Transitive Inference Task

Field	Value
Name	Transitive Inference Task
Version	v0.1.0-dev
URL / Repository	TaskBeacon/T000052-transitive-inference-task
Short Description	Classic five-symbol transitive inference task with premise-pair learning, repeat-until-criterion training, and a final BD/AE test
Created By	TaskBeacon build pipeline
Date Updated	2026-04-17
PsyFlow Version	0.1.12
PsychoPy Version	2025.2.4
Modality	Behavior
Language	English

1. Task Overview

This task implements a classic transitive inference paradigm in PsychoPy/PsyFlow. Participants learn a five-symbol Hiragana hierarchy by trial and error across four premise-pair training blocks. The training blocks repeat until every premise pair reaches the accuracy criterion, and the final test block mixes premise pairs with the BD transitive probe and the AE anchor probe.

The runtime keeps the execution auditable. Trial order is deterministic from the task seed, left/right symbol placement is randomized per trial, and training feedback is limited to the learning phase. The task uses text-rendered symbols and does not require external media assets.

2. Task Flow

Block-Level Flow

Step	Description
Load Config	Read the mode-specific config and task metadata.
Collect Subject Info	Collect subject ID in human mode or inject deterministic IDs in QA/sim.
Initialize Runtime	Create the PsychoPy window, keyboard, triggers, and stimulus bank.
Show Instructions	Present the transitive-inference instructions and key mappings.
Run Training Block 1	Present premise pairs with feedback until the criterion is met.
Run Training Block 2	Present the same premise-pair learning structure with a different schedule.
Run Training Block 3	Present the compact premise-pair schedule.
Run Training Block 4	Present the randomized premise-pair schedule.
Run Final Test	Present premise, transitive, and anchor probes with no feedback.
Finish	Present the goodbye screen, send the end trigger, and quit PsychoPy.

Trial-Level Flow

Step	Description
Trial Fixation	Show a centered fixation cross for a short pre-pair interval.
Pair Display	Show the two Hiragana symbols side-by-side with a response prompt.
Pair Response	Collect Z for left or M for right within the response window.
Training Feedback	Show correctness feedback only during the training blocks.
Trial ITI	Show the fixation cross again before the next trial.

Controller Logic

Feature	Description
Condition Scheduling	build_session_plan(...) expands the config-defined training and test patterns into trial lists.
Determinism	Trial order and left/right placement are deterministic from the overall seed, block index, trial index, and block attempt.
Criterion Control	Training blocks repeat until every premise pair reaches the accuracy threshold or the repeat limit is reached.
Trial Context	Every participant-visible phase in src/run_trial.py calls set_trial_context(...) before display or response capture.
Simulation	The scripted and sampler responders exercise the response path with high-accuracy training behavior and slightly lower transitive-probe accuracy.

Other Logic

Feature	Description
Asset Strategy	All stimuli are text-rendered Hiragana glyphs and prompt text; no external media assets are required.
Output Capture	Trial data are written to mode-specific CSV files in outputs/.
Repeat Handling	Failed training blocks are replayed with the same configured schedule.

3. Configuration Summary

All settings are defined in config/config.yaml.

a. Subject Info

Field	Meaning
subject_id	Three-digit participant identifier.

b. Window Settings

Parameter	Value
window.size	[1280, 720]
window.units	pix
window.bg_color	black
window.fullscreen	false
window.screen	0

c. Stimuli

Stimulus ID	Purpose
instruction_text	Intro screen that explains the hierarchy-learning task and key mapping.
training_block_intro_text	Training block intro screen with criterion reminder.
test_block_intro_text	Final test intro screen with no-feedback reminder.
pair_prompt_text	Response prompt displayed beneath the symbol pair.
pair_left_symbol / pair_right_symbol	The two Hiragana glyphs shown on each trial.
fixation	Center fixation cross used before pair screens and during ITIs.
training_feedback_*	Training-only feedback screens for correct, incorrect, and timeout outcomes.
block_summary_text	Training-block summary with pair accuracies and repeat/advance notice.
good_bye_text	Final summary screen and exit prompt.

d. Timing

Parameter	Value
timing.fixation_duration_s	0.4 s
timing.response_window_s	3.0 s
timing.feedback_duration_s	0.8 s
timing.iti_duration_s	0.4 s

e. Triggers

Event	Code
exp_onset / exp_end	1 / 2
block_onset / block_end	10 / 11
instruction_onset	12
trial_fixation_onset	20
pair_onset	21
response_z / response_m	31 / 32
trial_timeout	33
feedback_onset	34
trial_iti_onset	35
good_bye_onset	40

f. Adaptive Controller

Parameter	Value
task.training_accuracy_threshold	0.8
task.block_repeat_limit	3
Adaptive controller	Training blocks repeat until the 80% premise-pair criterion is reached or the repeat limit is hit.

4. Methods (for academic publication)

Participants completed a transitive inference task using five Hiragana symbols arranged in a learned hierarchy. They practiced adjacent premise pairs across four training blocks and received trial-level feedback during learning. Once the premise-pair criterion was reached, they moved to a final test block that mixed premise pairs with the novel BD transitive probe and the AE anchor probe.

The executable implementation keeps the protocol auditable. Trial order, randomization, and repeat decisions are deterministic from the config seed and the observed training accuracy. The displayed stimuli are text-rendered Hiragana glyphs, which makes the task portable and easy to localize or restyle without changing the state machine.