A Field Guide to Everyday Evidence

At the kitchen table, Mina notices a wet ring beneath a cold drinking glass. She says that the glass must be leaking. Her brother Arun thinks water has appeared on its outside. Both have noticed something real: there is water on the table. They have offered different explanations for it. The difference between noticing and explaining is where this field guide begins.

Mina and Arun are fictional. Their disagreement is ordinary enough to be useful. Many claims begin as a quick interpretation of something observed: the new lamp uses less electricity; the garden has fewer birds; a product works better; this month has been unusually hot. The claim may be correct. Before deciding, we need to understand what would make it correct and what evidence could help distinguish it from an alternative.

Start by giving the question an address. What object or group does it concern? What outcome matters? Over what period? Compared with what? “Is this better?” is too open to investigate well. “Does this lamp use less electrical energy during four hours of use than the other lamp?” names a measurable comparison. It still leaves questions about brightness and actual power use, but it gives the inquiry somewhere to begin.

The address prevents a common mistake: gathering information that is impressive but unrelated to the decision. A lamp's packaging may contain many technical terms. If the question concerns the energy required to provide useful light, a decorative claim about modern design does not answer it. A true statement can still be the wrong evidence for the question being asked.

Write three lines before looking for an answer. First, describe the observation without the explanation: water was found around the glass after ten minutes. Second, state a possible explanation: water vapour in the surrounding air condensed on the cold surface. Third, state something that would make you reconsider: perhaps the same vessel loses water even when its outside is not cold. These lines separate what you saw, what you think it means and what might challenge that interpretation.

An alternative explanation is not an accusation that someone is foolish. It is a way to make the question more demanding. A plant may grow more after being moved to a window because the light changed, because watering changed, because the weather changed or because it was already entering a new stage of growth. A useful investigation asks which differences were present, rather than assigning the whole result to the most memorable change.

This does not mean every everyday choice requires a formal experiment. The scale of the inquiry should match the consequences. Choosing a notebook can tolerate uncertainty that would be unacceptable when choosing a structural material or treating an illness. Knowing when your evidence is too limited is itself a practical result. Some questions belong with qualified professionals and established procedures.

Throughout this book, a field note will mean a record that another person could understand: a question, a method, observations, limitations and a conclusion that does not reach beyond them. You can use that structure with a ruler, a chart or a news report. Its purpose is not to make you sound scientific. Its purpose is to let evidence change your mind in a traceable way.

Arun measures the width of a shelf and writes “80.” Mina asks whether he means centimetres, millimetres or inches. He adds centimetres. She then asks whether he measured the inside space, the outside frame or the usable flat surface. The number has become more informative, but only because its method has become clearer.

A measurement connects a quantity to a procedure and a unit. Before comparing two results, check that they describe the same quantity. The width of an object and the width of the space available to contain it are related but different. A measurement across a curved surface may differ from a straight-line distance. More digits do not repair a mismatch in what was measured.

NIST's measurement terminology distinguishes accuracy from repeatability. Accuracy concerns agreement with the value of the quantity being measured; repeatability concerns agreement among results obtained under the same specified conditions. The terms should not be used interchangeably, and numerical estimates of uncertainty need their own clear description. [1]

Consider an invented scale that shows 102 grams every time a suitable reference object of 100 grams is placed on it. The readings agree with one another, yet they have a consistent offset from the reference. Repetition has exposed stability; it has not demonstrated correctness. Averaging the identical readings would still give 102 grams. This is why repeated measurement and comparison with a suitable reference answer different questions.

Now imagine another instrument giving 99, 100 and 101 grams for that object. The mean is 100 grams, but the spread remains relevant. Depending on the task, a one-gram difference might be unimportant or unacceptable. We cannot decide from the numbers alone. We need the purpose, method and uncertainty associated with the comparison. These examples illustrate ideas rather than certify any real instrument.

A display can also imply more confidence than the procedure deserves. A phone application may report several decimal places, but the number of displayed digits is not proof that the underlying measurement is equally exact. The instrument's design, conditions of use and the quantity being inferred all matter. Rounding a result sensibly is often more honest than copying every digit onto a report.

When using an ordinary ruler, define the endpoints and position it consistently. Record the unit and avoid pretending to resolve a difference smaller than you can read reliably. If the edge is worn, using its physical end as zero may introduce an offset. These are modest observations, but they show why the procedure belongs beside the result rather than hidden behind it.

There is also a difference between measuring a quantity and estimating one from a model. A device may sense one signal and calculate another value from assumptions. The displayed result can be useful while still depending on those assumptions. Ask what the device observes directly and what it infers. That question becomes especially valuable when comparing products that use the same familiar word for different measurements.

For your field note, record the quantity, unit, instrument, procedure and conditions. Add what you know about the limits. “Shelf opening, measured between the inner sides with a millimetre-marked ruler” is a better record than “80.” A number becomes useful when another reader can tell what it means and how much weight to place on it.

Mina wants to know which of two paper aeroplane designs travels farther. She makes one of each, throws the first gently indoors and launches the second hard from a balcony. The second travels farther. The observation is real, but it cannot isolate the effect of the design. Several things changed together.

Keep this as a safe thought experiment, or use a clear indoor space without people, fragile objects or sharp attachments. There is no need to launch anything from a height. The useful lesson is about comparison. The design is the factor Mina wants to examine; distance is the outcome she wants to record. Paper, launch method and surroundings can also affect that outcome.

NIST's introduction to experiments distinguishes factors from responses and separates association from causality. Seeing two quantities change together is not, by itself, enough to conclude that changing one caused the other. [2] Our plane example puts that distinction into a small practical setting: the measured difference could reflect the throw rather than the fold.

A better plan begins before the launches. Use comparable paper, define the starting line and decide how distance will be measured. Alternate or randomly choose the order of the designs rather than giving one all the early attempts and the other all the late attempts. Record each attempt, including awkward ones, according to a rule chosen in advance. If a launch hits a wall, note the event instead of silently removing it because the number is inconvenient.

Repeat the comparison because a single throw can be unusual. Repeated attempts with the same two planes tell you about those planes under that procedure. They do not automatically establish how every example of each design will perform. To learn about the designs more broadly, you would need additional independently made planes and a suitable plan. Repetition and independent replication are related, but not identical.

Here is an invented practice record. Design A travels 2.0, 2.2 and 2.1 metres. Design B travels 2.1, 2.0 and 2.2 metres. Both means are 2.1 metres. Selecting A's best attempt and B's worst would create a difference that disappears when all the recorded attempts are considered. The complete small record supports no difference in the means of these particular attempts; it does not prove the designs equivalent in every setting.