Perceptual Learning Enhances Discrimination but Distorts Stimulus Appearance

Core Concepts

Perceptual learning improves discrimination accuracy but simultaneously exacerbates biases in stimulus appearance, leading to increased distortion of perceived stimulus magnitude.

Abstract

The study examined the effects of perceptual learning (PL) on both discrimination performance and stimulus appearance. Observers were trained on either a discrimination or estimation task involving near-horizontal motion directions.
Key findings:

PL improved discrimination accuracy and reduced noise coherence thresholds, regardless of the training task.
However, PL also increased repulsive biases in stimulus appearance, leading to greater overestimation of motion directions.
These effects were observed not only for the trained directions, but also transferred to nearby untrained directions.
To explain these counterintuitive findings, the authors developed a computational observer model. The model proposes that PL enhances the precision of neural representations, which interacts with an implicit categorization process and a non-uniform internal representation of motion directions. This leads to improved discrimination performance but also increased distortion of stimulus appearance near category boundaries.
The results suggest that PL may enhance distinctions between perceptual categories at the expense of veridical stimulus representation. This has implications for theories of perceptual learning, as well as applications in areas like perceptual expertise and clinical rehabilitation.

Stats

"Discrimination accuracy improved significantly between the pre-training and the post-training sessions for observers who trained (Session: F(1,12)=17.35, p=.001), and to a similar degree in both training tasks (Training Task x Session: F(1,12)<1) and across directions (Training Task X Session X Direction: F(2,24)<1)."
"Training significantly reduced coherence threshold to a similar degree for both training groups (Session: F(1,10)=22.64, p<0.001; Training Task X Session: F(1,10)<1)."
"On average, ±2° motion was estimated as ≈±4.6°, ±4° as ≈±11.8°, and ±8° as ≈±19.1°."
"Training increased overestimation away from the horizontal for both the trained and untrained directions."

Quotes

"Strikingly, training also exacerbated the estimation biases (e.g., after training, a 4° motion stimulus was estimated as 18°), while reducing the frequency of misclassified estimates."
"Overwhelmingly, PL studies use two-alternative forced choice discrimination tasks and only a handful have used estimation tasks. Most of these provided feedback74–76, and reported reduced estimation variability, but did not examine effects on estimation bias (i.e., appearance)."
"Our empirical findings and observer model show that PL-induced increases in the precision of sensory encoding can interact with implicit categorization and a non-uniform internal representation to repel percepts away from the discrimination boundary."

Key Insights Distilled From

Perceptual learning improves discrimination while distorting appearance

by Szpiro,S. F.... at www.biorxiv.org 09-12-2022

https://www.biorxiv.org/content/10.1101/2022.09.08.507104v2

Deeper Inquiries

How might the findings of this study inform the design of perceptual expertise training protocols in applied domains like radiology or aviation?

The findings of this study shed light on how perceptual learning can lead to improvements in discrimination accuracy while simultaneously exacerbating estimation biases. In applied domains like radiology or aviation, where experts need to make quick and accurate judgments based on visual information, understanding these effects is crucial. Training protocols can be designed to not only focus on improving discrimination accuracy but also to address the potential biases that may arise in the estimation of visual stimuli. By incorporating tasks that challenge both discrimination and estimation abilities, training programs can better prepare individuals to make more accurate and reliable judgments in real-world scenarios. Additionally, the model developed in this study, which combines encoding and decoding components to simulate perceptual learning effects, can be utilized to tailor training programs to the specific needs of professionals in these domains.

What are the potential implications of these findings for clinical rehabilitation approaches that aim to improve perceptual abilities?

The implications of these findings for clinical rehabilitation approaches are significant, especially for conditions that involve deficits in perceptual abilities such as amblyopia or cortical blindness. By understanding how perceptual learning can impact both discrimination accuracy and estimation biases, rehabilitation protocols can be optimized to target these specific aspects of perception. For instance, rehabilitation programs can be designed to not only enhance sensitivity to certain stimuli but also to address any distortions in the appearance of those stimuli. By incorporating tasks that challenge both discrimination and estimation capabilities, rehabilitation approaches can better target the underlying mechanisms of perceptual deficits and promote more holistic improvements in perceptual abilities. Additionally, the model developed in this study can be used to tailor rehabilitation programs to the individual needs of patients, ensuring that the training is effective and targeted towards specific perceptual challenges.

Could the mechanisms underlying the observed effects of perceptual learning on appearance also play a role in other perceptual phenomena, such as adaptation or attentional modulation of perception?

The mechanisms underlying the observed effects of perceptual learning on appearance, such as changes in the precision of sensory encoding and implicit categorization, could indeed play a role in other perceptual phenomena like adaptation or attentional modulation of perception. For example, in adaptation, where exposure to a stimulus can lead to changes in perception, similar mechanisms of sensory encoding changes and implicit decision-making processes may be at play. The warping of neural representations to prioritize certain features of the environment, as seen in this study, could also be relevant in adaptation scenarios where the brain needs to adjust to new sensory inputs. Similarly, in attentional modulation of perception, the idea of efficient coding and implicit categorization could influence how attention shapes perception and biases judgments. By understanding these shared mechanisms, researchers can gain insights into how different perceptual phenomena interact and influence each other, leading to a more comprehensive understanding of perception and learning processes.