Primary 6 advanced vocabulary for experimental design, fair tests and scientific conclusions helps a learner explain how an investigation is built before deciding what its result means. This world-facing Grade 6 scientific method vocabulary manual develops precise language for questions, hypotheses, variables, controls, measurements, trials, evidence and conclusions. It is designed for learners who can follow an experiment but need stronger English for explaining why one comparison is fair, why another is confounded, and how a result can support a claim without proving more than the investigation actually tested.

Readers searching for 6th grade science vocabulary, scientific method words, experiment vocabulary, independent and dependent variables, controlled variables, fair test vocabulary, scientific evidence and how to write a scientific conclusion often meet short lists of definitions. This collection goes further. Each term is taught through a boundary: what it means, what it does not mean, and what changes when one condition in a fictional investigation changes. The aim is language control that improves reading, experimental reasoning and evidence-based writing together.

This article deliberately does not replace eduKateSG’s broader Grade 6 science-investigation vocabulary owner. Its narrower job is advanced application: experimental design vocabulary, fair testing, variable control, reliability, uncertainty, causal reasoning and scientific conclusion writing. The examples are original teaching cases rather than reports of real pupil experiments. Primary 6, Grade 6, sixth grade and Year 6 are different school-system labels, not a claim that their curricula are identical. Use the parts that match the learner’s actual science and language work.

Primary 6 Advanced Vocabulary Collection · Volume 7. Alicia, Tricia and Kai Kai are fictional learners used to make decisions visible. All investigations, apparatus records, results, class reports and dialogue created for this volume are fictional teaching material. Nothing here reports measured tuition outcomes, and no activity should override local laboratory, school or adult safety instructions. The vocabulary is an editorial selection rather than an official prescribed list.

Choose the experimental problem before choosing the vocabulary

Start with entries 1–10 when the investigation question is vague or untestable. Use entries 11–20 when more than one factor is changing. Later groups cover procedure and measurement, fairness and reliability, results and uncertainty, causal reasoning, scientific conclusions, models and explanations, responsible investigation and scientific writing actions.

The NOAA scientific-method resource for Grades 5–8 presents science as a sequence that includes wondering, defining, reviewing, designing, experimenting, analysing and concluding. The NGSS science and engineering practices describe investigations as systematic work that clarifies what counts as data and identifies variables or parameters. These references support the broad investigation language used here; they do not prescribe this vocabulary collection.

A fair test is a language problem before it is a calculation problem

Suppose two fictional paper spinners are timed from the same release height. Spinner A has long blades and is released without a push. Spinner B has short blades but is flicked downward. If B reaches the floor sooner, the timing may be perfectly recorded while the comparison remains poorly designed. Two conditions changed: blade length and release method. The issue appears in the nouns and verbs before it appears in a graph. What was changed? What was kept the same? What was actually measured?

A useful science answer therefore names the tested relationship. “The short spinner was faster” is narrower than “Short blades make all spinners fall faster.” The first describes this comparison. The second generalises to a causal rule. To move from one to the other, the design and evidence must justify the stronger wording. Advanced vocabulary should help a learner see that shift, not merely decorate the final paragraph with terms such as variable and reliable.

The Science Buddies fair-test explanation uses the practical rule of changing one factor while keeping other relevant conditions the same. Its variables guide distinguishes independent, dependent and controlled variables. This volume adopts those distinctions while repeatedly asking a second language question: what sentence is actually justified when the investigation is complete?

Entries 1–10: turn curiosity into a question that can be examined

1. Inquiry

Inquiry is the organised process of asking questions and seeking answers through evidence, reasoning or investigation. Scientific inquiry can include observation, experimentation, modelling, comparison and analysis. It is broader than one fixed seven-step recipe. A learner may begin from an unexpected pattern, an existing explanation that needs checking or a design problem that requires testing.

Inquiry does not mean accepting every question as scientifically testable. “Which paper spinner is prettier?” can support a preference survey but not the same kind of physical investigation as “How does blade length affect descent time under these conditions?” The question defines what kind of evidence will be relevant. Good inquiry narrows curiosity enough that observations can actually address it.

Use scientific inquiry, an inquiry into and the inquiry asks whether. Alicia begins by stating what the investigation is trying to find out. For practice, take a broad question such as “Why do things fall differently?” and write one narrower version that identifies a measurable feature and outcome without pretending that a single classroom comparison will explain every possible cause.

2. Observation

An observation is information noticed or measured about what occurs. “The spinner took 1.8 seconds to reach the floor” is a quantitative observation when the timing rule and measurement are supplied. “The spinner rotated while descending” is a qualitative observation. Both describe what was recorded rather than why it happened.

An observation is not automatically an explanation. Seeing condensation on the outside of a cold container does not by itself establish where the water came from. A learner can report the appearance first, then consider explanations that fit additional evidence. Separating observation from inference protects the record from being rewritten by the preferred theory.

Use recorded observation, observe that and the observation shows only for what the observation actually supplies. Tricia writes the measured value before discussing its cause. For practice, separate three statements into direct observations and interpretations. Explain which extra evidence would be needed to turn the interpretation into a supported conclusion.

3. Inference

An inference is a conclusion reached from observations and other information rather than a detail directly recorded. If a sealed paper packet becomes heavier after being placed in a humid chamber in a fictional model, a learner might infer that material entered or was retained by the packet. The mass change is observed; the proposed process is inferred.

Inferences can be well supported without being identical to observations. They can also be revised when new evidence appears. A learner should not weaken every inference with random uncertainty words, but should know which premises make it reasonable. “The object absorbed water” is stronger than “its mass increased” because it identifies a process, so the evidence burden is also stronger.

Natural phrases include infer from the pattern, a supported inference and cannot be inferred from this result alone. Kai Kai labels the step from record to explanation. For practice, write one statement that stays at the observation level and a second that adds a plausible inference. Underline the assumption connecting them.

4. Research question

A research question states what an investigation seeks to find out. In an elementary experimental design, it often names the factor being changed, the outcome being observed or measured, and the conditions under which the comparison is made. “How does blade length affect descent time for these paper spinners released from the same height?” is more operational than “Which spinner is best?”

A research question should not hide its desired answer. “Why are longer blades obviously better?” assumes the conclusion. Nor must every useful scientific question fit one identical sentence frame. Some investigations describe patterns, compare materials or test whether a proposed relationship appears under stated conditions. The design should match the actual question rather than force every inquiry into an unsuitable cause-and-effect mould.

Use the research question asks, refine the question and the investigation addresses. Alicia checks whether the recorded outcome can answer the question as written. For practice, rewrite a question about “better insulation” so that better is replaced by a measurable outcome such as temperature change over a specified interval.

5. Testable

Testable describes a question, hypothesis or prediction that can be examined using observations or measurements under defined conditions. “Increasing the number of paper layers will reduce the temperature change over ten minutes in this model container” is testable when the materials and measurement procedure are specified. The result can support, fail to support or complicate the statement.

Testable does not mean guaranteed to be easy, ethical or suitable for a school activity. A question may in principle be testable but require equipment, conditions or procedures outside the learner’s setting. A safe classroom simulation may test a narrower model instead. The final conclusion should remain attached to what was actually tested, not to the larger question that inspired it.

Use a testable question, made testable by defining and not testable with this packet. Tricia asks what observation would count against the proposed relationship. For practice, revise a claim such as “This material is always the best” into a statement that could be examined within one clearly defined comparison.

6. Hypothesis

A hypothesis is a proposed, testable explanation or relationship used to guide investigation. In school tasks it is often written before data are collected. A useful hypothesis might state that increasing blade area will increase descent time because greater interaction with the surrounding air will slow the fall under the model’s conditions.

A hypothesis is not a fact merely because it is written confidently, and it is not a random guess that needs no reasoning. After an investigation, avoid saying the hypothesis was proved true from one small result. A result can be consistent with it, support it within the tested conditions or fail to support it. Further testing can refine the explanation.

Use formulate a hypothesis, the evidence supports the hypothesis and revise the hypothesis. Kai Kai states the proposed relationship before examining the results. For practice, write a hypothesis and then identify one possible result that would not support it. If no imaginable observation could count against the statement, it may be too vague for the intended test.

7. Prediction

A prediction states an expected observation or result under specified conditions. It can follow from a hypothesis or model. If the hypothesis says that a longer paper-spinner blade increases descent time, a prediction may be that the long-blade spinner will have a greater mean descent time than the short-blade spinner when release height and method are controlled.

Prediction and hypothesis are related but not identical. The hypothesis expresses the proposed relationship or explanation; the prediction describes what should be observed if that relationship operates in the particular test. A prediction can turn out wrong without making the whole investigation useless. The mismatch can reveal a design issue, measurement problem or need to revise the explanation.

