THE LAST TRANSMISSIONStories from the Edge of Science

THE LAST TRANSMISSION / A READER’S GUIDE

Core principles of
scientific inquiry.

How does a signal become evidence? How does evidence become an explanation? These twelve principles help you follow that passage, from the first observation to the next question.

Ways to question. Room to revise.

This is a practical guide to recurring habits of scientific inquiry, rather than a universal list of twelve rules or a fixed sequence of steps. Real research moves among observations, experiments, models, criticism, and questions. Different fields use different tools, and a discovery may begin in any of those places. Read the process reference.

The examples below lead into this site’s investigations, Evidence Files, image stories, and interactive Atlas. They are explanations of published research and teaching illustrations; this site does not claim to have performed the underlying measurements.

PRINCIPLE 01 / Begin with what was measured.

Empirical evidence

An explanation earns scientific support through observations and measurements of the world. Keep the recorded signal distinct from the story used to explain it. A chemical ingredient, a change in brightness, or an orbit can be measured even when its cause remains uncertain.

Ask yourself What was actually observed, and what has been inferred from it?

FOLLOW AN EXAMPLEBennu: ingredients without a detection of life Method reference: UC Berkeley

PRINCIPLE 02 / Give the idea something to risk.

Testability and falsifiability

A useful scientific hypothesis makes predictions that evidence could challenge. Ask which result would count against it and whether that result could be investigated. A mismatch calls for scrutiny of the hypothesis, instruments, and assumptions; it does not automatically identify which one failed.

Ask yourself Which possible observation would make this explanation less convincing?

FOLLOW AN EXAMPLETabby’s Star: using color to test an explanation Method reference: Government of Canada

PRINCIPLE 03 / Find the signal before naming it.

Careful measurement and controls

Measurements depend on instruments, calibration, sampling, and analysis choices. Comparison conditions help separate the proposed effect from background changes. Where a laboratory intervention is impossible, scientists can compare independent instruments, populations, or observations and examine possible sources of bias.

Ask yourself What comparison or calibration helps rule out an instrumental or background effect?

FOLLOW AN EXAMPLEALPHA: measuring gravity while checking magnetic effects Method reference: NIH

PRINCIPLE 04 / Keep another answer in view.

Competing explanations

Several explanations can fit the same observation. Compare their predictions and look for a measurement that separates them. An appealing account gains little from being repeated if the same evidence also fits an ordinary alternative. The next useful question is often what the alternatives predict differently.

Ask yourself What else could produce this signal, and how could we distinguish the possibilities?

FOLLOW AN EXAMPLEPluto: an ocean interpretation and an alternative Method reference: Government of Canada

PRINCIPLE 05 / A pattern still needs a cause.

Correlation and causation

Two quantities changing together can suggest a causal relationship, a common cause, or a coincidence. Timing, interventions, comparison groups, and physical mechanisms can help distinguish these possibilities. Controlled experiments are powerful where feasible, but causal inquiry also uses carefully designed observational evidence.

Ask yourself What evidence connects the proposed cause to the effect beyond their appearing together?

FOLLOW AN EXAMPLEPlant responses: tracing a proposed causal pathway Method reference: UC Berkeley

PRINCIPLE 06 / Let the result keep its boundaries.

Uncertainty and limits

A reported value needs context: how it was measured, how uncertain it is, and which assumptions could shift it. A detection limit can constrain a phenomenon without showing it is absent everywhere. Distinguish uncertainty in a measurement from uncertainty in the explanation built around it.

Ask yourself How large are the uncertainties, and which claim would go beyond the data?

FOLLOW AN EXAMPLELHS 1140 b: what a missing helium signal can constrain Method reference: NIST

PRINCIPLE 07 / Let someone else check the route.

Reproducibility and replication

Using the National Academies’ terminology, computational reproducibility means obtaining consistent results from the same data and analysis; replication uses data to investigate the same question. Terms vary among fields. Replaying an animation is neither. Some historical or astronomical events are checked through independent records and methods rather than recreated.

Ask yourself Can the analysis be checked, and has another study tested the claim with additional evidence?

FOLLOW AN EXAMPLEThe Wow! Signal: a record that has not become a repeated detection Method reference: National Academies

PRINCIPLE 08 / Ask whether different routes meet.

