DRDB McMaster Baby Lab logo: a hatching eaglet

Developmental Research Database · open source · MIT

Research happens in a changing world. Its infrastructure should keep the science constant.

Behind every behavioral experiment sits an operations layer: a participant pool, records of who is eligible and who has been contacted, schedules, rooms, and the people who run sessions. DRDB is a self-hosted system that turns that coordination from etiquette into enforced rules.

View on GitHub Deploy with Docker

Fig. 1 · The layers
Three nested layers: society, the DRDB operations layer, and the experiment. Three nested rings: society, containing the lab operations layer provided by DRDB, containing the experiment. SOCIETY pandemics · new technology · privacy law · how families communicate LAB OPERATIONS LAYER · DRDB the participant pool · eligibility · contacts · schedules · rooms · the people THE EXPERIMENT the task · the data · the analysis the same procedure, the same year after year
Fig. 1The operations layer absorbs outside change so the experiment stays the same.
Fig. 1 · The layers
Three stacked layers: society changes, the DRDB operations layer absorbs it, the experiment stays the same. Society changes at the top, the DRDB operations layer absorbs it in the middle, and the experiment at the bottom stays the same. SOCIETY pandemics · new technology privacy law · how families communicate CHANGES LAB OPERATIONS LAYER · DRDB the participant pool · eligibility contacts · schedules · rooms · people ABSORBS THE EXPERIMENT the same procedure, year after year
Fig. 1The operations layer absorbs outside change so the experiment stays the same.

01 · The problem

The invisible operations layer

Before a single trial is run, someone must find families who meet the inclusion criteria, confirm they have not already completed a related study, contact them without overwhelming them, book a room and an experimenter, remind them to come, and record what happened.

In most laboratories this work is carried by spreadsheets, shared drives, inboxes, and the memory of a lab coordinator. Its failures are rarely reported, but they shape the data. Each of the six questions below is a place where the layer quietly does the work — and where doing it badly changes the result.

01

Who is eligible?

Age windows close, and families who have already done a related paradigm are no longer naïve. Participants who turn out to be ineligible are excluded after testing, which costs power (Frank et al., 2017) and turns into a protocol deviation on paper.

02

Who contacts whom?

Several labs draw on one pool, so one lab's call is another lab's problem (Peeler et al., 2021). Over-contacted volunteers stop responding (Hurwitz et al., 2017), and the order of rows in a shared spreadsheet quietly decides who gets invited at all.

03

Who can see what?

Records kept on a lab machine cannot be reached from home; records shared with everyone are too open for participant data. Access and access-by-accident are the same problem viewed from two sides.

04

Who does the work?

New assistants arrive every term, and procedural fidelity depends on onboarding (Byers-Heinlein et al., 2020). Detailed records of who contacted which family, and when, are what let a lab keep track of its own work (Hurwitz et al., 2017).

05

Who shares the space?

Rooms, equipment, and parking are shared. A missed infant visit costs more than an empty hour: if the child ages out of the window before the session can be rebooked, the participant is lost entirely.

06

Who is in the pool?

Samples still over-represent easily-contacted families (Nielsen et al., 2017), and how a lab recruits is itself a methodological choice, not an administrative detail (Sugden & Moulson, 2015).


02 · The stress test

Why developmental research is the hardest case

Developmental research takes the hardest version of every one of these problems. Eligibility is defined by age windows that are often only weeks wide, so the list of who can be invited changes daily. A single family can hold several participants, each with their own history and their own rules. And families are recruited from the community rather than from a course-credit pool, so their goodwill has to survive years of repeated invitations.

Whatever manages this has to be right every day, not most days.