Use predict that, the prediction follows from and the observed result differed from the prediction. Alicia writes the expected direction before seeing the data. For practice, derive two different predictions from two competing explanations. Identify the observation that would help distinguish between them rather than merely confirm both.

8. Objective

An objective is the specific purpose an investigation is intended to achieve. “To compare heat loss from identical model containers wrapped with one, two and three paper layers over ten minutes” gives a concrete task. It does not state which condition will perform best. Keeping purpose separate from expected outcome helps prevent the method from being silently adjusted to produce a preferred result.

An objective is not the same as an evaluation of success. Completing every planned trial achieves the procedural objective of collecting the specified data, but it does not guarantee that the results are clear or that the hypothesis is supported. A well-designed investigation can produce an inconclusive result and still teach something important about the question or method.

Natural phrases include the objective is to compare, meets the stated objective and outside the investigation’s objective. Tricia checks whether a proposed extra claim belongs to the planned question. For practice, write an objective for a fictional absorption test without using words such as best or superior unless the criterion for those evaluations has already been defined.

9. Operational definition

An operational definition states how a concept or variable will be recognised, measured or recorded in a particular investigation. If a paper towel’s absorbency is defined as grams of water retained after a stated dipping and draining procedure, the word absorbency now has a specific measurement rule inside that test. Another investigation might operationalise it differently.

An operational definition does not claim that its chosen measure captures every meaning of the concept. “Cooling performance” might be represented by temperature change over ten minutes, but that does not automatically include cost, durability or safety. The definition makes the comparison workable while also revealing what has been left outside the measure.

Use define operationally as, the operational definition specifies and measured using the stated rule. Kai Kai asks what counts as one completed outcome. For practice, take a vague term such as strength and give it two different possible operational definitions for two safe paper-based tests. Explain why their results should not be treated as identical measures.

10. Criterion

A criterion is a stated standard used for judging whether a condition, result or design meets a requirement. A fictional packaging test might define success as protecting a paper model from a drop with no visible tearing. A different criterion could focus on lowest mass or least material used. The winning design can change when the criterion changes.

A criterion should be established before the result is judged whenever possible. Choosing a new success rule after seeing which design performs well can make the comparison unfair. Several criteria can also conflict. A design may minimise mass while performing less well on protection. The conclusion should state which criterion produced the ranking rather than call one design universally best.

Use meets the criterion, evaluation criterion and under the stated criterion. Alicia writes the rule beside the result table. For practice, compare two fictional designs under two different criteria and explain why a change in ranking does not mean the measurements contradict one another.

Entries 11–20: identify what changes, what responds and what must stay comparable

11. Experiment

An experiment is a structured investigation in which conditions are deliberately arranged or changed so their effects can be examined. In a simple fair-test design, the investigator changes one main factor and observes or measures an outcome while keeping other relevant conditions comparable. The experiment is the whole organised test, not just the moment when a material is moved or timed.

Not every scientific investigation is an experiment. Some questions are addressed through observation, field records, models or existing data without deliberately changing a variable. Calling every science activity an experiment can hide that distinction. The method should follow the question and ethical, practical or safety constraints.

Use conduct an experiment, experimental condition and the experiment tests. Tricia identifies which condition was deliberately changed. For practice, compare a survey of naturally occurring leaf sizes with a controlled test changing light exposure in a fictional plant model. Explain why both can be investigations while only the second is experimental in this sense.

12. Investigation

An investigation is an organised attempt to answer a question using evidence. It can include an experiment, but it may also involve systematic observation, comparison, measurement, modelling or analysis of supplied records. The word is therefore broader than experiment and is useful when the learner should not assume that a factor was deliberately manipulated.

The investigation’s boundary matters. A study of five paper materials under one moisture condition cannot automatically answer how those materials behave at every temperature, thickness or duration. A useful conclusion names the question and tested conditions rather than presenting the activity as a complete investigation of everything about the material.

Natural phrases include plan an investigation, the investigation examines and within this investigation. Kai Kai writes the scope before summarising the results. For practice, rewrite “The investigation proves paper is weak” as a narrower sentence that names the particular property and test actually supplied.

13. Trial

A trial is one run or repetition of a procedure under a stated condition. If a paper spinner is released five times with the same blade length, those are five trials for that condition. The resulting five times are separate observations. Their variation can show why one run alone may provide an unstable picture of performance.

Repeated trials do not automatically create five independent experimental conditions. They repeat the same condition to examine consistency and reduce the influence of chance variation. If the release method changes between trials, the supposed repetitions may no longer be comparable. A trial count is meaningful only when the procedure defining one trial remains clear.

Use five trials per condition, repeat the trial and trial-to-trial variation. Alicia records every valid trial rather than keeping only the fastest result. For practice, explain why choosing the best of five trials answers a different question from calculating a mean across all five under the same rule.

14. Condition

A condition is a defined set of circumstances under which an observation or trial occurs. One condition might use a ten-centimetre blade; another a fifteen-centimetre blade. Other relevant features such as paper type, release height and timing method can be kept the same so the conditions differ mainly in the factor being investigated.

Condition does not always mean treatment. A baseline or comparison condition may involve no added material or the ordinary setup. A condition can also contain several stated features, so a label such as Group A is not enough to understand it. The method should tell the reader what the label represents.

Use under each condition, comparison condition and conditions were kept the same except for. Tricia expands vague group labels into meaningful descriptions. For practice, compare two trials and list every relevant condition that differs. Decide whether the design still isolates one intended factor.

15. Variable

A variable is a characteristic or condition that can take different values or categories. Blade length, release height, paper type and descent time can all be variables in the spinner example. The word itself does not say which variable is changed deliberately, which is measured as an outcome or which should be kept constant.

A variable need not be numerical. Paper type can vary by category. A variable can also stay at one chosen value during the experiment because the investigator is controlling it. Calling it controlled does not remove its capacity to vary in other circumstances; it describes how that factor is handled in this design.

Use identify the variables, value of the variable and a categorical variable. Kai Kai names the characteristic before its particular value. For practice, distinguish the variable release height from the value one metre. Explain why confusing the characteristic with one value can make a method difficult to reproduce.

16. Independent variable

The independent variable is the main factor deliberately changed between conditions in a simple controlled experiment. If blade length is changed while release height, paper type and timing method are kept comparable, blade length is the independent variable. Naming it clarifies what relationship the experiment is intended to test.

Independent does not mean unaffected by every other feature in nature. It describes the variable’s role in the design. Nor should a learner label every first column in a table independent automatically. A table can list observation dates, identifiers or categories unrelated to deliberate manipulation. Read the method before assigning experimental roles.

Use the independent variable is, levels of the independent variable and change the independent variable. Alicia checks whether only one intended factor differs. For practice, take a flawed design where blade length and paper type both change. Explain why neither change can be isolated cleanly as the sole independent-variable contrast.

17. Dependent variable

The dependent variable is the outcome measured or observed to examine how it changes across the experimental conditions. In the spinner test, descent time can serve as the dependent variable. The method needs a measurement rule: where timing starts, where it stops and which unit is recorded.

Dependent does not mean that the investigator has already proved what causes the outcome. The design aims to test a relationship. A measured difference can still be influenced by uncontrolled conditions, measurement variation or chance. The conclusion should therefore depend on the quality of the comparison as well as the numerical difference.

Use measure the dependent variable, the outcome variable and changes in the dependent variable. Tricia checks that the outcome answers the research question. For practice, compare “Which paper is strongest?” with a specific dependent variable such as maximum supported mass under a defined paper-bridge test.

18. Controlled variable

A controlled variable is a relevant factor deliberately kept the same across the conditions being compared. In the spinner test, paper type, release height and release method might be controlled while blade length changes. Controlling them reduces alternative explanations for a difference in descent time.

A controlled variable is not the same as a control group. The controlled variable is a factor kept constant; a control or baseline group is a comparison condition. Also, keeping a factor nominally the same requires an operational rule. “Same height” should mean the same defined release point, not roughly somewhere above the desk.

Natural phrases include keep the variable controlled, controlled variables include and held constant across trials. Kai Kai lists the factors whose changes could plausibly affect the outcome. For practice, explain why colour may be irrelevant in one test but important in another. Control what matters to the question and mechanism rather than every imaginable feature.

19. Control condition

A control condition is a comparison condition representing a baseline, ordinary state or absence of the tested treatment when that comparison is meaningful. In a fictional insulation test, an unwrapped but otherwise identical container can serve as a baseline against which wrapped containers are compared. The control condition helps the reader interpret how much change is associated with the added layer.

