Two-Step Task

An adaptation of the task described by Nussenbaum, K., Scheuplein, M., Phaneuf, C., Evans, M.D., & Hartley, C.A. (2020) - Moving developmental research online: comparing in-lab and web-based studies of model-based reinforcement learning..

They collected data from 151 participants on two tasks: the two-step task, as described in Decker et al. (2016) and the Matrix Reasoning Item Bank (MaRs-IB) as described in Chierchia, Fuhrmann et al. (2019).

The adaptation of the two-step task, developed using jsPsych, can be found in the src directory.

Task Overview

This sequential decision-making task was originally described in Decker et al. (2016), and is based off of an adult task originally described in Daw et al. (2011). Participants make a series of sequential decisions to try to gain as much reward as possible. In this version, on each trial, participants first must select a spaceship, which then transports them to one of two planets where they can ask an alien for space treasure.

The jsPsych version of the task was originally coded by the Niv Lab at Princeton, adapted by the Hartley Lab at NYU for use online with children, adolescents, and adults, and adapted here by the Brain Development and Disorders Lab at Washington University in St. Louis.

Major Changes

• Complete refactor and rewrite of core experiment architecture using jsPsych v7.0
• Pavlovia integration was removed from all source code
• Gorilla platform integration
  • Packaging of images, audio and CSV files
  • Integration of Neurocog library for integration with Gorilla
• Rocket images and backgrounds updated
• Comprehensive data collection

Data Collection

All raw data and analysis code has been moved into the analysis directory. All analyses and results reported in the Nussenbaum et. al. (2020) manuscript can be reproduced by running the R scripts (for all data summary statistics and regression analyses) and MATLAB code (for the computational modeling of the two-step task data).

The experiment collects comprehensive data for each trial. The following table details all data points collected:

Category	Variable	Type	Description
Trial Layout	trialLayout	string	Structure of trial: 'training-rocket', 'training-alien', 'training-full', or 'full'
	leftKey	string	Key mapping for left choice (default: 'f')
	rightKey	string	Key mapping for right choice (default: 'j')
	commonTransition	boolean	Participant experiences a common ('true') or rare ('false') transition ('true' or 'false')
	rewardLikelihoods	number[]	Array of 4 reward probabilities for each alien
	responseWindow	number	Maximum response time allowed in milliseconds
Participant Response	levelOneChoice	0|1|2	Rocket choice: 0=timeout, 1=left, 2=right
	levelTwoChoice	0|1|2	Alien choice: 0=timeout, 1=left, 2=right
	levelOneRT	number	Reaction time for rocket choice in milliseconds
	levelTwoRT	number	Reaction time for alien choice in milliseconds
	timeout	boolean	Whether the trial timed out
Transition & Reward	transitionType	'none'|'common'|'rare'	Type of transition that occurred
	wasRewarded	boolean	Whether participant received a reward
Timing	trialStartTime	number	Timestamp when trial began (Date.now())
	trialEndTime	number	Timestamp when trial ended (Date.now())
Comprehension Questions	question.prompt	string	The comprehension question text
	question.correct	string	Correct answer ('true' or 'false')
	response	string	Participant's answer ('true' or 'false')
	correctAnswer	string	The correct answer for the question
	responseTime	number	Time to answer comprehension question
	isCorrect	boolean	Whether participant answered correctly
Fixation Trials	duration	number	Duration of fixation display in milliseconds
Counterbalancing	counterbalancing.swapMainRockets	boolean	Whether main rocket positions are swapped
	counterbalancing.swapTrainingRockets	boolean	Whether training rocket positions are swapped
	counterbalancing.swapRedAliens	boolean	Whether red planet alien positions are swapped
	counterbalancing.swapPurpleAliens	boolean	Whether purple planet alien positions are swapped
	counterbalancing.swapGreenAliens	boolean	Whether green planet alien positions are swapped
	counterbalancing.swapYellowAliens	boolean	Whether yellow planet alien positions are swapped
	counterbalancing.swapRocketPreference	boolean	Whether rocket-to-planet mapping is swapped

Configuration

Trial counts and parameters can be easily modified in src/config.ts:

export const config: ExperimentConfig = {
  debug: {
    enableDebugLogging: true, // Show extra logging messages
  },
  trainingTrials: {
    rocket: 8,           // Number of rocket-only training trials
    alien: 8,            // Number of alien-only training trials
    full: 8,             // Number of complete training trials
  },
  mainTrials: {
    blockSize: 50,       // Trials per block
    blockCount: 4,       // Number of blocks
  },
  timing: {
    fixation: 1000,      // Fixation cross duration (ms)
    choice: 3000,        // Response window duration (ms)
    reward: 1000,       // Reward display duration (ms)
    transition: 1500,   // Transition animation duration (ms)
  },
  controls: {
    left: 'f',           // Left choice key
    right: 'j',          // Right choice key
  },
  commonTransition: true,  // Participant experiences a common transition
  name: 'Two-Step Task',
  studyName: 'task_two_step',
  contact: 'henry.burgess@wustl.edu',
  counterbalancing: {
    swapMainRockets: true,
    swapTrainingRockets: false,
    swapRedAliens: true,
    swapPurpleAliens: false,
    swapGreenAliens: false,
    swapYellowAliens: false,
    swapRocketPreference: false,
  },
};

License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Issues and Feedback

For questions about the task used by Nussenbaum et. al., please contact <katenuss@nyu.edu>. Please contact Henry Burgess <henry.burgess@wustl.edu> regarding this adaptation or other code-related issues and feedback.