Fig. 2 · Eligibility moves
Age-window chart: studies own eligible age ranges on a shared axis; two children move along it over time. Five study age windows plotted from birth to sixty months, with markers for two siblings in one family and a vertical line at today. 0 MO 1 YR 2 YR 3 YR 4 YR 5 YR Lab A · eye tracking Lab B · word learning Lab C · parent report Lab B · motor play Lab C · school readiness Children 4 MO TODAY · IN SIX MONTHS 36 MO TODAY · IN SIX MONTHS both children move right at the same rate
Fig. 2Studies own eligible age ranges on a shared axis. The accented window is the one this infant enters next; the 3-year-old sibling already sits inside two of them.
Fig. 2 · Eligibility moves
Age-window chart: studies own eligible age ranges on a shared axis; two children move along it over time. Five study age windows from birth to five years, with two siblings in one family moving along the axis over six months. 0 MO 1 YR 2 YR 3 YR 4 YR 5 YR Lab A · eye tracking Lab B · word learning Lab C · parent report Lab B · motor play Lab C · school readiness CHILDREN 4 MO TODAY 10 MO IN SIX MONTHS 36 MO TODAY 42 MO IN SIX MONTHS both children move right at the same rate
Fig. 2Studies own eligible age ranges on a shared axis. The accented window is the one this infant enters next; the 3-year-old sibling already sits inside two of them.
Study editor for Lab C's school-readiness study, with an age group of 44 to 58 months and a rule requiring a completed parent-report study
Screen 1Set the age window and participation rules. Lab C's school-readiness study takes children aged 44–58 months who have already completed its parent-report study. Health criteria are set on the same screen. Fictional demo data.
Family record for a one-year-old who completed Lab A's eye-tracking study, with a note listing why each Lab C study is not offered, including face perception: participation rules not met
Screen 2See why a study is not offered. Mia completed Lab A's eye-tracking study, so Lab C's face-perception study, which excludes it, is not offered to her. The other two Lab C studies fail on age. Fictional demo data.

03 · History

Born in March 2020

The DRDB repository was created on 29 March 2020, about two weeks after the province declared an emergency. The problem that week was narrow and practical: the participant records lived on laboratory computers that nobody could reach, and the people who knew them were scattered. The first version was a web database so that a dispersed team could work from one copy of the truth.

That emergency set the shape of everything after it. In 2020 the question was can we reach families at all? By 2023 it was who gets which room, and when? By 2026 it was can a new assistant run this lab correctly on day one, and can a new lab adopt this without us? Each phase came from a real change in the world outside the lab.

Fig. 3 · How DRDB got here
DRDB history in three phases: remote recruitment, return to the lab, and full lab operations, with dated milestones. A timeline in three phases: remote recruitment, return to the lab, and a lab operations platform, with dated milestones. PHASE 1 · REMOTE RECRUITMENT PHASE 2 · RETURN TO THE LAB PHASE 3 · FULL LAB OPERATIONS MAR 2020 first commit NOV 2020 training mode JAN 2021 v1.0.0 ships FEB 2021 real-time updates JUN 2023 testing rooms APR 2026 setup wizard · v3 SEP 2026 Docker · local AI NOV 2026 participation rules stable routine use
Fig. 3Every step was a response to something that changed outside the lab. The quiet stretch is quiet because nothing had to be built: the system had simply settled into being used.
What changed outside the lab, and what DRDB grew in response
External shiftWhat it brokeWhat DRDB added
Pandemic lockdown
Mar 2020
Records on lab computers became unreachable; the team was dispersed. Web-based shared database (Mar 2020); real-time updates so remote users don't collide (Feb 2021).
Testing moves online
2020–21
Sessions move to video; families are no longer expected to be local (Sheskin et al., 2020). Zoom links for online studies (Nov 2020); time-zone support (Mar 2021).
Remote, shifting workforce
2020–21
Training can't happen side by side; new assistants arrive mid-semester. Training mode with practice families (Nov 2020); phone scripts stored per study (Jan 2021).
Return to the lab
2022–23
Shared rooms and multiple labs back in one building; double-bookings return. Testing-room management and multi-lab features (Jun 2023); lab locking and auto-release.
Campus logistics
2026
Parking and shared space create friction for families and late visits. Shared seven-day parking board across labs (Oct 2026).
How families communicate
2021–26
Email fatigue; some families are reachable only by phone. Next-contact dates and scheduled follow-up prompts; filters for families reachable only by phone.
Staff turnover over years
2026
Who did what is lost when people leave. Retired experimenters stay in the record (May 2026); personnel history with role and project dates (Aug 2026); workload analytics (Aug 2026); an optional weekly activity report for lab management (Nov 2026).
Labs beyond the founders
2026
Adopting institutions don't have the developers on hand. First-run setup wizard (Apr 2026); Docker deployment (Sep 2026); per-lab email branding (Jun 2026).
AI and privacy expectations
2025–26
Pressure to use AI tools sits against participant confidentiality. Optional AI drafting that can run on a local model, restricted to training data by default (Sep 2026).

04 · Principles

Design principles that emerged over six years

Adapt the infrastructure so the science can stay constant.

The other principles are how that one is achieved. None of them was decided in a meeting; each was forced by a season where the alternative was worse.