Not every experiment requires a separate untreated control group. Some designs compare several levels of one quantitative variable directly. The important question is what comparison allows the research question to be answered. Do not add the word control merely because it sounds scientific; identify the actual baseline and explain its role.

Use control condition, baseline comparison and compared with the control. Alicia distinguishes a control condition from controlled variables. For practice, write both for one simple fictional experiment: one sentence naming the baseline and another naming three factors kept the same across all conditions.

20. Confounding variable

A confounding variable is an additional factor that differs in a way that can provide an alternative explanation for the observed outcome. If longer blades are always made from thicker paper while shorter blades use thinner paper, blade length and paper thickness are tangled together. A descent-time difference cannot be attributed cleanly to blade length alone.

Finding a confound does not tell the learner which factor actually caused the result. It identifies why the current design cannot separate them. Repeating the same flawed comparison many times can produce a consistent difference without repairing that causal ambiguity. The redesign should make the competing factors comparable or vary them in a way that allows their effects to be distinguished.

Use a potential confounding variable, the comparison is confounded by and an alternative explanation. Tricia names the specific factor rather than calling the whole experiment bad. For practice, repair a two-factor design so that one intended contrast remains. Then state which earlier causal claim must be withdrawn until the cleaner comparison is tested.

Entries 21–30: make the method reproducible enough to understand

A result cannot be separated from the procedure that produced it. “We timed the spinner” leaves open the release height, timing start, timing end and whether the same timer was used. The language below helps a learner state those choices. This is not about making every method paragraph long. It is about preserving the details another reader needs to understand what was compared.

21. Procedure

A procedure is the ordered method used to carry out an investigation. It tells the reader how conditions are prepared, what is changed, what is measured and how a trial is completed. A useful procedure is specific enough that another person can understand the intended comparison without depending on unspoken classroom knowledge.

A procedure is not automatically good because it contains many numbered steps. It can be detailed and still change two relevant factors at once. Conversely, a short procedure can be adequate for a simple task when all important conditions are defined elsewhere. Evaluate the logic of the comparison as well as the formatting.

Use follow the procedure, revise the procedure and the procedure specifies. Alicia checks whether the method makes the independent and dependent variables clear. For practice, add the missing information to “Drop each spinner and time it” until another reader can identify the release point, release method, timing rule and number of trials.

22. Protocol

A protocol is a defined set of procedures or rules for carrying out and recording a task consistently. In an advanced classroom investigation, a timing protocol might specify the same release cue, device placement and invalid-trial rule for every condition. The word emphasises standardisation across people or repeated measurements.

Protocol is broader than one instruction. “Use the stopwatch” names an instrument, while a protocol can state who operates it and how timing begins and ends. A protocol also does not guarantee that the measurement is valid for the intended concept. Consistency can reproduce the same wrong method if the rule itself is unsuitable.

Use measurement protocol, follow the same protocol and the protocol defines. Tricia compares two groups’ timing rules before combining their data. For practice, explain why results collected under different stop rules should not automatically be merged as though one second meant the same event boundary in both groups.

23. Apparatus

Apparatus means the equipment or set of materials used to carry out an experiment. In a safe paper-based timing investigation, the apparatus might include paper templates, clips, a measuring ruler and a timing device. Naming the apparatus helps another reader understand what made the measurement possible.

Apparatus does not tell the reader how the equipment was used. The same ruler can measure blade length from different reference points. A timer can be started at different moments. The method therefore needs both the equipment and the operational rule. A more sophisticated instrument cannot compensate for an unclear procedure.

Use the apparatus includes, set up the apparatus and apparatus limitation. Kai Kai distinguishes the object being tested from the equipment used to measure it. For practice, list the minimum apparatus needed for a fictional paper-bridge load comparison and identify which details belong in the procedure instead of the equipment list.

24. Instrument

An instrument is a tool used to measure or detect a quantity. A ruler can measure length; a digital balance can measure mass; a thermometer can measure temperature. The instrument gives a reading under a measurement procedure. The reading’s usefulness depends on the instrument, its scale, the operator and the definition of the quantity being measured.

Instrument does not mean that a reading is perfectly accurate. A device can have limited resolution or an offset. Nor does a number displayed to two decimal places prove that both digits are justified. Read the instrument’s meaningful increments and the task’s required precision rather than treating all visible digits as equally trustworthy.

Natural phrases include measurement instrument, instrument reading and instrument resolution. Alicia records which instrument produced each quantity. For practice, compare measuring a paper strip with a ruler marked every millimetre and one marked every centimetre. Explain what differs before any arithmetic is done.

25. Calibrate

To calibrate is to check or adjust a measuring instrument against a known reference or defined standard so its readings can be interpreted appropriately. In an elementary classroom discussion, calibration might mean checking whether a balance reads zero when empty according to its instructions or whether a ruler’s zero mark is intact.

Calibration is not the same as setting every result to the value the investigator expects. It is a reference-based check performed independently of the desired outcome. A device can also be calibrated yet unsuitable for a particular task because its measurement range or resolution is inappropriate. Calibration addresses one aspect of measurement quality.

Use calibrate the instrument, calibration check and requires recalibration. Tricia records the check before collecting the main measurements. For practice, explain why finding a consistent two-gram offset after an experiment raises a different problem from seeing readings vary unpredictably by several grams from trial to trial.

26. Unit

A unit is the agreed quantity used to express a measurement, such as centimetres for length, grams for mass, seconds for time or degrees Celsius for temperature. A value without its unit can be impossible to interpret correctly. “The strip measured 12” does not tell the reader whether twelve means centimetres, millimetres or something else.

Units also protect comparisons. Two lengths expressed in different units must be converted before their numerical values are compared directly. A time measured in seconds should not be averaged with one accidentally recorded in milliseconds as though the numbers are on the same scale. Unit errors can survive correct arithmetic and still produce a meaningless result.

Use measurement unit, convert to the same unit and record the unit beside the value. Kai Kai writes units on every data-table heading rather than in only the first row. For practice, repair a table whose mass column contains 5, 6000 and 7 when the middle value was entered in milligrams while the others were grams.

27. Measurement

A measurement is a value obtained for a quantity using a stated method and unit. A descent time of 1.82 seconds is one measurement when it comes from a defined start and stop rule. Measurements can be repeated to study variation, compared across conditions or combined in a summary when the underlying procedures are compatible.

Measurement differs from counting in some contexts, although both produce quantitative data. It also differs from the true quantity the investigator is trying to estimate. Every real measurement has finite resolution and may include error. The goal is not to pretend uncertainty disappears, but to make the method suitable enough for the comparison being made.

Use take a measurement, measurement error and repeat the measurement. Alicia writes the method beside an unusual value before deciding whether it should be excluded. For practice, explain why a surprising measurement should first be checked against the procedure and source record rather than deleted because it spoils the expected pattern.

28. Interval

An interval is the distance or time between specified values or observations. An investigation might record temperature every two minutes or test blade lengths at two-centimetre intervals. The interval affects what changes can be observed. Large gaps can hide short-lived behaviour that a finer schedule would reveal.

Equal intervals do not guarantee equal change in the measured outcome. Recording every two minutes standardises when measurements are taken; it does not imply the temperature must fall by the same amount during every interval. Likewise, a one-centimetre difference in blade length is not automatically a one-second difference in descent time.

Use at regular intervals, measurement interval and interval between trials. Tricia checks whether the time points are comparable across conditions. For practice, compare data taken at 0, 2, 4 and 6 minutes with another group measuring at 0, 3 and 6. Identify which direct time-point comparisons are possible without interpolation.

29. Duration

Duration is the length of time for which an event, condition or procedure continues. A test may expose each paper sample to the same fictional moisture condition for five minutes before measurement. Keeping duration consistent can be important when the outcome changes with time.

Duration is not the same as measurement interval. A ten-minute experiment can contain measurements every minute or only at the start and end. It is also different from a response time such as how long a spinner takes to fall. The sentence should make clear whether time describes the overall procedure, the spacing of observations or the outcome itself.

Use test duration, for a duration of and equal exposure duration. Kai Kai checks whether one sample remained in the condition longer than another. For practice, explain how unequal duration could become a confounding factor in a comparison intended to test material type.

30. Standardise

To standardise is to make a procedure, condition or recording rule consistent across relevant trials or groups. Using the same release method, measuring from the same reference point and applying the same invalid-trial rule are forms of standardisation. The goal is comparability, not making every physical object identical.

Standardisation can improve consistency while still leaving a design question unresolved. If every group uses the same inaccurate ruler, the method is standardised but may still contain systematic measurement error. If the chosen procedure does not represent the concept being studied, repeating it faithfully does not establish validity. Consistency is one requirement, not the whole quality of an investigation.

