In a second-floor lab at Harbor City Community College, a dozen students stood over clipboards instead of beakers, asked to commit their science to paper before touching a single instrument — a move organizers said is meant to curb common mistakes that can turn a class project into “a story you wished happened.”

The session, hosted by the college’s new Open Bench program, was billed as a crash course in writing a minimal experimental plan — not a thesis, not a grant proposal, but a one-page set of decisions meant to hold up when results are messy.

“We’re not trying to make this complicated,” said Dr. Lila Sato, the program’s director, as she walked participants through a template pinned to a corkboard. “We’re trying to make it difficult to fool ourselves.”

Sato said most novice projects fail in predictable ways: they don’t repeat measurements, they change procedures midstream, they decide who “counts” after seeing the data, or they quietly swap the outcome they were excited about once the original outcome looks flat.

Replication: repeats that survive a bad day

At one table, student Mateo Rivas described a recent classroom experiment testing whether a fertilizer boosted plant growth.

“I got one plant that shot up, and I thought, ‘That’s the answer,’” he said. “Then the next one didn’t. We didn’t know what to trust.”

Sato asked the group to write down, in plain language, what they meant by “repeat.” She differentiated between repeating within the same batch and repeating on different days.

“A repeat tells you if an effect is steady or if you caught a fluke,” she told them, pointing to a row of coffee cups on a side bench. “You’ll have days when the room’s warmer, your hands shake, the timing slips. Repeats show whether your finding survives those little insults.”

Sample size: planning for variability and precision

The template avoided equations, but the discussion returned repeatedly to two words printed in bold: variability and precision.

“If the measurement jumps around, you need more observations to pin down what’s typical,” said Sato, who asked participants to think about how noisy their outcome might be.

A teaching assistant, Noor Patel, held up a stack of anonymized student charts from last semester: some measurements clustered tightly; others looked like scattershot.

“Think about whether you want a quick read or a sharper estimate,” Patel said. “A small sample can show a huge difference if it’s there. But if the effect is modest and the data are bouncy, a handful of trials won’t tell you what you want to know.”

Several participants wrote notes to run a short pilot first — not as a “try until it works,” Patel said, but as a way to learn how variable their measurement is and how consistent their procedure feels.

Protocol consistency: the same steps, every time

On a whiteboard, Patel wrote: “Same person, same equipment, same timing, same settings.”

A student planning a bacterial growth project admitted she had been switching between two incubators depending on which was free.

“That’s not always fatal,” Patel said, “but write it down. If you can’t keep it constant, at least make it deliberate and consistent.”

He urged students to draft a step-by-step protocol that could be followed by a classmate.

“If your friend can’t replicate your steps from your notes, you don’t have a protocol,” Patel said.

Measurement procedures: decide how you’ll measure before you look

Across the room, participants practiced timing, counting and recording methods using mock data sheets.

Sato asked students to specify who would take measurements and how, including when to start a timer, how to position a sensor, and where to record values.

“Don’t decide after you see the numbers that you ‘meant’ the peak value, not the average,” she said. “Pick your measurement rule now.”

The group also discussed calibration and blinding in everyday terms. If a device drifted, note the check; if expectations might sway a reading, have someone who doesn’t know the condition record it.

Inclusion/exclusion criteria: who counts, and why

Sato passed around a checklist labeled “Include/Exclude,” warning participants against ad hoc decisions.

“If you’re going to toss out a sample because the sensor fell off, say so in advance,” she said. “If you’re going to toss it because you don’t like what it does to your chart, that’s a different thing.”

Students were asked to write criteria that were tied to observable events: a missed dose, a broken seal, a measurement taken outside the time window, a subject failing to follow instructions.

Rivas wrote that any plant knocked over during watering would be excluded, while a plant that grew slowly would remain included.

“That’s the line,” Sato said, scanning his page. “Exclude for a documented mishap, not for an inconvenient outcome.”

Stopping rules: when you’ll stop collecting data

A recurring theme was the temptation to keep going until a desired pattern appears.

Patel asked students to write down the planned number of observations and the planned duration, plus what would justify stopping early.

“Sometimes you stop because of a safety issue, equipment failure, or because you’ve reached your planned endpoint,” he said. “What you don’t want is ‘we stopped when the result looked good.’”

One student, working on a reaction-time project, wrote a stopping rule that ended data collection after a set number of sessions even if results were “almost significant,” a phrase Patel discouraged.

