Range

This is the repository where I am storing the data and code used to generate behavioural analyses for the publication:

Two sides of the same coin: Beneficial and detrimental consequences of range adaptation in human reinforcement learning
Sophie Bavard, Aldo Rustichini, Stefano Palminteri
https://advances.sciencemag.org/content/7/14/eabe0340/

Behavioral analyses

Run the behavioral_analyses.m script to produce the data_figures.mat.
The script loads the raw matrix data.mat and extracts the variables of interest for the figures.

Generate the figures

Run the figures.m script to generate the figures.
The script loads the file data_figures.mat.

Data

All behavioral data are stored in the raw matrix data.mat. The columns are ordered as follows:

• COLUMN 1: participant number (1:800)
• COLUMN 2: phase number (0:2)
  • 0: training phase
  • 1: learning phase
  • 2: transfer phase
• COLUMN 3: trial number (1:120)
• COLUMN 4: context number (1:8)
  • 1 = 7.50 vs 2.50 learning test
  • 2 = 7.50 vs 2.50 learning test
  • 3 = 0.75 vs 0.25 learning test
  • 4 = 0.75 vs 0.25 learning test
  • 5 = 7.50 vs 0.75 transfer test
  • 6 = 2.50 vs 0.25 transfer test
  • 7 = 7.50 vs 0.25 transfer test
  • 8 = 2.50 vs 0.75 transfer test
• COLUMN 5: left symbol (1:8)
• COLUMN 6: right symbol (1:8)
• COLUMN 7: choice left/right
  • -1: left
  • 1: right
• COLUMN 8: choice accuracy
  • 0: incorrect
  • 1: correct
• COLUMN 9: outcome chosen option
  • 0: null
  • 1: positive (1pt or 10pts)
• COLUMN 10: outcome unchosen option (complete feedback only)
  • 0: null
  • 1: positive (1pt or 10pts)
• COLUMN 11: trial reaction time (ms)
• COLUMN 12: cumulated reward
• COLUMN 13: participant ID
• COLUMN 14: experiment number (1:8)
  • 1 = partial / no feedback, interleaved trials
  • 2 = partial / partial, interleaved trials
  • 3 = complete / no feedback, interleaved trials
  • 4 = complete / complete, interleaved trials
  • 5 = partial / no feedback, blocked trials
  • 6 = partial / partial, blocked trials
  • 7 = complete / no feedback, blocked trials
  • 8 = complete / complete, blocked trials
• COLUMN 15: trial order (1:2)
  • 1 = interleaved trials
  • 2 = blocked trials

Model fitting

Participants' choices were modeled using likelihood maximization. Run the Model_Fitting.m script to fit the 4 models presented in the main text (ABSOLUTE, RANGE, HABIT, UTILITY models). The script uses data.mat and function_model_fitting_simulations.m, and saves fitted variables in the dedicated file Optimization.mat.

Model simulation

Run the Model_Simulation.m script to produce generated data for each model. The script loads Optimization.mat and uses function_model_fitting_simulations.m to generate figures with data overlayed with model simulations. Optionally, the script saves the generated data in a dedicated file.

Functions

File function_model_fitting_simulations.m contains the algorithms used to fit and simulate the data for all 4 models presented in the main text. Files SurfaceCurvePlotSmooth.m, SurfaceCurvePlotSmooth2.m, skylineplot.m, skylineplot_model.m, structure_matrix_to_plotmatrix.m were created and used for visual purposes. Created by Sophie Bavard & Stefano Palminteri.