Use standardise the method, standardised conditions and standardised recording rule. Alicia asks which parts must remain comparable. For practice, design a short standardisation checklist for three groups timing the same paper-spinner condition without specifying what numerical result they should obtain.

Entries 31–40: distinguish repeating a method from improving the method

31. Repeat

To repeat is to carry out the same measurement or trial again under the intended same condition. Repetition provides more observations and can reveal trial-to-trial variation. If five timings differ slightly, their spread tells the learner something that one timing cannot show.

Repeating does not correct a confounded design. Ten repetitions of a comparison in which blade length and paper thickness both change still leave those factors entangled. Repetition addresses variation within the method; redesign addresses whether the comparison isolates the intended question.

Use repeat the trial, repeated measurements and repeat under the same condition. Tricia asks what remains unchanged across repetitions. For practice, compare a repeated timing with a changed-condition timing and explain why only the first contributes another observation to the same condition.

32. Replicate

To replicate is to carry out an investigation again to see whether a similar finding appears under the stated method and conditions, often with a fresh set of materials, people or occasions. In advanced school discussion, replication is broader than simply pressing the timer again on the same object.

A replication can differ in small practical details while preserving the essential design, but those differences should be recorded. If the second group changes several core conditions, the work becomes a related new investigation rather than a direct replication. Agreement between replications can strengthen confidence without making the conclusion universal.

Use replicate the investigation, replication attempt and replicated under comparable conditions. Kai Kai distinguishes five repeated trials in one session from another class repeating the whole method later. For practice, state which parts of a spinner investigation must be preserved for the later work to count as a meaningful replication of the original question.

33. Repeatability

Repeatability concerns how closely results agree when measurements are repeated under the same or very similar conditions. A timer obtaining 1.81, 1.79, 1.82 and 1.80 seconds under a controlled model shows close agreement in those readings. The term focuses on consistency within the repeated setup.

High repeatability does not guarantee accuracy. A timing system that consistently starts late can produce tightly grouped but biased values. Nor does repeatability prove that the experimental conclusion is valid. The method could repeatedly measure the wrong outcome for the research question.

Use good repeatability, repeatability of the measurement and low trial-to-trial variation. Alicia separates consistency from correctness. For practice, create two fictional sets: one tightly grouped around the wrong reference value and one more variable around the correct value. Explain which appears more repeatable and which is more accurate.

34. Reproducibility

Reproducibility concerns whether a result or pattern can be obtained again when the investigation is carried out independently under appropriately comparable conditions. At this level, use the term as an advanced extension: another group following the documented method with fresh materials obtains a similar ordering of spinner conditions.

Reproducibility is not identical to repeating the same reading. Independent reproduction can reveal hidden details that the original procedure failed to state. If only the first group knows an unrecorded release trick, another group may fail to reproduce the result. The failure can therefore teach the writers that the method needs clarification.

Use reproducible result, reproducibility across groups and the method was not sufficiently documented for reproduction. Tricia asks whether another reader has enough information to try the comparison. For practice, remove one critical procedure detail and predict which part of the result might become difficult to reproduce.

35. Reliability

Reliability concerns the consistency and dependability of measurements or findings under stated conditions. Repeated measurements that vary wildly can indicate low measurement reliability for the present method. Increasing the number of carefully standardised trials may help the investigator judge whether an apparent difference is stable.

Reliable does not mean true in every sense. A method can reliably produce the same biased result. A reliable measure can also be valid for one quantity but not another. Avoid using reliable as a general compliment that replaces the specific evidence about repeatability, calibration, design and scope.

Use measurement reliability, a reliable pattern across repeated trials and the reliability is limited by. Kai Kai points to the repeated values that justify the description. For practice, rewrite “The experiment was reliable” as a more specific claim naming what was repeated and how consistent the relevant results were.

36. Accuracy

Accuracy concerns how close a measurement or statement is to an accepted, checked or intended reference. A ruler reading can be accurate when the correct endpoints and units are used and the instrument is suitable. A copied value can be accurately transcribed even if the original measurement has another limitation.

Accuracy is not the same as repeatability or precision. Several readings can agree closely with each other yet all be offset from the reference. A method can also report a highly accurate length while the experiment’s causal interpretation remains invalid. State which layer of the investigation is being evaluated.

Use measurement accuracy, accurate against the reference and accuracy of the recorded value. Alicia checks both the reading and the label. For practice, explain why writing 10.000 cm does not make a measurement more accurate when the instrument only distinguishes millimetres.

37. Precision

Precision can refer to the fineness of numerical expression or to close agreement among repeated measurements, depending on context. In this volume, the sentence should make the intended sense clear. Reporting a value to the nearest millimetre concerns resolution and expression; obtaining closely grouped repeated values concerns consistency.

More digits do not create more precision than the measurement process supports. Nor does precise agreement prove accuracy. A group can repeatedly use the wrong zero point and obtain consistent values. Precision should therefore be discussed alongside the instrument, procedure and reference rather than praised in isolation.

Use reported precision, closely grouped readings and precision limited by the instrument. Tricia matches the number of reported decimal places to the justified measurement resolution. For practice, revise a table that shows ruler measurements to six decimal places and explain why the edit improves honesty rather than reduces the quality of the data.

38. Validity

Validity concerns whether the investigation or measure supports the interpretation it is being used to make. A fair comparison of insulation conditions can be valid for the question of temperature change under the specified model. It may not be valid for a broad claim about comfort, cost or real-building performance that the setup never measured.

Validity is therefore about fit between question, method, evidence and conclusion. Repeating a poorly targeted measure more often can increase the amount of data without making it a better measure of the intended concept. A valid conclusion can also be narrow. Scope is part of strength, not an embarrassment to hide.

Use valid for this comparison, threat to validity and the conclusion exceeds the study’s validity. Kai Kai asks whether the measurement actually represents the word used in the conclusion. For practice, operationalise “stronger paper” in two different ways and explain why one test cannot automatically substitute for the other.

39. Fair test

A fair test is a comparison designed so that the main intended difference is isolated while other relevant conditions are kept comparable. In a simple classroom experiment, this often means changing one independent variable at a time and controlling factors that could influence the dependent variable. Fairness here is about design logic, not moral fairness between people.

A fair test does not guarantee a particular result. It can produce no clear difference. Nor does keeping every visible object the same necessarily create fairness if an important hidden condition changes, such as one sample receiving more exposure time. The learner must understand which variables matter to the outcome.

Use design a fair test, fair comparison and the test is unfair because. Alicia names the second changing factor when identifying a flaw. For practice, improve a fictional two-material absorption test in which one sample is twice as large as the other and explain why equal sample size or area matters to the intended comparison.

40. Systematic

Systematic means organised according to a consistent plan or method. A systematic investigation records the same defined quantities for every condition, labels trials, preserves units and applies the same inclusion rules. This organisation makes checking easier because another reader can see what happened at each stage.

Systematic does not mean free of error. A systematic method can contain a systematic error that shifts every reading in the same direction. The shared word root should not hide the difference between orderly procedure and a consistent measurement bias. Context determines whether systematic is a compliment to organisation or part of an error label.

Use systematic procedure, systematic recording and later systematic error with its separate technical meaning. Tricia keeps a trial log in order. For practice, explain why orderly record-keeping makes a later mistake easier to diagnose even though organisation alone cannot prevent every experimental flaw.

Entries 41–50: describe the evidence before explaining it

41. Qualitative

Qualitative data describe qualities or categories rather than measuring a quantity on a numerical scale. A paper sample might be recorded as torn, creased or intact after a defined model test. Those categories can be meaningful when their criteria are stated clearly enough that different observers can apply them consistently.

Qualitative does not mean vague or unscientific. It can be collected systematically and coded carefully. Nor does adding numbers to category labels automatically turn them into genuine quantitative measurements. If intact is coded 1 and torn 2 for convenience, averaging the codes may not represent a meaningful halfway condition.

Use qualitative observation, qualitative category and qualitative evidence. Alicia defines the categories before the test. For practice, improve “The paper looked bad” by creating observable category criteria that another person could use without knowing which material the writer hopes will perform best.

42. Quantitative

Quantitative data express a measured or counted quantity numerically, such as mass in grams, time in seconds or number of completed folds. The number must remain attached to its unit and measurement rule. A quantitative value is not automatically more informative than a qualitative observation if it measures the wrong thing.

A number can also be approximate, rounded or uncertain. Writing a numerical result does not make the evidence exact. The instrument’s resolution and the procedure’s consistency still matter. A strong report uses quantitative detail where it answers the question and qualitative description where that captures a relevant feature the number does not.