  • One source of truth. Every family, child, contact, and appointment lives in one place that every authorized person sees in real time.
  • Encode etiquette as rules. Lab locking, auto-release after fourteen days, next-contact dates and follow-up prompts, prerequisite and exclusion rules — coordination that used to depend on people remembering is enforced by the system.
  • Automate the routine; keep humans for judgment. Reminders, calendar invitations, and closing out past sessions are automatic. Deciding whom to invite stays with researchers.
  • Safe to learn. A training mode with practice families, kept out of statistics and out of automated reminders, so a new assistant can click anything.
  • Privacy by default. Self-hosted; the optional language model can run on a local deployment so drafting never leaves the institution's hardware.
Family record for Daniela Rivera with two children, Leo and Sofia, contact details, a completed study on the timeline and a conversation log
Screen 3One family record. Both siblings, their contact details, participation history and every conversation are in one record that every researcher sees. Fictional demo data.

05 · In practice

Following one family

A family registers at a community event with a four-month-old and a three-year-old. Both children are entered under one family and linked as siblings.

When the infant enters the window for Lab A's study at six months, she appears in a randomly ordered eligible list — randomized so that the chance of being invited doesn't depend on when a family registered. When another researcher has opened the same family in the eligible list, a warning asks the assistant to pause before booking. A week later Lab A books the visit: the family is now locked to Lab A, a calendar invitation goes to the family and both experimenters, the testing room is reserved, and a reminder with parking instructions goes out the day before.

After the visit the session is marked complete and the family is released. At twelve months the infant is eligible for Lab B's study but excluded from Lab C's, because she has already completed a related paradigm.

The same path fits an adult cohort: an older adult re-invited each year who must be excluded from paradigms they have already seen.

Scheduling screen for Lab A's eye-tracking study, with its 6-to-12-month window, a shuffled list of seven eligible children and Sofia Rivera's family record
Screen 4Find eligible participants. Choosing a study applies its age window and rules; here seven infants qualify, shown in shuffled order. Phone-only families can be hidden. Fictional demo data.
Booking dialog for Sofia with the date, study and two experimenters, and chips showing one more eligible study for Sofia and one for her brother Leo
Screen 5Plan a visit. While booking Sofia, DRDB shows one more study she could do in the same visit and one that her brother Leo is eligible for. The next steps send the confirmation email and set the next contact date. Fictional demo data.
Fig. 4 · One family's path
Lifecycle phase map from registered to completed, with an auto-released recovery state and a no-more-contact terminal state. Registered, eligible, contacted, scheduled, visited, completed, with an auto-released recovery state and a no-more-contact terminal state. INTERRUPTION & RECOVERY PRIMARY PHASES TERMINAL OUTCOMES AGE CALL BOOK VISIT DONE NO REPLY · 14 DAYS BACK IN THE POOL FAMILY ASKS TO STOP Registered Eligible Contacted Scheduled LOCKED TO LAB Visited Completed Auto-released No more contact
Fig. 4Booking a family locks it to one lab; a completed visit releases it automatically. A family that asks to stop is never contacted again.
Fig. 4 · One family's path
Lifecycle phase map from registered to completed, top to bottom, with an auto-released recovery state and a no-more-contact terminal state. Registered, eligible, contacted, scheduled, visited and completed from top to bottom, with an auto-released recovery state and a no-more-contact terminal state. PRIMARY PHASES RECOVERY TERMINAL AGE CALL BOOK VISIT DONE NO REPLY 14 DAYS BACK IN THE POOL FAMILY ASKS TO STOP Registered Eligible Contacted Scheduled LOCKED TO LAB Visited Completed Auto-released No more contact
Fig. 4Booking a family locks it to one lab; a completed visit releases it automatically. A family that asks to stop is never contacted again.

06 · Etiquette as rules

Lab courtesy, written into software

Much of what DRDB enforces was already lab etiquette — the things a good recruiter does without being told. The difference is that DRDB does not rely on anyone remembering. The left column below is how the founding lab's training materials teach it; the right column is what the software does.

