;DR

This study used visual perceptual learning (VPL) to investigate whether the brain processes directional motion (from random dot kinematograms, RDKs) and non-directional motion (from dynamic Glass patterns, GPs) using the same neural mechanisms. The findings suggest that VPL is highly specific to the trained stimulus, indicating that these two types of motion are processed by distinct neural pathways.

What Problem Does This Paper Address?

The human visual system is thought to be organized into distinct streams for processing "what" (form) and "where/how" (motion). While it's known that both dynamic Glass patterns (GPs) and random dot kinematograms (RDKs) activate motion-sensitive brain regions like hMT+, it's unclear whether they rely on the same underlying neural mechanisms. GPs create a sense of motion from static patterns of dots (dipoles) that change configuration, leading to a perception of non-directional motion. RDKs, on the other hand, use dots moving in a consistent direction to create a clear perception of directional motion. This paper aims to resolve this ambiguity by using visual perceptual learning (VPL) to see if learning to process one type of stimulus transfers to the other. If VPL transfers, it would suggest shared mechanisms; if it doesn't, it implies separate processing.

Methods

The researchers employed a visual perceptual learning (VPL) paradigm. VPL involves repeated practice on a specific visual task, leading to significant and often long-lasting improvements in performance.

Here's a breakdown of their approach:

  • Participants: 31 participants were divided into two groups: one trained on dynamic GPs and the other on RDKs.
  • Stimuli:
    • Dynamic Glass Patterns (GPs): Sequences of static GP frames, where dipoles are randomly arranged. Observers perceive motion even without direct dipole correspondence between frames. These stimuli are associated with non-directional motion and global form.
  • Random Dot Kinematograms (RDKs): Dots moving in a consistent direction. These stimuli are associated with clear directional motion.
  • Task: Participants performed a two-interval forced-choice (2IFC) task. In each trial, they were presented with two intervals, one containing a coherent stimulus (either a GP or an RDK) and the other containing a noisy/random version. Their task was to identify which interval contained the coherent stimulus.
  • Procedure:
    • Pre-test: Assessed baseline performance (coherence threshold) for both GPs and RDKs for all participants.
  • Training: Participants underwent eight training sessions, with one group trained exclusively on GPs and the other exclusively on RDKs. Each session involved many trials of the 2IFC task with their assigned stimulus type.
  • Post-test: Re-assessed performance on both GPs and RDKs to measure any learning and transfer.
  • Data Analysis: A Generalized Linear Mixed Model (GLMM) was used to analyze the coherence thresholds, accounting for individual differences and the experimental design. They also calculated a "Magnitude of Learning" index to quantify performance changes.

Datasets

Not discussed in the paper. The stimuli (GPs and RDKs) were generated for the experiment.

Key Insights and Ideas

  • Specificity of Perceptual Learning: The most striking finding is that VPL was highly specific to the trained stimulus. Learning to discriminate GPs did not significantly improve performance on RDKs, and vice versa. This strongly suggests that the neural mechanisms underlying the processing of these two stimulus types are largely independent.
  • Distinct Neural Pathways: The lack of transfer implies that directional motion (RDKs) and non-directional motion/form-motion integration (GPs) are likely processed by separate neural circuits, even though they might converge in higher-level visual areas.
  • Form vs. Motion Dominance: The results suggest that dynamic GPs might be more heavily reliant on global form processing, with VPL occurring at earlier visual stages. In contrast, RDKs, which clearly represent directional motion, might engage higher-level motion processing areas, leading to VPL that doesn't transfer to form-based motion cues.
  • Asymmetric Influence: While there was little transfer, the authors speculate about a potential asymmetric influence, suggesting motion might influence form more than form influences motion, though this requires further investigation.

Key Results

  • VPL Specificity: Participants trained on GPs showed significant improvements in GP discrimination but not RDK discrimination. Similarly, participants trained on RDKs showed significant improvements in RDK discrimination but not GP discrimination.
  • Magnitude of Learning: The "Magnitude of Learning" index was significantly below zero (indicating learning) only for the trained stimulus in both groups. For the untrained stimulus, the magnitude of learning was not significantly different from zero.
  • Coherence Thresholds: Generally, participants found it easier to discriminate RDKs than GPs (lower coherence thresholds for RDKs), suggesting RDKs might be a "simpler" task in this context.
  • Learning Curves: The learning curves (coherence thresholds over training sessions) could be described by a power function with similar learning rates but different asymptotic performance levels for the two groups. This indicates that while the learning process might share some characteristics, the endpoint (how good they get) differs based on the stimulus.

Limitations and Future Work

  • Stimulus Complexity: The study focused on translational GPs and RDKs. Future work could explore whether more complex configurations (e.g., circular or radial GPs and RDKs) show different patterns of learning transfer.
  • Task Difficulty: While the authors aimed to keep task difficulty consistent, they acknowledge that task difficulty can influence transfer. Further research could systematically vary difficulty to see its impact.
  • Neural Mechanisms: While the VPL results strongly suggest distinct mechanisms, direct neural evidence (e.g., fMRI or TMS studies) could further elucidate the specific brain regions and pathways involved for each stimulus type.
  • Asymmetry of Influence: The potential asymmetric influence between form and motion warrants further investigation.

This study provides compelling psychophysical evidence that our visual system employs distinct mechanisms for processing clear directional motion versus motion cues embedded within form-based stimuli, highlighting the sophisticated and specialized nature of visual perception.

Summary updated

Abstract

Dynamic Glass patterns (GPs) are visual stimuli commonly employed to study form-motion interactions. There is brain imaging evidence that non-directional motion induced by dynamic GPs and directional motion induced by random dot kinematograms (RDKs) depend on the activity of the human motion complex (hMT+). However, whether dynamic GPs and RDKs rely on the same processing mechanisms is still up for dispute. The current study uses a visual perceptual learning (VPL) paradigm to try to answer this question. Identical pre- and post-tests were given to two groups of participants, who had to discriminate random/noisy patterns from coherent form (dynamic GPs) and motion (RDKs). Subsequently, one group was trained on dynamic translational GPs, whereas the other group on RDKs. On the one hand, the generalization of learning to the non-trained stimulus would indicate that the same mechanisms are involved in the processing of both dynamic GPs and RDKs. On the other hand, learning specificity would indicate that the two stimuli are likely to be processed by separate mechanisms possibly in the same cortical network. The results showed that VPL is specific to the stimulus trained, suggesting that directional and non-directional motion may depend on different neural mechanisms.
BibTeX
@inproceedings{donato2022mechanisms,
  author = {Rita Donato and Andrea Pavan and Giovanni Cavallin and Lamberto Ballan and Luca Betteto and Massimo Nucci and Gianluca Campana},
  title = {{Mechanisms Underlying Directional Motion Processing and Form-Motion Integration Assessed with Visual Perceptual Learning}},
  booktitle = {Vision},
  year = {2022},
  doi = {10.3390/vision6020029},
}
Figure