Use quantitative measurement, quantitative data and quantify the change. Kai Kai pairs every value with its unit. For practice, turn “Sample A absorbed more” into a quantitative statement using a fictional before-and-after mass difference, then state what that number still does not reveal about the absorption process.

43. Raw data

Raw data are the original recorded observations before they are summarised into means, percentages, graphs or selected examples. Five individual descent times are raw measurements; their mean is a derived summary. Keeping the original values allows the learner to inspect variation and check a calculation.

Raw does not mean untouched by every process. An instrument may already convert a physical event into a numerical reading. The practical distinction here is between the recorded observations and later transformations applied for analysis. A report should not discard inconvenient values without documenting the rule and reason.

Use inspect the raw data, raw measurements and retain the raw record. Tricia checks an unusual mean by returning to the original trials. For practice, compare two groups with the same mean but different raw values. Explain what the summary hides and why preserving the individual observations improves interpretation.

44. Result

A result is an outcome obtained from an investigation, calculation or analysis. A result can be a measured value, a comparison between conditions or a pattern observed across trials. “The long-blade condition had a greater mean descent time” is a result when it follows from the supplied measurements.

Result is not the same as conclusion. The result describes what the data show; the conclusion connects that result to the research question and may discuss support, limitations or explanations. Keeping them separate helps prevent a descriptive difference from becoming an unsupported causal claim.

Use the result shows carefully, result of the calculation and the observed result. Alicia states the numerical comparison before interpreting it. For practice, write one result sentence and one conclusion sentence from the same fictional dataset. Underline the extra reasoning added by the conclusion.

45. Pattern

A pattern is a recurring or organised relationship in observations. If descent time increases across several increasing blade lengths, the data may show an upward pattern under the tested conditions. A pattern can guide explanation and further testing without by itself identifying the mechanism responsible.

Two points make a difference, but they provide limited evidence of a broader pattern. More levels can show whether the relationship continues, levels off or reverses. A visual pattern can also be affected by graph scale, missing values or one extreme observation, so the learner should inspect the numerical record as well as the picture.

Use pattern across the conditions, consistent pattern and the pattern is interrupted by. Tricia names exceptions rather than hiding them. For practice, compare a smooth four-point sequence with a sequence that shares the same endpoints but reverses twice. Explain why endpoint change and overall pattern are not identical descriptions.

46. Trend

A trend is a general direction of change across ordered observations, such as increasing, decreasing or remaining broadly stable. In an experiment with several levels of an independent variable, a trend can summarise how the dependent variable changes across those levels.

A trend need not mean every neighbouring value moves in the same direction. Nor does a trend guarantee that the same relationship will continue beyond the tested range. Extrapolation is a new claim. State the observed range before extending a pattern into conditions the experiment did not include.

Use an increasing trend, trend across the tested range and the trend is not strictly monotonic only when that technical phrasing is useful. Kai Kai first describes the values in ordinary language. For practice, write a trend statement that includes one exception without turning the exception into a reason to ignore the overall relationship.

47. Anomaly

An anomaly is an observation that differs unexpectedly from the main pattern or from what the procedure suggests should occur. One unusually long timing among several close trials may deserve checking. The word identifies its unusual relationship to the rest of the data; it does not establish that the value is wrong.

An anomalous result can arise from genuine variation, an unrecorded condition, measurement error or another cause. Investigate before excluding it. If the source record confirms that the procedure was followed and the value is valid, the anomaly may be scientifically informative. Deleting it solely because it weakens the expected conclusion is not a fair analysis rule.

Use an anomalous observation, investigate the anomaly and retained after verification. Alicia checks the original trial note. For practice, write two possible outcomes after checking an anomaly: one where a transcription error is corrected and one where the unusual value is confirmed and remains in the dataset.

48. Outlier

An outlier is a value unusually far from the other observations according to an appropriate context or rule. In a small classroom dataset, a learner can identify a possible outlier descriptively and investigate it. A formal statistical outlier rule, when required, should be stated rather than invented after seeing which value is inconvenient.

Outlier and anomaly overlap but are not exact synonyms in every context. Outlier emphasises numerical distance from the main group; anomaly can refer more broadly to an unexpected observation or event. Neither word justifies automatic removal. A valid extreme observation can change the mean precisely because it is part of the phenomenon being studied.

Use a possible outlier, extreme observed value and outlier rule. Tricia reports the summary both with and without a questioned value only when the comparison helps show its influence and the reason is explained. For practice, state why that sensitivity check is different from secretly choosing whichever version supports the preferred hypothesis.

49. Uncertainty

Uncertainty is the remaining limit on what is known about a measurement, estimate or conclusion. A timer’s finite resolution, variation among repeated trials and incomplete control of conditions can each contribute to uncertainty of different kinds. Naming the source helps the reader understand what the result does and does not establish.

Uncertainty does not mean that all answers are equally possible or that nothing has been learned. A result can strongly favour one condition while still having finite measurement limits. The useful task is to calibrate the conclusion to the evidence rather than alternate between false certainty and total doubt.

Use measurement uncertainty, uncertainty about the cause and the conclusion remains uncertain because. Kai Kai distinguishes numerical uncertainty from causal uncertainty. For practice, write one secure fact and one remaining uncertainty from the same experiment. Explain why recognising the second does not cancel the first.

50. Error

Error, in measurement and experimental work, means a difference between a measured, recorded or inferred result and the relevant reference or intended procedure. It does not automatically mean someone behaved carelessly. Measurement error can arise even when people follow a sensible method because instruments and procedures have limitations.

Error also needs a type. A one-off transcription mistake differs from random variation among readings and from a systematic offset affecting every measurement. Calling all three human error hides the mechanism needed for improvement. The repair should follow the particular source: correct the transcription, improve repeatability or recalibrate the instrument.

Use measurement error, recording error and source of error. Alicia names what went wrong rather than blaming a person generally. For practice, take the sentence “The experiment had human error” and rewrite it as a specific, observable procedural problem that another group could actually address.

Entries 51–60: move from differences to causes without skipping the design

51. Random error

Random error is unpredictable variation that causes repeated measurements to scatter around a value or pattern. Small differences in human reaction time when starting a stopwatch can contribute random timing variation. Repeated trials and improved procedures can help estimate or reduce its influence on a summary.

Random does not mean lawless or necessarily large. It describes variation whose direction is not consistently the same. Repeating measurements can reduce the effect of some random errors on a mean, but it does not repair a systematic offset that shifts every reading similarly.

Use random measurement error, trial-to-trial scatter and reduce random variation. Tricia compares the spread of repeated values. For practice, explain why collecting ten readings from a stopwatch that is consistently two seconds slow will not remove that systematic problem even if the average becomes very stable.

52. Systematic error

A systematic error shifts measurements in a consistent direction because of the instrument, calibration, procedure or another recurring factor. A balance that reads two grams high for every object can create such an offset. Repetition can make the shifted average look very precise without making it accurate.

A systematic error is not simply any mistake that happens repeatedly. The key idea is a directional bias built into the measurement process. Its effect may depend on the instrument or procedure. Checking against a known reference can reveal some systematic errors that repeated unknown samples alone would not expose.

Use systematic measurement error, consistent offset and check calibration. Kai Kai distinguishes systematic as orderly from systematic error as directional measurement bias. For practice, write a dataset that is tightly clustered but shifted from a known reference, then explain why precision and accuracy lead to different evaluations.

53. Bias

Bias is a systematic tendency in selection, measurement, procedure or interpretation that can shift a result away from the intended quantity or comparison. Measuring only the strongest-looking paper samples would bias a material test if the question concerns ordinary samples. The mechanism should be named rather than assumed to involve dishonest motives.

Bias can occur unintentionally. A researcher who knows which condition is expected to win may unconsciously choose borderline outcomes differently unless the scoring rule is clear. A strong method reduces opportunities for such influence through standardisation, predefined criteria or, where suitable, masking of condition labels.

Use selection bias, measurement bias and the procedure may bias the comparison. Alicia describes the route by which the result could be shifted. For practice, improve a sample-selection method without predicting what the newly included cases will show. Reducing bias does not supply the missing result in advance.

54. Association

An association is an observed relationship between variables. Across several fictional conditions, longer blades may be associated with longer descent times. This describes how two variables vary together in the data. Whether blade length caused the difference depends on the design and competing explanations.

Association can appear in observational data where no variable was deliberately changed. It can also appear in an experiment. The word is useful because it reports a relationship without automatically claiming a mechanism. A strong answer can later make a causal interpretation if the experimental design and evidence justify it.

Use an observed association, associated with and the association alone does not establish. Tricia states the variables and tested range. For practice, rewrite a causal headline as a descriptive association statement, then list the additional design features needed before returning to the stronger causal wording.

