Guidelines for Keeping a Lab Notebook

You must maintain a lab notebook as you conduct your research. Here are the required practices when maintaining your notebook.

Lab Notebook Practices

To make timely progress in research, manage your data, and collect meaningful results, everyone MUST document their work. This includes what materials, tools, instruments, software and analysis you use, as well as what data you collect.

Introduction and Motivation: Why are we documenting everything?

Research requires communication.

Communicate your work to the field: Science is all about reproducibility. Someone else, without any prior knowledge of your experiments, needs to be able to reproduce your findings based off of the description you produce. This requires transparency, consistency, and accountability.

Communicate your work to yourself: You need to report how you did your work in your thesis and papers, the writing of which may occur months or years after you collect the data. Whether you work on a project for days or for years, you will eventually forget the details of your work. Documenting your work as you go is critical for accurate reporting.

Communicate your work to your colleagues: Finally, you or I may need to communicate how your work is done while training new students. You may wish to train students to assist you, or I may train students to continue your work after you graduate.

Research requires repeatability and consistency.

• To compare multiple approaches, experimental variables must be kept as consistent as possible. The only way to do this is to document them and check that they are the same every time.

  • For example, you may have multiple power supplies for your robot. Sometimes you use one and sometimes you use the other. But these power supplies have different voltages (12V and 15V), meaning that you are introducing a new variable.

    • Best case: Use a labelmaker or colored electrical tape to mark which power supply you use. Mark that down in your experimental guide. Always check that you use that one.

    • Acceptable, but pretty bad case: Write down which power supply was used for each experiment. You use a different power supply each time, but at least you have recorded which is used each time, enabling you to identify differences in your data.

    • Unacceptable: You use a different power supply each time and do not record it anywhere.

• To quantify how one approach differs from another, multiple devices or test samples may need to be constructed. Devices/samples from each approach must be manufactured the same way every time.

  • For example, to test the friction in a robot joint, you make 10 robot joints of one kind and 10 robots joints of another kind. To say anything meaningful, each group of 10 must be made up of identical samples.

    • Best case: Make Lego-style instructions. Take photos at each step. Explain what must be done. Include the catalog numbers or manufacturer numbers of the parts you use and from where you ordered them. Add arrows and other callouts to the photos to connect the text to the photo. Look at the Innovation Hub’s Standard Operating Procedure (SOP) documents as an example.

    • Acceptable, but pretty bad case: Write a text-only list of steps to follow when making each device/sample. List the catalog numbers for the parts.

    • Unacceptable: Assemble each device with the parts you can find that day. Write down nothing.

Research requires record keeping.

• Experiments may take months or years to complete. You must keep track of which version of which script you used to collect and analyze your data, what version of hardware you used, what version of the control program you used, which data go with which experiment, and a video of what happened. This record keeping is for YOU.

• If you have any uncertainty about your data (e.g., the script you used to run your robot, the version of the software you used to analyze data, the sample rate of a sensor you used), then we cannot draw any conclusions. Drawing conclusions is the point of doing research, so documentation is critical.

How are we documenting everything?

Medium

Different people prefer different methods, so I do not enforce usage of a specific system. I will offer some options. But your system must:

1. Be shareable

2. Be backed up

When you leave this lab, a copy of your notebook will remain here, and you will take a copy. Therefore, it must be shareable. Options include:

1. Lab notebook with carbon-copy pages

2. Digital document such as Word/Google doc, GitHub page, etc.

3. Digital lab notebook such as SciNote, Benchling, eLabJournal, Labstep.

Your lab notebook is the stepping stone between you and your research products, e.g., published papers, thesis. Therefore, losing it is not an option, and it must be backed up. Options include:

1. Ripping out carbon-copy pages from your notebook and putting them into a 3-ring binder

2. Storing your digital document in an auto-synchronizing backup folder such as Google Drive or OneDrive

3. Ensuring your digital lab notebook has Internet access to sync to the cloud.

Motivation and Theory

What do you wish to measure? What will you actually measure? For example, you may wish to measure a part’s moment of inertia about a particular point, but this cannot be measured directly. Instead, you may actually measure the mass and length of a pendulum and its natural frequency of oscillation. How are these calculations related? Lay this out clearly in ink or with typeset equations. Define each variable. I encourage you to consider how sources of error in your actual measurement will propagate into what you wish to measure, e.g., through perturbation theory.

Materials

What do you use to build the thing you are testing? Who is the distributor who sold you the item, and what is the item’s brand and product number? What are the other parameters (e.g., for hardware, what size, pitch, and length machine screw)? For example, if you are 3D printing a leg segment then attaching a strain gauge to it, at the bare minimum, record:

• 3D printer filament(s) brand and type, with product number

• Time needed to print the leg segment

• Eiger file for printing the leg segment

• Strain gauge manufacturer and model number, Digikey catalog number

• Sandpaper grit and brand used to smooth the leg segment surface

• Adhesive used to glue strain gauge to the leg segment

• Tape used to align, raise, and lower strain gauge onto the leg segment

This information will help us make sure there are no uncontrolled variables (e.g., different leg segments’ strain gauges are glued with different types of glue). This information is necessary for someone else to replicate your work, or even for you to replicate your own work, later. This information can be used for inventory stocking and determining whether we have enough materials to finish building what you start.

Process

How do you build the thing you are testing? In what order is everything completed? What should it look like at different stages? Write step-by-step instructions for manufacturing what you are making. Dr. S should be able to walk in and build what you are building using only your instructions. It must be a numbered list, in order. It must include photos of what it should look like at every step. The photos should have arrows, circles, and other callouts to help the reader connect the text to the photos.