“Almost is just your brain bargaining,” Patel said.

Stay aligned to your original question

Near the end, Sato brought the room back to the first line on the template: Your question, in one sentence.

“Your question is the anchor,” she said. “When you’re tired and you’ve got a spreadsheet full of surprises, you’ll be tempted to reinterpret what you were ‘really’ testing.”

She said exploratory analyses can be valuable — as long as they are labeled as such.

“If you discover something you didn’t ask, write it down as a new question for next time,” she said. “Don’t rewrite history and pretend that was your plan all along.”

A participant, high school senior Erin Cho, said she had once changed her outcome from “average score” to “best score” after her tutoring intervention didn’t improve the class mean.

“I didn’t think of it as cheating,” Cho said. “I thought I was being flexible.”

Sato nodded.

“Flexibility is fine when you’re brainstorming,” she said. “It’s a problem when you’re making claims.”

A fully worked mini-plan handed out at the session

Organizers provided a sample one-page plan for a simple biology question, written to be copied and adapted. The example was discussed line by line, with instructors pointing out where students tend to improvise mid-experiment.

Question

Does caffeine change resting heart rate after 10 minutes compared with a caffeine-free drink?

Outcome (decide first)

Primary outcome: heart rate (beats per minute) measured at baseline and exactly 10 minutes after drink consumption.

Secondary notes (not primary claims): reported jitteriness on a simple 0–10 self-rating recorded at the same two time points.

Design overview

Within-person comparison on two days: each participant completes one caffeine condition and one caffeine-free condition, with the order assigned before the study begins.

Replication plan (repeats)

• Each participant provides two paired measurements on each day (baseline and 10-minute).
• The full two-day procedure is repeated across multiple participants and scheduled across at least two different calendar weeks to reduce dependence on a single day’s lab conditions.

Instructor note, read aloud: “If one person shows a change, it might be that person. If everyone shows it across days, you’re closer to something general.”

Sample size reasoning (conceptual)

• Start with a small pilot group to confirm timing, comfort, and whether the heart-rate readings vary widely under the same condition.
• Enroll enough participants so that the final estimate is not dominated by one or two unusually variable individuals.
• Plan recruitment with the expectation that some participants will be excluded for missed timing or protocol deviations.

Patel told the group: “The noisier your measurement, the more people you need before you can say what’s typical.”

Protocol consistency

• Same drink volumes, same cup type, same room, same chair position.
• No vigorous exercise, nicotine, or energy drinks for a defined period before each session; the same instruction sheet is read each time.
• Same timing: baseline measured after a seated rest period; the 10-minute measurement taken within a narrow time window.
• Same device and settings for all readings; device checked against a reference at the start of each session.

Measurement procedures

• Heart rate measured using a single specified method (e.g., wearable sensor or manual pulse count) chosen before sessions begin.
• One trained measurer records heart rate for all sessions when possible; if a substitute measurer is used, their name is recorded.
• Baseline reading taken after seated rest; 10-minute reading begins at the pre-specified time.
• Data recorded immediately on a standardized sheet, then entered into a spreadsheet with a second person confirming entries.

Inclusion criteria

• Adults who consent and can attend two sessions.
• Able to refrain from restricted activities and follow timing instructions.

Exclusion criteria (pre-specified)

• Missed or late 10-minute measurement outside the allowed window.
• Consumed caffeine, nicotine, or other stimulants within the restricted period before a session.
• Sensor failure or incomplete recording at either time point.

Sato’s comment: “Exclude because something broke or a rule was violated — not because the number is inconvenient.”

Stopping rules

• Stop enrollment when the planned number of completed two-day participants is reached.
• Stop a participant’s session if they report discomfort, dizziness, or other safety concerns; document the reason.
• Do not extend data collection solely because early results look weak or inconsistent.

Analysis plan (kept minimal)

• Compare each participant’s baseline-to-10-minute change in the caffeine condition with their change in the caffeine-free condition.
• Keep the primary outcome as heart rate at 10 minutes, as written above. Any additional patterns observed are written up as follow-up questions, not as the main claim.

Documentation plan

• Save the dated protocol, randomization/order list, and blank data sheets before the first session.
• Record any deviations immediately, with a brief reason.

By the time the workshop ended, students had filled in their own one-page plans, many with fewer flourishes and more checkboxes than they expected.

“Writing it down makes it feel real,” Cho said as she packed up. “It’s like future-you can’t pretend you meant something else.”