55. Correlation

Correlation is a statistical relationship describing how numerical variables vary together. At this level, a learner can understand the conceptual boundary without calculating a formal correlation coefficient: a consistent upward relationship is different from proof that one variable causes the other.

Correlation can arise because one variable influences another, because a third factor affects both, because of shared trends or by chance in a small dataset. A controlled experiment can provide stronger causal evidence than a simple correlation when its design successfully isolates the manipulated factor. The word should not become a synonym for any pair of lines that look similar.

Use positive correlation, correlated variables and correlation does not by itself establish causation. Kai Kai names the alternative explanations. For practice, invent a harmless observational example in which two quantities rise together because both respond to time or another third factor rather than directly causing each other.

56. Causation

Causation concerns whether a change in one factor produces or contributes to a change in another under the relevant conditions. Experimental designs aim to strengthen causal reasoning by deliberately changing a factor and controlling alternatives. A measured difference after that change is evidence, but the strength of the causal claim depends on how well the design isolates the relationship.

Causation should not be reduced to “A happened before B, so A caused B.” Timing is relevant but insufficient. Nor does failure to establish causation prove that no causal relationship exists. It may mean the current test cannot distinguish the possibilities. A careful conclusion separates absence of evidence from evidence of absence.

Use causal effect, evidence for causation and causation is not established by this comparison alone. Alicia identifies the design feature that supports the causal interpretation. For practice, compare a controlled test with an observational before-and-after record and explain why identical numerical changes can support different strengths of conclusion.

57. Mechanism

A mechanism is a process or set of interactions that explains how one condition could produce an outcome. A paper spinner with greater blade area might interact differently with the surrounding air, providing a proposed physical mechanism for a change in descent time. The mechanism connects a pattern with an explanatory model.

A plausible mechanism is not automatically established by the experiment. The measured outcome may fit several mechanisms. The investigation might test the effect while leaving the detailed process unmeasured. Write “consistent with the proposed mechanism” when the data fit it but do not distinguish it from alternatives.

Use proposed mechanism, mechanism linking and the mechanism was not directly measured. Tricia asks which observations would discriminate between explanations. For practice, write one effect claim and one mechanism claim from the same fictional result, then identify the extra evidence required by the mechanism statement.

58. Isolate

To isolate a factor is to design the comparison so its contribution can be distinguished from relevant alternatives. Keeping paper thickness the same while varying blade length helps isolate blade length in the spinner investigation. The word describes a design goal rather than a guarantee that every hidden influence has been eliminated.

Isolation can be partial. A classroom design may control the most important visible conditions while still facing human timing variation. State what has been controlled and what remains a limitation. Do not call the result perfectly isolated merely because one checklist has been completed.

Use isolate the effect of, better isolated and the design does not isolate. Kai Kai identifies the competing factor in a flawed comparison. For practice, redesign an experiment changing both material and thickness so that one version isolates material type and another isolates thickness. Explain why the research question changes with the redesign.

59. Attribution

Attribution is the act of assigning an observed change or outcome to a particular cause or source. In experimental writing, causal attribution should follow from the design. If only blade length differs meaningfully and repeated timings show a consistent difference, the result may support attributing part of that difference to blade length under the tested conditions.

Attribution can be too strong when other factors changed. “The improvement was due to the new material” is unsupported if the new sample was also thicker and tested longer. A better report names the confounded comparison and withholds the single-cause attribution until the factors are separated.

Use attribute the difference to, causal attribution and cannot be attributed solely to. Alicia checks what else differs. For practice, replace solely with wording that matches a design in which two factors changed. Explain why removing one word can materially alter the scientific claim.

60. Alternative explanation

An alternative explanation is another plausible account of the observed result. If one spinner falls faster, blade length may be one explanation; a different paper thickness or release push can provide alternatives. Considering them helps the learner test whether the intended variable has really been isolated.

An alternative explanation should be plausible and connected to the evidence, not invented merely to avoid every conclusion. Once a factor has been carefully controlled, that particular alternative may become less credible. Scientific reasoning narrows possibilities through evidence rather than treating all explanations as permanently equal.

Use an alternative explanation is, rules out this alternative cautiously, and the design leaves open. Tricia asks which additional comparison would distinguish the possibilities. For practice, propose one alternative explanation for a result and one new condition that would help test between the competing accounts.

Entries 61–70: write the strongest conclusion the experiment can actually carry

A scientific conclusion is not a ceremonial paragraph added after the graph. It answers the research question by connecting the observed result with a claim whose scope matches the design. The words in this group help the learner distinguish support from proof, disagreement from refutation, and a useful limitation from a vague statement that “the experiment may have errors.”

61. Support

To support a claim is to provide evidence that increases its credibility within the relevant context. If longer-blade spinners repeatedly show greater descent times under a fair comparison, those results support the hypothesis that blade length affects descent time under the tested conditions.

Support is not the same as prove. One small classroom investigation does not establish a universal law for every spinner design, material and environment. The word is useful because it lets the writer state that the evidence matters without exaggerating what has been settled.

Use the results support, evidence supporting and support within the tested range. Alicia names both the claim and the evidence. For practice, replace “proved” in a conclusion with support language that retains confidence in the actual result while preserving the experiment’s boundaries.

62. Contradict

To contradict is to be inconsistent with a statement, prediction or expectation. If a prediction says the long-blade condition will have the longest descent time but the repeated results consistently show the opposite ordering, the data contradict that prediction under the test conditions.

A contradictory result does not automatically identify why the expectation failed. The hypothesis may be wrong, the proposed mechanism incomplete, the method flawed or the measured range unusual. The correct next move is investigation, not altering the data so the original prediction survives.

Use contradicts the prediction, inconsistent with and the data do not match. Tricia preserves the unexpected result. For practice, write a conclusion that acknowledges contradiction and proposes one specific design check without presenting the check as an already established explanation.

63. Refute

To refute a claim is to provide evidence or reasoning that shows the claim, as stated, cannot be sustained. The word is stronger than contradict. One result that differs from a prediction may not refute a broad explanation when measurement uncertainty or alternative conditions remain open.

Refutation should therefore match the claim’s precision. A universal statement such as “all longer blades always fall more slowly” can be refuted by a valid counterexample within its stated scope. A narrower probabilistic or conditional claim may require a different evaluation. Do not use refute simply as an impressive synonym for disagree.

Use refute the universal claim, evidence is insufficient to refute and the observation provides a counterexample. Kai Kai checks the wording of the original claim. For practice, compare a universal statement with a conditional statement and explain why the same new observation can have different logical effects on them.

64. Inconclusive

Inconclusive describes evidence that does not allow the intended question to be answered with sufficient confidence or separation between alternatives. If two spinner conditions have heavily overlapping, highly variable timings and only one trial each, the comparison may be inconclusive about which condition tends to descend more slowly.

Inconclusive does not mean useless. The investigation may reveal that the measurement method is too variable, that more trials are needed or that the effect is smaller than expected. A report should state what remains unresolved and what was learned about the method rather than end with “nothing can be concluded.”

Use the evidence is inconclusive about, inconclusive under this design and further trials could address. Alicia identifies the specific unresolved comparison. For practice, revise “The experiment failed” into a more informative statement naming why the present evidence cannot distinguish the two conditions.

65. Conclusion

A conclusion is the reasoned answer drawn from an investigation’s results in relation to its question. A strong conclusion identifies the main finding, explains whether it supports the hypothesis and states important limitations or conditions. It should not introduce a new measurement that never appeared in the method or results.

The conclusion is not simply a summary of every number. It selects the evidence that answers the question and interprets that evidence. It also should not become a story about motives or mechanisms the experiment did not test. A concise, bounded conclusion can be scientifically stronger than a dramatic paragraph claiming more.

Use draw a conclusion, the conclusion applies to and the conclusion is limited by. Tricia checks every major noun against the variables actually measured. For practice, turn a three-row results table into a two-sentence conclusion containing the comparison, direction and scope without copying all three rows.

66. Limitation

A limitation is a feature of the design, data or scope that restricts what can be concluded. Human reaction time in manual timing may limit precision. Testing only one paper type may limit generalisation to other materials. A limitation should connect directly to the claim it affects.

“There may be human error” is often too vague to be useful. Name the specific process and likely consequence. A limitation also does not erase a valid narrow result. If the same timing method was used across all conditions, relative differences may still be informative even when the absolute times have finite measurement uncertainty.

Use a limitation of the method, limits the conclusion about and does not invalidate the measured comparison. Kai Kai links each limitation to one affected inference. For practice, write one limitation that concerns measurement and another that concerns generalisation, then explain why their remedies differ.

