Try the work before you join
A hands-on test run for prospective graduate students and rotation students: work with a real neuroscience dataset, see how you like the process, and experience the kind of reasoning our lab uses every day.
This is not a test of how much free time you have, and it is not a puzzle with one approved answer. It is a structured way to show something a CV cannot: how you inspect evidence, make decisions, revise a claim, and communicate uncertainty.
It should also help you evaluate us. The notebooks expose you to the kinds of imperfect recordings, quantitative choices, and scientific arguments that make up everyday work in our lab. If the process is not enjoyable, that is useful information before either side makes a larger commitment.
A realistic sample of the work
Coding and data analysis are central to research in our lab.
The data and workflow in this challenge are representative of the work you would do here: write and debug code, organise and visualise data, evaluate statistical evidence, and turn results into scientific claims. These are not peripheral tasks; they are part of our everyday research.
You do not need to know everything already, but you should want to develop these skills and use them regularly. If you are not interested in writing code, analysing datasets, troubleshooting, and reasoning quantitatively, our lab is probably not the right fit for you.
Reasoning, not speed
Time spent, polish, and total notebook count are not scores. Careful partial work can be stronger than a complete analysis.
Independent setup
Installing the environment, reading documentation, and troubleshooting ordinary errors are part of the work sample.
- Core
- 4–6 hours
- Optional
- Open-ended
- Language
- Python
- Format
- Jupyter
The science
An animal sees a cue, then has to wait through a delay before it can respond. The longer the delay, the more chances there are to forget, to lapse, or to respond too early — and performance falls off accordingly.
Two neuromodulators are obvious candidates for carrying the animal through that gap. Noradrenaline tracks arousal and attention. Dopamine tracks motivation and reward. We recorded both at once, using fibre photometry, in two cortical regions, while mice did the task.
The question you will work on
Does the strength of the dopamine or noradrenaline signal on a single trial tell you whether the animal is about to get that trial right?
This is genuinely open. We have our own view, and we are curious whether you arrive at it — or talk us out of it.
What you will do
You get four recording sessions and a set of Jupyter notebooks that walk you through the analysis line by line. Each one explains the reasoning, then asks you to write a short piece of code yourself. Built-in checks tell you immediately whether your function behaves correctly, so you are never left stuck wondering.
Required
Load and inspect the data, build averaged response traces, quantify them, test a hypothesis, and analyse the behaviour. Budget roughly 4–6 hours.
Optional · not scored by volume
Try to predict single-trial outcomes, and ask one question of your own choosing. Do this only if it interests you and you have the time. Completing more material is not treated as greater commitment to the lab.
You send back your notebooks and a one-page write-up. If you advance to a conversation, we will ask you to walk through selected choices and one of your figures. We care much more about whether you understand and can revise your work than whether your code resembles ours.
What we are looking for
We use the same qualitative dimensions for every submission. The applicants who stand out are not necessarily those with the cleanest plots; they are the ones who argue thoughtfully with their own results.
- Attention to the raw data, including anomalies and limitations that affect later claims.
- Scientific reasoning that connects an analysis to the experimental design and considers alternative explanations.
- Statistical judgment about sample size, aggregation, uncertainty, effect size, and what the evidence can generalise to.
- Clear communication: explaining choices, identifying weak points, asking useful questions, and being honest about tools and unfinished work.
A submission that gets halfway and reasons carefully is stronger than one that finishes everything and believes all of it. The section of the write-up we read most closely is the one where you list what you do not trust.
How the review works
Work sample
We review your reasoning, written caveats, code choices, and whether the analysis can handle a held-back dataset.
Consistent rubric
Everyone is considered using the same dimensions. There is no bonus for speed, optional volume, or a particular coding style.
Conversation
The challenge is one part of a holistic decision. Research interests, mentoring fit, prior opportunities, and discussion all matter too.
Ground rules
Use AI assistants if you want to. We do. But you should be able to explain and defend every line, because we will go through your code with you. If a tool saved you real time, tell us where — that is useful information about how you work, not something to hide. Because the recordings are unpublished, do not upload the dataset or raw excerpts to an external AI service.
Troubleshoot before asking. Use the setup guide, read the error, consult documentation, and make a reasonable attempt to diagnose routine problems yourself. We do not provide introductory instruction in installing Python, pandas, or Jupyter. If you believe the supplied files or environment are broken, send one concise debugging report with your operating system, exact command, complete error, and what you already tried.
Partial work is welcome. Tell us what you did not get to and why. Running out of time is normal; pretending you did not is not.
Your circumstances are not a score. Prior access to Python, free time, and familiarity with neuroscience vary widely. We consider the opportunities you have had, and we separate readiness from commitment.
Accessibility accommodations
Access is separate from technical independence.
Independent environment setup and routine debugging are intentional parts of this challenge. Separately, if you need a disability-related or other reasonable accommodation to access the application process, contact us. An accommodation may change the format, timing, or access method, but it does not remove the core scientific and technical skills being evaluated. Requesting one will not count against you.
Request an accommodationUW–Madison accommodation informationHow to get started
- 01
Download Neurochallenge
Get the complete ZIP immediately. No request form, email, or permission is needed.
- 02
Unzip and set up
Everything you need is in one folder. Start with the README, create the Python environment, and open the orientation notebook.
- 03
Work through it
At your own pace. There is no deadline unless we have agreed one with you.
- 04
See whether it fits
Notice whether you enjoy the coding, troubleshooting, quantitative reasoning, and scientific interpretation. If you were invited to submit the exercise, follow the instructions in the ZIP.
The ZIP includes unpublished recordings. By downloading it, you agree to keep the data private, not share or repost it, not upload it to external AI systems, and delete your copy when you are done.
Questions people ask
Do I need to have done neuroscience before?
No. The notebooks assume no background in photometry and explain the biology as they go. You do need to be comfortable enough with Python to write a short function and make a plot.
How good does my Python need to be?
You should already be able to install and run a Python environment, open a Jupyter notebook, use pandas to group and average a table, make a plot with matplotlib, and investigate ordinary error messages. The challenge explains the neuroscience and specialised analysis, but it is not an introductory Python course.
Will the lab troubleshoot my Python environment?
Routine setup and debugging are part of the work sample, so we do not provide step-by-step help installing Python, pandas, or Jupyter. If you have followed the setup guide and believe the supplied environment or files are broken, send the exact command, complete error message, operating system, and what you already tried.
Is this required to apply?
It is optional unless we explicitly invite you to complete it as part of a later-stage selection process. We will always tell you which applies. An unsolicited application is not rejected simply because it does not include this challenge.
What if I run out of time?
Send what you have with a note about where you stopped. We would much rather see three careful notebooks than six rushed ones.
Does completing the optional work improve my chances?
Not simply because it took more time. Optional analysis can give us more to discuss, but the core review focuses on the quality of your reasoning. Extra hours are not treated as evidence that you are more serious.
Can I request an accommodation or another format?
Yes. Email amohebi@wisc.edu if you need a disability-related or other reasonable accommodation in the application process. An accommodation changes access to the exercise; it does not remove the core technical skills being evaluated. Asking will not count against you.
Will I get feedback?
Every submission is acknowledged and reviewed. Candidates who advance discuss their work with us. We aim to provide concise feedback when capacity permits, but we cannot promise a detailed code review for every submission.