Study Overview

Task: Participants watch animated ball collisions and predict where Ball B will land after impact, then rate their confidence in their prediction. Various physics manipulations test different aspects of physics intuition:

• Egress angle: Ball B exits at modified angles (±4.5° to ±15° from natural physics)
• Temporal delay: Ball B's response is delayed (1-5 frames after collision)
• Spatial overlap: Ball collision positions are spatially misaligned (±10-20%)
• Mass/size ratio: Ball mass and size ratios vary (0.6x to 1.4x)
• Combined conditions: Egress + overlap, egress + temporal

Key Measures:

• Prediction accuracy (directional error toward physics-aligned vs physics-violating endpoint)
• Confidence ratings (0-100 scale)
• Reaction time and mouse trajectories

Hallucinatory Quotient (HQ): Individual differences measure derived from 12 subscales across three psychometric questionnaires (PDI, CAPS, SPQ). K-means clustering divides participants into High HQ vs Low HQ groups.

Sample: 121 participants (Prolific)

Project Ecosystem

This repository is the analysis component of a three-part research project:

Component	Purpose	Repository
physicsTask	Physics simulation engine — generates stimulus data	GitHub
physics_dev (this repo)	Analysis pipeline — processes behavioral data	GitHub
physics_v4	PsychoPy experiment — data collection via Pavlovia	Pavlovia GitLab

Install the simulation engine: pip install git+https://github.com/adammanoogian/physicsTask.git

Quick Start

Environment Setup

# Option 1: Conda (recommended)
conda env create -f environment.yml
conda activate physics_dev

# Option 2: Pip
pip install -r requirements.txt

Run the Full Pipeline

# All analyses (frequentist only, ~15-30 min)
python scripts/pipeline/00_run_pipeline.py

# Paper figures only (~10-20 min)
python scripts/pipeline/00_run_pipeline.py --paper-only

# Include Bayesian analyses (~5-7 hours additional)
python scripts/pipeline/00_run_pipeline.py --paper-only --bayesian

Pipeline Stages

The full pipeline runs 11 stages sequentially. Scripts use stage-based naming: S{stage}_{seq}_{name}.py

Stage	Name	Description
S00	Stimulus Generation	Generate physics simulation data using physicsTask
S01	Data Extraction & Cleaning	Extract behavioral data from raw Prolific CSVs
S02	Quality Control & HQ Scoring	HQ PCA clustering, attention checks
S03	Mouse Tracking	Trajectory metrics (AUC, velocity, direction changes)
S04	Trial Selection	Subcondition filtering, correlation analyses
S05	Frequentist GLMM	Standard + weighted GLMM, diagnostics
S06	Bayesian Models	HQ-outcome, metacognition, bivariate, latent class
S07	Subcondition Analysis	Per-subcondition breakdowns and regressions
S08	Slow Trials & Mouse Trajectories	Slow-motion analyses, path plots
S09	Learning & Baseline	Learning curves, pre/post comparisons
S10	Reporting	Publication figures, summary tables

Data

Primary Data File: data/human_v5/human_data_cleaned.csv

• Trial-level data with quality filters applied
• Includes psychometric scale scores (PDI, CAPS, SPQ)
• HQ composite and cluster assignments

HQ Scores: data/hallucinatory_quotient_scores.csv

• Authoritative source for all HQ-related columns
• Contains 12 subscales, cluster assignments, composite scores