67. Generalise

To generalise is to extend a finding beyond the exact observations or conditions studied. A learner may wish to generalise from one paper-spinner design to other spinners, or from one classroom material sample to a broader class of materials. That extension requires justification; it is not contained automatically in the original measurements.

Generalisation is stronger when the tested cases represent the wider range of interest and when similar results are reproduced under varied relevant conditions. A result from one specific setup can still be useful even when broad generalisation is not justified. State the narrower finding rather than pretending that lack of universality makes it worthless.

Use generalise beyond the tested conditions, limited generalisability as advanced extension, and the result applies directly only to. Alicia asks what population or range the writer wants to include. For practice, identify three extra conditions that would need testing before a narrow paper result could reasonably support a wider material claim.

68. Implication

An implication is a consequence, meaning or next consideration suggested by a result. If one insulation model loses less heat under the tested setup, an implication may be that its construction is worth testing under additional durations or materials. The implication extends the usefulness of the finding without pretending the new test has already been completed.

An implication is not the same as a guaranteed practical effect. A tabletop model does not automatically establish building performance. A scientific implication should follow logically from the result and remain proportionate to the model. Recommendations for real-world use may require evidence about cost, durability, safety and scale that the experiment did not collect.

Use one implication is, the result suggests testing and practical implications remain uncertain. Tricia separates next question from established answer. For practice, write one implication that proposes a future investigation and one overstatement that claims the future result already. Explain the difference.

69. Claim

A claim is a statement that asserts something that can be supported, challenged or evaluated. In scientific writing, “Increasing blade length increased mean descent time under the tested setup” is a claim. Its strength depends on the experimental design, measurements and reasoning connecting them.

A claim is not simply an opinion because a person believes it strongly. Nor is every sentence in a report a claim of equal importance. Method statements describe what was done; result statements report observations; the central claim answers the scientific question. Distinguishing these roles makes evidence easier to match.

Use scientific claim, claim supported by and claim exceeds the evidence. Kai Kai underlines the sentence that actually answers the question. For practice, extract one claim from a paragraph containing method, data and explanation, then identify the two specific pieces of evidence intended to support it.

70. Evidence

Evidence is information used to support or evaluate a claim. In an experiment, evidence can include relevant measurements, observations and patterns produced under the design. Not every recorded fact is equally relevant to every claim. The colour of the stopwatch is usually not evidence about blade-length effects on descent time.

Evidence gains meaning through method and context. A numerical difference from a confounded comparison is still data, but it provides weaker evidence for a single-cause claim. More data of the same flawed kind may increase quantity without resolving the design problem. Quality and relevance matter alongside amount.

Use evidence for the claim, relevant evidence and the evidence does not distinguish. Alicia asks how each selected result bears on the conclusion. For practice, remove one irrelevant detail from a scientific paragraph and add one missing methodological detail that changes how the evidence should be interpreted.

Entries 71–80: connect observations with reasoning and models

71. Reasoning

Reasoning is the logical connection explaining why the evidence supports a claim. In a claim–evidence–reasoning structure, evidence identifies the relevant result while reasoning uses scientific ideas to connect that result to the conclusion. It is the bridge, not a repetition of the number.

Reasoning should not introduce a mechanism as fact when the experiment did not test it. A learner can use an established concept to explain why the observed relationship is plausible, while identifying a proposed mechanism as a model if necessary. The strength of the reasoning depends on whether its principles apply to the actual conditions.

Use the reasoning connects, reason from the evidence and the reasoning requires the assumption that. Tricia asks what sentence explains why the selected measurement matters. For practice, take a claim and evidence pair and add one reasoning sentence without simply restating either.

72. Justify

To justify is to give reasons and evidence showing why a choice, claim or interpretation is warranted. A learner can justify selecting a longer testing duration by explaining that the shorter duration produced changes too small for the instrument to distinguish reliably. The justification links the decision to the investigation’s purpose.

Justification is not the same as defending every decision after the fact. Some choices may need revision when evidence reveals a weakness. A strong scientist can justify a change of method by identifying the earlier limitation rather than pretending the first design was already optimal.

Use justify the choice, justified by the evidence and insufficiently justified. Kai Kai distinguishes preference from reason. For practice, compare “We used five trials because five is better” with a justification connected to observing variation and reducing dependence on one unusual result.

73. Model

A model is a simplified representation used to describe, explain or predict aspects of a system or phenomenon. A paper spinner is a model object for studying relationships between shape and descent, not a complete representation of every falling object. Models deliberately include some features and omit others.

A model can be physical, conceptual, diagrammatic or mathematical. Its usefulness depends on the question. A simple model can clarify one mechanism while being inappropriate for direct prediction of a complex real system. State which relationship is being modelled instead of treating model as another word for exact copy.

Use physical model, model represents and limitation of the model. Alicia names one feature preserved and one omitted. For practice, explain why a model-container insulation experiment can compare heat-loss conditions without automatically establishing the performance of a full building.

74. Representation

A representation is a way of expressing or displaying information about something else. A diagram can represent apparatus, a graph can represent measured results and an equation can represent a relationship between quantities. The representation is useful because it selects and organises relevant features.

The representation should not be confused with the phenomenon itself. A smooth line between measured points can suggest a trend without showing that every intermediate value was observed. A schematic diagram may not preserve physical scale. Read the convention before extracting information the representation was not designed to carry.

Use graphical representation, represents the measured values and schematic representation. Tricia asks what the symbols encode. For practice, compare a table and a graph of the same results. Identify one relationship each makes easier to see and one detail the graph could hide if the raw values are not labelled.

75. Simulation

A simulation is a model that imitates selected behaviours or processes so they can be explored without reproducing the full real system. A computer or classroom simulation can vary conditions repeatedly and show predicted outputs under its programmed rules. Those outputs belong first to the simulation.

A simulation result is not automatically an empirical observation of the real world. Its usefulness depends on the assumptions, equations and data built into the model. Comparing simulated predictions with actual observations can test the model, but agreement in one case does not guarantee accuracy everywhere.

Use simulation output, simulate the process and the simulation assumes. Kai Kai labels simulated and measured values separately. For practice, write a sentence that reports a model prediction without presenting it as something already observed in a physical experiment.

76. Assumption

An assumption is a condition accepted for the purpose of reasoning, modelling or simplifying an investigation. A calculation may assume that a paper layer has uniform thickness across the sample. The assumption allows the model to proceed but should be recognised when it affects the conclusion.

Assumptions are not automatically weaknesses. Every model simplifies. The important question is whether the assumption is reasonable enough for the intended use and whether the conclusion changes substantially when it is relaxed. Hidden assumptions are harder to evaluate than explicit ones.

Use under the assumption that, model assumption and the conclusion depends on. Alicia writes the assumption beside a derived result. For practice, identify one assumption in a fictional equal-layer comparison and explain how non-uniform layer thickness could change the interpretation.

77. Explain

To explain is to show how or why a result, process or relationship occurs using relevant evidence and concepts. A scientific explanation goes beyond describing that one condition produced a larger value. It connects the pattern to a mechanism, model or principle appropriate to the question.

An explanation should not be stronger than the evidence. If the mechanism was not directly tested, it can be presented as a plausible account consistent with established concepts. The learner should distinguish “the result was greater” from “the result was greater because…” and check what supports the because clause.

Use explain the observed difference, explanation consistent with and the data describe but do not fully explain. Tricia marks where the paragraph shifts from description to explanation. For practice, add one scientifically relevant explanation sentence to a result while explicitly identifying one untested mechanism detail.

78. Evaluate

To evaluate is to judge the quality, significance or suitability of evidence, a method or a conclusion using stated criteria. A learner can evaluate whether the number of trials was sufficient to reveal the observed variation, whether the control condition addresses the question or whether the conclusion overgeneralises.

Evaluation is more than saying good or bad. It needs a reason and consequence. “The timing method has limited precision because manual reaction time contributes variation; using an automated sensor could reduce this source” is more useful than “The stopwatch was inaccurate.” The first identifies the mechanism and a plausible improvement.

Use evaluate the method, evaluate against the criterion and evaluation of the evidence. Kai Kai supports each judgement with a feature of the design. For practice, write one strength and one limitation that can both be true of the same investigation.

79. Compare

To compare is to examine two or more conditions or results using the same relevant measure. A fair comparison of insulation models uses the same temperature-change definition and duration across conditions. The word should direct attention to a dimension, not merely place two numbers beside one another.

Comparability can fail when units, procedures or baseline conditions differ. A five-degree change over ten minutes and a six-degree change over an hour answer different time-based questions. Standardising or converting may help in some cases, but a conversion cannot repair fundamentally different experimental definitions.