The lab's practice, and the rule it became
How the lab teaches itWhat DRDB doesSource
Record every call “so other researchers don't call back with the same information.” A conversation log on each family, and a warning and disabled booking action when the same family is already open in another eligible list. recruitment
“Make the most of the participant's time” — bundle studies into one visit, and ask about eligible siblings. Additional eligible studies and eligible siblings are highlighted while booking. family info
Follow up “in a timely manner.” Every contact sets a next-contact date and feeds a follow-ups queue. recruitment
Mark completed studies so a family is “not contacted again” for that study. Completion status feeds prerequisite and exclusion rules at booking time. studies
Respect a family's wish to stop being contacted. A “No More Contact” flag that holds across every lab drawing on the pool. recruitment
Email templates “should be the same as the ones in the approved study ethics protocol.” Per-study templates, generated from the study's details and proofread before sending. lab setup
Block the times you are unavailable. Experimenter calendars are linked; invitations go to the family and the experimenters. researchers
Mark people inactive when they leave the lab. Retired personnel are removed from active lists but kept in the history. researchers

07 · Learning DRDB

How a new assistant starts

  • An admin or lab manager adds them to the lab and gives them a role.
  • They are assigned to a study as an experimenter.
  • They practise in training mode, on practice families that are kept separate from the real data and statistics.
  • They switch to working mode and start calling families.

The full walkthrough lives in the founding lab's training materials: recruitment · family information · researchers · studies · lab setup · FAQs. It is the founding lab's own, and other labs are expected to adapt it.


08 · In use

What the founding lab's numbers look like

13,310
families in the database
15,295
children, across those families
2,600+
families who have visited for a study
10+
universities running DRDB
~39
visits a month on average, since 2021

Database figures are the founding lab's, as of 2026. The university count covers all known deployments.

Analytics for Lab B's motor-play study: 18 recruited, 13 completed, a 22 percent drop-off rate, an outcome chart and weekly recruitment history
Screen 6Review study progress. Recruitment, completed sessions, drop-off and weekly recruitment history for each study. Fictional demo data.

09 · Beyond developmental research

Any lab with a standing pool has these problems

A cognitive-aging lab re-inviting the same older adults each year has time-varying eligibility. Several EEG and imaging groups sharing one registry have contact coordination and non-naïve exclusions. They just don't have the word developmental in the title.

Subject-pool systems manage course-credit participants. REDCap and Qualtrics manage data capture. DRDB manages the pool and the operations around it — eligibility, contact history, scheduling, and the people — which is the part that used to live in a spreadsheet.


10 · Under the hood

For the technical reader

Fig. 5 · What DRDB stores
Entity relationship diagram: families and children, appointments and contact, and lab relationships, joined through the appointment record. Families and children, appointments and conversations, and labs, studies and personnel, joined through the appointment record. FAMILIES & CHILDREN APPOINTMENTS & CONTACT LAB RELATIONSHIPS HAS 1 N HAS 1 N ATTENDS 1 N RUNS 1 N EMPLOYS 1 N BOOKS 1 N STAFFS 1 N ENTITY Family # id contact details languages · English % last · next contact date ENTITY Child # id → family date of birth sex · gender ENTITY Conversation # id → family date · who spoke what was agreed ENTITY Appointment # id → family · → child → study → experimenter ENTITY Lab # id · name time zone appearance · address ENTITY Study # id · → lab age groups · windows eligibility rules prerequisites · exclusions ENTITY Personnel # id · → lab role · active workload history
Fig. 5Seven of DRDB's tables, grouped the way the lab works. The appointment is where the family, the study, and the experimenter meet.
Fig. 6 · What it runs on
Architecture diagram: a browser talks to a Node.js API which writes to MariaDB, sends email and reads availability through Google services, and may use an optional language model. A browser talks to a Node.js API, which writes to MariaDB and reaches Gmail, Google Calendar, and an optional language model. SAVES WRITES EMAIL AVAILABILITY OPTIONAL Gmail REMINDER EMAIL · OAUTH Google Calendar BLOCKED AVAILABILITY Language model A LOCAL MODEL WORKS Browser VUE 3 · VUETIFY API EXPRESS · NODE.JS SCHEDULED JOBS MariaDB MYSQL-COMPATIBLE
Fig. 6The parts DRDB is built from: a web app, a database, the Google services the lab already uses, and scheduled jobs that run on their own.

The parts that get used daily: duplicate detection and merge, Excel and CSV import, per-lab email branding, Zoom links, per-lab time zones, the shared parking board, encrypted migration between deployments, Swagger API docs, and a first-run setup wizard so a new lab can start without a developer.

It deploys with Docker Compose; the next section has the steps.


11 · Get started

Running DRDB in your lab

If your lab already runs DRDB, ask your lab manager for its address and an account; there is nothing to install. To try it yourself, you need:

  • A Mac or Windows computer with Docker Desktop, or a Linux server with Docker Engine and Compose. Docker runs the web app and its database together.
  • For a lab deployment, a server your institution approves for participant data, and someone from IT for an hour or two.
  • A Google Cloud project with the Gmail and Calendar APIs enabled, if DRDB should send emails and calendar invitations. You can try everything else without it.

