Experimental Code for PUnCH Project
Project Details
The PUnCH (or Parametric Uncertainty in the Control Hierarchy) experiment investigates the behavioral and neural effects of parametrically manipulating uncertainty at multiple levels of task structure.
The basic task used here is a context-dependent random dot motion perceptual decision-making task.
Code Structure
The code below uses PsychoPy for the presentation of experimental stimuli. It also makes heavy use of the cregg package for experimental utilities. Design generation uses the moss package. Otherwise the standard SciPy stack should be sufficient to control the experiment.
punch.py
This script exposes the main interface for collecting data. There are several sub-modes of this script for presenting different stages of the experiment (these are detailed below).
All modes take a -subject (can be shortened to -s) argument. Several
aspects are randomized across subjects based on the subject ID, so any
interaction with a subject should use this flag. The -cbid option can be
used to seed the counterbalance with a different ID than will be associated
with the data.
Run with the -debug flag to present in a separate GUI window. Debug mode
also adds textual information about the stimulus parameters on the screen.
The punch.py interface has the following modes (usage is python punch.py <mode>):
instruct
Self-paced participant-specific instructions for the experiment.
learn
Full-coherence stimuli, blocked, with feedback. Terminates when participant reaches a set criterion of performance (in terms of accuracy).
staircase
Blocked trials that staircase the coherence values (independently for each dimension) based on response accuracy. Generates a json file in the data directory to be used for subsequent sessions. Presents response feedback.
practice
Balanced, interleaved context-switching version of the experiment. The stimulus values and context choices are generated at runtime, so this might not be perfectly balanced for short runs.
Feedback can be turned on (it is off by default) with the -feedback command
line switch. The number of trials in the run and the number of trials in between
breaks can also be set on the command line. Can also run in scanner mode.
scan
The main experiment. Reads trial schedule information produced by
generate_designs.py. Can be collected in scanner mode or computer mode,
depending on usage of the -fmri switch.
params.py
File with dictionaries containing experimental parameters.
generate_designs.py
Module for creating static schedules for the different modes of the experiment.
At the moment it is only used for the scan mode; everything else is generated
at runtime.
monitors.py
File with monitor information.
design/
Static design files from generate_designs.py execution are stored here.
calib/
Contains monitor calibration information.
data/
Experimental data are generated here