Probabilistic Reversal Learning (PRL) Task – PsyFlow Version

This task implements a probabilistic reversal learning (PRL) paradigm to measure flexible decision-making and adaptation to changing reward contingencies. Participants must learn to choose the better option based on probabilistic feedback and adapt when the reward contingencies reverse. The task is built using the PsyFlow framework.

---

Task Overview

Participants are presented with two visual stimuli and must select one on each trial. One option initially yields a higher probability of winning points. However, across the task, the correct option will occasionally reverse. Participants must track feedback and flexibly adjust their choices based on observed outcomes.

This task is useful for studying reinforcement learning, flexibility, decision-making under uncertainty, and neuropsychiatric conditions like OCD, addiction, and depression.

---

Task Flow

Step	Description
Instruction	A textbox (instruction_text) presents task instructions in Chinese. Participants press the space bar to start.
Fixation	A fixation cross "+" is shown before each trial.
Cue Display	Two options appear; participants choose using the left/right arrow keys.
Feedback	Feedback is probabilistic: +10 points (win) or -10 points (loss), with a chance of reversal.
Block Break	After each block, participants are given feedback on total score and allowed to rest.
Goodbye	A final thank-you and score screen is displayed.

---

Configuration Summary

All key settings are stored in the config/config.yaml file.

Subject Info (subinfo_fields)

Participants are registered with:

• Subject ID (3-digit number from 101–999)
• Session Name
• Experimenter Name
• Gender (Male or Female)

Localized prompts are available via subinfo_mapping.

---

Window Settings (window)

• Resolution: 1920 × 1080
• Units: deg
• Fullscreen: True
• Monitor: testMonitor
• Background color: gray

---

Stimuli (stimuli)

Stimulus Name	Type	Description
fixation	text	Central fixation cross
stima, stimb	image	Left and right selectable stimuli
highlight_left, highlight_right	rect	Highlight box drawn around selected option
win_feedback	text	+10 points feedback
lose_feedback	text	-10 points feedback
no_response_feedback	text	Penalty for missing a response
block_break	text	Mid-task rest break screen
instruction_text	textbox	Detailed task instructions
good_bye	text	Final thank-you and score display

---

Timing (timing)

Phase	Duration
Fixation	0.6–0.8 seconds (randomized)
Cue Display	1.5 seconds
Feedback Display	0.8 seconds

---

Conditions (task.conditions)

Participants are randomly assigned two option orders:

• AB: Stimulus A (left) vs Stimulus B (right)
• BA: Stimulus B (left) vs Stimulus A (right)

Choices and correct sides depend dynamically on the reversal controller.

---

Probabilistic Feedback and Reversals (controller)

Parameter	Value
Initial Win Probability	0.8
Reversed Win Probability	0.9
Sliding Window Size	10 trials
Sliding Window Hits for Reversal	9

Participants must maintain high accuracy within a moving window to trigger a reversal.

• Feedback is probabilistic: even correct choices sometimes lose points.
• After sufficient correct trials, reward mappings flip.

---

Triggers (triggers)

Trigger events are sent for synchronization:

Event	Code
Experiment start	98
Experiment end	99
Block start	100
Block end	101
Fixation onset	1
Cue onset	2
Key press	3
No response	4
Win feedback onset	5
Lose feedback onset	6
No response feedback onset	7

A marker padding system (multiplying reversal count) is used to differentiate reversal stages.

---

Running the Task

python main.py