Tools

What tools do you use to build the thing you are testing? These should be specified by their technical characteristics, not their cosmetic characteristics. For example, specify a ¼” Allen wrench, not “the green one”. They should not be specified by their location, either, because location may change if the lab is rearranged. Photos of each tool would be helpful for specificity.

Instruments

What technical instruments are you using to collect data? This could include, for example, scales, force plates, force gauges, cameras, fiducial markers (e.g., Aruco markers, April tags), reflective motion trackers, tensile testers, rheometers. What additional fixtures or hardware did you require to utilize the instrument? Document the manufacturer and model number of any instruments. Include photos of your device within the instrument.

Conditions

What “environmental” conditions could be affecting your results? Are the lights on (which could introduce 60 Hz electrical interference)? Are other people walking around (which could introduce vibration)? What are the temperature and humidity in the lab (which could alter material properties)?

Software

What software did you use to design what you are building? What version of the software did you use? What is the name of the specific file you used to design what you are building? Once you have a version you will turn into reality (e.g., print it, machine it), save this version in a new place.

• This is the point of services such as GitHub; changes are appended, not overwritten. It has a record of every previous version, but only one is the current version when you click “clone” or “download”. Therefore, once you are ready to collect data and publish a paper, make sure the “final” version is on GitHub and there is no confusion which file on your desktop is an old version.

• At the very least, frequently copy your work and keep old versions. Keep a record of what changed in each version and what needs to be fixed. Once you publish a paper, you can stash away all the previous versions and just keep the most recent one as the version of record.

Where is that file stored on your computer and/or the cloud? Include a hyperlink to that file so anybody with this document can access it.

What software did you use to collect data (e.g., to interface with a machine)? What version of the software did you use? If it is a proprietary piece of software, what is the URL I can use to download it myself? Are there any tricks I need to know to set up the software and use it? What is the name of the specific configuration file you used to collect data? What settings were used (e.g., for a serial port, what was the baud rate, termination character, checksum)?

Did you write your own software to collect data (e.g., an Arduino sketch that sends data to Matlab over a serial port)? Which version of each IDE did you use? What is the name of the specific file you used to design what you are building? What “special tricks” do I need to know to make it work?

• Follow the advice above about how to version this software. It is critical that you know exactly which file was used to collect data for each experiment.

• Before committing to a final piece of code, it must:

  • Have no executable code that is commented out. This raises the questions: Should this be commented out? How would the output change if this were uncommented? Was it commented or uncommented when I ran that experiment? I understand that code gets commented out during debugging, but once you are confident that it works, this must be removed.

  • Be commented throughout to explain the operations being performed. This will help others understand and build on your code. You may have a clever way to solve a problem, but if others can’t make sense of it, they may remove it in favor of something they understand. If your code has a clunky part, someone may be able to replace it with faster or more accurate code based on your commented explanation of what it is meant to do.

Data

Store all experimental data with extreme care. This is all you have to show for your hard work. If you cannot interpret your data or rely on it being accurate, you have wasted time and money. It may take a few tries to figure out exactly what you need to record for a specific experiment. Before running 100 trials, check that you can use the data you save from a couple experiments to generate the types of figures you need to prove your point.

What data did you collect? What were the environmental variables, e.g., time, date, location, temperature, humidity, etc.? At what times were each sample collected? What specific software did you use to control the instrument or robot?

• If using Matlab, store your data and the metadata listed above inside a “struct”, then save it as a .mat file. Structs use dot-indexing to store different fields. For example, for a struct called “str”, you could have:

  • str.t = time (e.g., vector with 1000 elements)

  • str.data = data (3 columns, 1000 rows, one at each time in str.t)

  • str.column_headers = {‘theta 1’,’theta 2’,’theta 3’} (string that names each column)

  • str.data = 15 Feb 2025

  • str.video_filename = ‘Google Drive\Experimental Data\video1.mp4’

• If using Python, you can do something similar with a “dictionary” and .json file.

Save all your data into a folder with the date and time. Use Matlab’s datetime() to automatically give each folder a unique name.

Make a script that generates the figure with no input from you, given a list of .mat files, .json files, or however your data is stored. Define a list of files you wish to pull data from.

When you ran your experiment, which files did you run, and which version of those files? Filetypes like .m or .py are very small; just save a copy of everything in the folder with the data. In what order did you run the files? Automate this process as much as possible! If you have multiple scripts to run for your robot to operate and collect data, write just one script that executes all the other scripts or programs, then automatically saves the data.

Analysis

How did you process the raw data you collected? This could be a simple or complicated operation and it could take place as the data is collected or after the fact.

As a simple example, you may smooth the data you collect to remove noise or jitter. Is this part of your data collection protocol, or do you smooth the data after collecting it and before plotting it to make a figure? Are the data you save unsmoothed or already smoothed? If smoothed, what algorithm did you use, e.g., low-pass filter, moving mean, moving median? What were the parameters of that algorithm, e.g., time constant for a low-pass filter, number of timesteps averaged for a moving mean or moving median? All this information would be necessary to understand your data in the context in which it is stored.

As a complicated example, you may wish to know how the peaks of a periodic dataset change from cycle to cycle. Because this is such a complicated operation, your dataset may store only the original periodic signal. You should write a script that loads the datasets you specify and plots the figure you put in your paper. Ideally, the figure in the paper can be generated in its entirety by running this analysis program. This will ensure that others see precisely how you processed your raw data into the figures you presented.

Conclusion

Research is a human endeavor and it will never go perfectly smoothly. However, our job as scientists is to do our work as thoroughly and carefully as possible and strive for excellent record keeping and data collection. This document is meant to teach and guide you how to conduct your experiments, ensure they are repeatable, and help others interpret and use your data. Share suggestions for improving this document with Dr. S so we can all learn something new about the world and share it with each other!