diff --git a/README.md b/README.md
index be87829fc6954a175b638dabc47d2510025f615b..76e09de4f0015ff222bb151218c832b4f54367f0 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,29 @@
-# Instructions for Reproducing Results
+# Numerical experiments: A fast algorithm with error bounds for Quadrature by Expansion
+
+This repository contains the code for numerical experiments in the paper 'A fast
+algorithm with error bounds for Quadrature by Expansion,' available at
+[doi:10.1016/j.jcp.2018.05.006](https://doi.org/10.1016/j.jcp.2018.05.006) or on
+[arXiv](https://arxiv.org/abs/1801.04070).
+
+The code that is included reproduces the following parts of the paper:
+
+* Tables 3, 4, 5, and 6
+* Figures 14 and 15
+* Data presented in Section 5.2.2
+
+## Running Everything
+
+Install the [Docker image](#docker-image), and from the repository directory in
+a shell running in a container, type:
+```
+./run.sh
+```
+This script re-runs all experiments, and generates an output file
+`summary/summary.pdf` containing all the generated figures, tables, and
+data.
+
+It's also possible to have more selective control over what gets run. See
+[below](#running-experiments).
 
 ## Installation Hints
 
@@ -9,16 +34,21 @@ necessary software and the code in this repository is available.
 
 ### Manual Installation
 
-The script `install.sh` downloads and installs all necessary software in an
-isolated environment using Conda and pip.
+If you don't want to use the Docker image, you can install the software manually
+using the command:
+```
+./install.sh
+```
+This script downloads and installs everything in an isolated environment in this
+directory using Conda and pip.
 
-For producing figures and outputs, you will also need LaTeX and the
+For producing figures and outputs, you also need LaTeX and the
 [latexrun](https://github.com/aclements/latexrun) script.
 
 ### Post-Installation
 
-Before running, activate the Conda environment and instruct PyOpenCL to use
-POCL, as follows:
+Before directly running the Python scripts in this directory, activate the
+Conda environment and instruct PyOpenCL to use POCL, as follows:
 ```
 source .miniconda3/bin/activate inteq
 export PYOPENCL_TEST=portable
@@ -34,24 +64,14 @@ in `src/pytential/test`), which should also pass.
 
 ## Running Experiments
 
-To re-run all experiments and regenerate all experimental outputs, use the
-command `./run.sh`. This may take a long time. It is also possible to
-selectively run experiments, as detailed below.
-
-The code in this directory generates the following outputs:
-
-* experimental data stored in the `raw-data` directory;
-* tables and figures stored in the `out` directory;
-* a PDF file `summary/summary.pdf` which summarizes results.
-
-The script `generate-data.py` generates the files in the `raw-data`
-directory.
+The scripts `generate-data.py` and `generate-figures-and-tables.py` can be used
+to run individual experiments or groups of experiments, and postprocess
+experimental outputs, respectively. Pass the `--help` option for more
+documentation and the list of available experiments.
 
-The script `generate-figures-and-tables.py` postprocesses the output in the
-`raw-data` directory to generate figures and tables in the `out` directory.
-
-The Makefile `makefile.summary` controls the build of the summary PDF, whose
-source is `summary.tex`.
+The `raw-data` directory is written to by `generate-data.py` and holds
+experimental outputs. The `out` directory contains generated figures and tables
+and is written to by `generate-figures-and-tables.py`.
 
 To regenerate all outputs from the data that is already in the `raw-data`
 directory, run
@@ -70,4 +90,39 @@ regenerate the results for the `bvp` experiment, run
 ./generate-figures-and-tables.py -x bvp
 ```
 
-To see the list of available experiments, run `./generate-data.py --help`.
+## Contents
+
+The following files and directories in this repository are included and/or
+generated:
+
+| Name | Description |
+|----------------------------------|------------------------------------------------------------------------------------------------------------|
+| `Dockerfile` | Used for generating the Docker image |
+| `README.md` | This file |
+| `install.sh` | Installation script |
+| `generate-data.py` | Script for running experiments. Puts output in the `raw-data` directory |
+| `generate-figures-and-tables.py` | Script for postprocessing experiments and producing figures and tables. Puts output in the `out` directory |
+| `makefile.summary` | Makefile for generating the summary PDF |
+| `utils.py` | Extra code used by `generate-data.py` |
+| `summary.tex` | Source code for summary PDF |
+| `run.sh` | Script for re-running all experiments and generating all outputs |
+| `env` | Files used by the installer to set up the Conda and pip environments |
+| `.miniconda3` | Conda install directory |
+| `src` | Pip install directory |
+| `raw-data` | Holds data generated by experiments |
+| `out` | Holds generated figures and tables |
+| `summary` | Holds generated summary PDF and auxiliary files |
+
+## Citations
+
+```
+@article{gigaqbx2d,
+  title = "A fast algorithm with error bounds for {Quadrature} by {Expansion}",
+  journal = "Journal of Computational Physics",
+  volume = "374",
+  pages = "135 - 162",
+  year = "2018",
+  doi = "10.1016/j.jcp.2018.05.006",
+  author = "Matt Wala and Andreas Klöckner",
+}
+```