First run

  1. Download the repository from GitHub and open a terminal in its folder.
  2. Copy .env.example to .env and set three long random values for DB_PASS, DB_ROOT_PASS and JWT_KEY.
  3. Run docker compose up --build -d, then open http://localhost:8080.
  4. Sign in with the default administrator account given in the README. The setup wizard then asks you to set a new password, name your lab and connect Google.

Start with fictional families while you learn the system. The deployment section of the README has the full walkthrough, Google setup, backups and upgrades. For questions or bugs, open an issue on GitHub.

Cite DRDB

If DRDB supports research you publish, please cite the software. A methods paper is in preparation; this entry will point to it once it is out.

Xiao, N. (2026). DRDB: Developmental Research Database (Version 3.2.3) [Computer software]. https://github.com/naiqixiao/DRDB
@software{xiao_drdb_2026,
  author  = {Xiao, Naiqi},
  title   = {{DRDB}: Developmental Research Database},
  year    = {2026},
  version = {3.2.3},
  url     = {https://github.com/naiqixiao/DRDB}
}

12 · References

References

  1. Byers-Heinlein, K., Bergmann, C., Davies, C., Frank, M. C., Hamlin, J. K., Kline, M., Kominsky, J. F., Kosie, J. E., Lew-Williams, C., Liu, L., Mastroberardino, M., Singh, L., Waddell, C. P. G., Zettersten, M., & Soderstrom, M. (2020). Building a collaborative psychological science: Lessons learned from ManyBabies 1. Canadian Psychology, 61(4), 349–363. https://doi.org/10.1037/cap0000216
  2. Frank, M. C., Bergelson, E., Bergmann, C., Cristia, A., Floccia, C., Gervain, J., Hamlin, J. K., Hannon, E. E., Kline, M., Levelt, C., Lew-Williams, C., Nazzi, T., Panneton, R., Rabagliati, H., Soderstrom, M., Sullivan, J., Waxman, S., & Yurovsky, D. (2017). A collaborative approach to infant research: Promoting reproducibility, best practices, and theory-building. Infancy, 22(4), 421–435. https://doi.org/10.1111/infa.12182
  3. Hurwitz, L. B., Schmitt, K. L., & Olsen, M. K. (2017). Facilitating development research: Suggestions for recruiting and re-recruiting children and families. Frontiers in Psychology, 8, Article 1525. https://doi.org/10.3389/fpsyg.2017.01525
  4. Nielsen, M., Haun, D., Kärtner, J., & Legare, C. H. (2017). The persistent sampling bias in developmental psychology: A call to action. Journal of Experimental Child Psychology, 162, 31–38. https://doi.org/10.1016/j.jecp.2017.04.017
  5. Peeler, A., Miller, H., Ogungbe, O., Lewis Land, C., Martinez, L., Guerrero Vazquez, M., Carey, S., Murli, S., Singleton, M., Lacanienta, C., Gleason, K., Ford, D., & Himmelfarb, C. R. (2021). Centralized registry for COVID-19 research recruitment: Design, development, implementation, and preliminary results. Journal of Clinical and Translational Science, 5(1), e152. https://doi.org/10.1017/cts.2021.819
  6. Sheskin, M., Scott, K., Mills, C. M., Bergelson, E., Bonawitz, E., Spelke, E. S., Fei-Fei, L., Keil, F. C., Gweon, H., Tenenbaum, J. B., Jara-Ettinger, J., Adolph, K. E., Rhodes, M., Frank, M. C., Mehr, S. A., & Schulz, L. (2020). Online developmental science to foster innovation, access, and impact. Trends in Cognitive Sciences, 24(9), 675–678. https://doi.org/10.1016/j.tics.2020.06.004
  7. Sugden, N. A., & Moulson, M. C. (2015). Recruitment strategies should not be randomly selected: Empirically improving recruitment success and diversity in developmental psychology research. Frontiers in Psychology, 6, Article 523. https://doi.org/10.3389/fpsyg.2015.00523

Developed at

The McMaster Baby Lab

DRDB is developed by researchers at the McMaster Baby Lab, McMaster University, where it has run the lab's participant pool since 2020. Learn more about the lab and its work on the lab website.

Visit babylab.mcmaster.ca