Independent lines of evidence

Confidence can grow when methods with different weaknesses support the same explanation. Independence matters: ten articles repeating one measurement are still reporting one measurement. Compare what each line of evidence contributes and whether shared assumptions could make them agree for the wrong reason.

Ask yourself Do these sources contain separate measurements, or are they retelling the same result?

FOLLOW AN EXAMPLEEuropa: comparing magnetic, surface, and geological clues Method reference: UC Berkeley

PRINCIPLE 09 / Know what the model can answer.

Model scope and predictions

A model selects features of reality and leaves others out. Judge it by its assumptions, the observations it explains, and predictions that can be checked. A teaching illustration may clarify a mechanism while omitting the scale, complexity, or interactions needed for a research forecast.

Ask yourself Which assumptions drive this result, and where would the model stop being useful?

FOLLOW AN EXAMPLESagittarius A*: exploring an orbit within a stated teaching model Method reference: Government of Canada

PRINCIPLE 10 / Make the evidence trail inspectable.

Transparency and research integrity

Describe the methods, assumptions, analysis choices, and limitations needed to assess a result. Share data and code where ethical and lawful, acknowledge sources, and distinguish measured results from interpretation or illustration. Privacy and other legitimate restrictions should be explained rather than disguised as missing evidence.

Ask yourself Can I trace this claim to its evidence and understand the choices behind it?

FOLLOW AN EXAMPLEAn Enceladus cutaway: separating observation from artwork Method reference: National Academies

PRINCIPLE 11 / Invite the difficult question.

Critical review and open questioning

Other researchers can challenge methods, calculations, and interpretations. Peer review helps scrutinize work before publication, and criticism continues afterward. Publication alone does not establish correctness. Evaluate the substance of a criticism, the evidence behind it, and the response rather than counting supporters.

Ask yourself What is the strongest substantive criticism, and has it been answered with evidence?

FOLLOW AN EXAMPLEFaint galaxies: checking distances, tracers, and mass estimates Method reference: UC Berkeley

PRINCIPLE 12 / Allow a better answer to arrive.

Revision with additional evidence

Scientific conclusions can change when stronger observations, improved instruments, or better explanations become available. Revision need not erase everything learned before. Keep track of which part of a claim changed, why it changed, and what evidence still supports the wider picture.

Ask yourself What changed the interpretation, and which parts remain supported?

FOLLOW AN EXAMPLECassini’s archive: questions for an existing record Method reference: UC Berkeley

TAKE THE QUESTION WITH YOU

Three questions for the next story.

  1. What was measured, and what was inferred?
  2. What other explanation could fit, and what test could distinguish it?
  3. What uncertainty or assumption might change the conclusion?

Use the Evidence Files to inspect an argument, the investigations to follow its context, and the Atlas to explore a mechanism. Artwork and animations can make an idea memorable; their apparent detail does not increase the strength of the underlying evidence.

Open the Evidence Files Choose an investigation Explore the Atlas

CHECK THE METHOD

The reference trail.

Each source explains a method used in this guide. A Method Reference link highlights its matching card below. Read the description to see why it is relevant, open the source for more detail, or use the principle links to return to the guide.

Selected method reference

UC Berkeley: How science works

This overview follows research as it moves among observations, tests, discussion, and questions. It explains why science can take several routes and why conclusions can change when evidence improves. Read it for the reasoning behind this guide’s emphasis on revision.

Used in this guide for

Selected method reference

NIH: Rigor and reproducibility guidance

Written for research applicants, this guidance examines careful study design, relevant biological variables, and the quality of research materials. Those considerations help readers ask whether a reported effect could depend on how a study was conducted. Read it for a fuller account of research rigor.

Used in this guide for

Selected method reference

UC Berkeley: Correlation and causation

These teaching materials examine how interventions and comparison groups help investigate causes. They explain why a shared pattern can also arise from a common cause or a different causal relationship. Read them when a story moves from two things changing together to one causing the other.

Used in this guide for

Selected method reference

NIST: Reporting measurement uncertainty

This section of NIST Technical Note 1297 explains what should accompany a measurement and its reported uncertainty. It connects the number to the methods and uncertainty evaluation behind it. Read it to understand why a precise-looking value still needs context.

Used in this guide for

THE SIGNAL CONTINUES

Some discoveries stay with you.

Get the illustrated investigations delivered to your inbox. Then follow the questions that refuse to leave.

Subscribe on Substack