This repository contains code and preprocessed data for reproducing analyses in "What is a face neuron," Bardon et al. 2022.


# Analysis results and code

The main analyses are documented in these Jupyter notebooks:
1. `code/1_psy_analyses.ipynb`, for analyses using only psychophysics data (e.g., Figs. 2--5);
2. `code/2_neur_analyses.ipynb`, for analyses involving neuronal data (e.g., Fig. 6).

All figures (main and supplementary) are documented.

The notebooks already contain evaluation results for inspection. Output plots are saved in `plots/fig*/`.

The notebooks can also be executed. To do so, please:
- Set up [Jupyter Notebook (or JupyterLab)](https://jupyter.org/);
- Install standard data analysis libraries (e.g., pandas, scipy, statsmodels). A full list of libraries used is printed at the end of the two notebooks;
- Copy (or symlink) the directories `data/`, `plots/` under `code/` since some files/paths as required for input/output.


# Other files

- `code/utils.py` contains functions for plotting and statistical tests.
- `code/0_combine_data_batches.ipynb` documents some preprocessing steps.
- `code/*.csv` files contain support info for analysis (e.g., image categories).
- `data/*.csv` files are needed for psychophysics analysis.
- `data/evo_resps/*.h5` files are additionally needed for neuronal response-related analyses.
- `data/cached/lexvec.h5` is needed for analyses using LexVec embeddings (Fig. S2).
- `data/cached/*.npz` contain memoized analysis results for slow steps; they can be regenerated.
- `data/batch2` contains minimally processed trial-level data; `data/batch1` contains similar data, slightly more processed, from a previous iteration of the experiments.
- `plots/assets` are needed to make some plots.
- `plots/fig*/` directories are used by the analysis notebooks for output (overwriting the contained files).