Use compare under the same conditions, direct comparison and not directly comparable. Alicia names the measure in the comparison sentence. For practice, identify one valid and one invalid comparison from a mixed table containing different durations and units, then explain why the invalid one needs more than arithmetic.

80. Contrast

To contrast is to highlight a relevant difference between conditions, results or explanations. A report can contrast two datasets that share the same mean but differ in variation, or two methods that measure the same outcome using different instruments. The contrast should help answer a question rather than collect unrelated differences.

Contrasting does not require treating the conditions as complete opposites. Two procedures can differ in one important feature while sharing many others. Preserving common features can make the specific contrast clearer. In experimental design, the most informative contrast often isolates the intended difference while holding the rest comparable.

Use in contrast, contrast the methods and the key contrast is. Tricia states the shared baseline before the difference. For practice, write one sentence contrasting a fair comparison with a confounded one without implying that every other part of the two investigations differs.

Entries 81–90: keep safety, ethics and record integrity inside the investigation

Scientific quality is not only a matter of getting a clean graph. A useful investigation must also be appropriate to its setting, document what was done and avoid exposing people, animals or the environment to unjustified harm. The examples here remain conceptual and use safe fictional classroom contexts. Follow local adult, school and laboratory requirements for any real activity.

81. Safety

Safety concerns preventing or reducing avoidable harm during an investigation. A classroom design should use materials, tools and procedures appropriate to the learners and setting. Safety constraints can legitimately change the experimental question: a safe model may be chosen instead of a direct real-world test.

Safety is not an obstacle to scientific thinking. It is one of the conditions under which responsible investigation occurs. A safer procedure may also improve standardisation by controlling handling and exposure. Never treat a more dangerous experiment as automatically more authentic or scientifically impressive.

Use safety requirement, safe procedure and within the permitted safety conditions. Alicia checks the allowed method before planning measurements. For practice, convert a risky imagined question into a safe paper or simulation model while stating which part of the original real-world phenomenon the model no longer represents.

82. Hazard

A hazard is something with the potential to cause harm. Sharp edges, heat, unstable equipment or unsuitable substances can be hazards in different settings. Identifying a hazard is not the same as calculating how likely harm is under a particular procedure; that broader judgement belongs to risk assessment.

A hazard can often be removed from an educational investigation by changing the apparatus or using a model. If the question can be answered with ordinary paper, room-temperature materials or supplied datasets, adding unnecessary hazardous elements provides no scientific advantage. The safest adequate design is generally the better classroom design.

Use identify the hazard, remove the hazard and hazard associated with. Tricia does not assume that protective equipment makes every hazardous activity appropriate. For practice, distinguish eliminating a hazard from reducing exposure to it and explain why elimination can be preferable when the scientific question still remains answerable.

83. Risk

Risk concerns the chance and possible severity of harm occurring under stated conditions. A low-hazard classroom material used carelessly can still create risk, while a potentially hazardous item may be excluded entirely from a student investigation. Risk depends on the situation and procedure, not only on the name of the object.

Scientific curiosity does not override reasonable risk controls. An investigation can use existing data, simulations or safer analogues when direct testing is inappropriate. The educational objective is learning to ask and answer questions responsibly, not proving willingness to accept avoidable danger.

Use assess the risk, reduce risk and risk is unnecessary for this question. Kai Kai asks whether the same relationship can be studied more safely. For practice, propose a safe substitute for a direct test and state the limitation introduced by the substitute.

84. Precaution

A precaution is an action taken to reduce the chance or consequence of harm or error. In a safe paper experiment, clearing the floor around a drop zone can be a simple precaution against tripping. In measurement, placing the ruler flat and reading from the correct mark can be a procedural precaution against avoidable error.

A precaution should address a specific risk or measurement problem. Writing “be careful” is less useful than naming the action and what it prevents. Precautions also do not justify an activity that remains inappropriate despite them. Risk should be reduced by design first where possible.

Use take the precaution of, measurement precaution and safety precaution. Alicia links every listed precaution to a reason. For practice, replace three vague safety instructions with specific actions appropriate to a harmless classroom timing activity.

85. Ethical

Ethical describes conduct that follows relevant principles about responsible treatment, consent, honesty and avoidance of unjustified harm. An investigation involving people or animals raises considerations beyond whether the measurement would be interesting. At Primary 6 level, the practical lesson is that some questions should use anonymised supplied data, simulations or non-living models instead.

Ethical does not mean whatever the investigator personally prefers. Schools, institutions and communities may have formal rules governing what may be studied and how. A learner should not attempt to design human or animal experiments independently from appropriate adult and institutional guidance.

Use ethical consideration, ethical constraint and ethically appropriate model. Tricia chooses a fictional dataset when direct collection would be intrusive. For practice, explain why replacing identifiable personal responses with a supplied anonymous dataset changes what can be learned while also reducing privacy concerns.

86. Consent

Consent is agreement given for participation under appropriate information and conditions. In formal research involving people, consent has ethical and procedural requirements that exceed an ordinary classroom word definition. This volume does not provide a protocol for conducting human-subject research; it uses the term to teach why people cannot simply be turned into experimental objects.

A classmate answering a casual question does not automatically create permission for their response, image or identity to be published or used in a wider study. Where school-approved activities involve people, follow the institution’s rules and adult supervision. When those requirements are not in place, use fictional or already-authorised data for practice.

Use informed consent as an advanced recognition term and participation requires appropriate consent. Kai Kai keeps the discussion conceptual. For practice, identify which scientific questions in this article can be answered entirely with paper models or supplied fictional records, avoiding personal data collection altogether.

87. Transparency

Transparency means making relevant methods, decisions, assumptions and changes visible enough for others to understand how a result was produced. A transparent report states that one anomalous trial was excluded and explains the predefined rule rather than silently presenting only the remaining values.

Transparency does not require publishing private information or every irrelevant detail. The principle concerns information necessary for evaluating the work. A clear method, data-handling rule and revision record can make a short report more trustworthy than a long report that hides consequential choices.

Use transparent reporting, state the decision transparently and transparent about limitations. Alicia marks where a method changed after a pilot trial. For practice, write a two-sentence change note that preserves what was altered, why, and which data were collected under each version.

88. Document

To document is to record information in a form that allows later checking, interpretation or reproduction. A lab notebook can document the procedure, condition labels, measurements, anomalies and changes. Documentation is part of the evidence trail because memory can change after the result is known.

Documenting is not the same as copying everything without structure. The record should preserve consequential information with dates or trial identifiers where useful. A photo can supplement a written description but may not show the measurement rule or timing conditions. Different record forms serve different purposes.

Use document the procedure, document each change and documented observation. Tricia writes notes during rather than after all trials. For practice, list four items that should be recorded immediately in a repeated timing test and explain which later interpretation each item helps check.

89. Provenance

Provenance is information about where data, materials or records came from and how they were produced. In a school investigation, provenance can include which group collected a dataset, under which version of the procedure and on which date. The term helps prevent one table from becoming detached from the conditions that give its values meaning.

Provenance is especially useful when combining records. Two files may have identical column headings but different measurement rules. Knowing their origin can reveal that they should not be merged directly. Source information does not guarantee quality, but it allows quality to be assessed more intelligently.

Use data provenance, record the provenance and provenance note. Kai Kai labels each dataset with method version. For practice, explain why “final-results.csv” is an inadequate source description if several groups used different timing rules.

90. Traceability

Traceability is the ability to follow a result back through the records, measurements and method that produced it. A graph point is traceable when the learner can identify the original trials, their units and any calculation used to create the plotted value. Traceability helps diagnose errors without guessing where they entered.

A polished chart with no underlying record has weaker traceability than a chart linked to its source values. Traceability does not mean every reader must see all raw notes in the main report. It means the evidence chain is preserved so that important results can be checked when necessary.

Use traceable to the raw measurements, maintain traceability and the result cannot be traced to. Alicia assigns trial identifiers before averaging. For practice, describe how one reported mean should connect to five raw timings and one calculation rule.

Entries 91–100: turn the investigation into a checkable explanation

91. Record

To record is to capture an observation, measurement or decision in a durable form. Recording should happen according to a defined rule, with units and condition labels where relevant. “1.72” is not a complete timing record if the reader cannot tell which condition, trial and unit it belongs to.

Recording differs from remembering. Notes made only after the whole activity can be influenced by the final result or omit inconvenient detail unintentionally. A structured table or notebook keeps the evidence available for later analysis and verification.

Use record the observation, recorded immediately and recording rule. Tricia writes values without rounding them prematurely. For practice, design one data-table row that preserves condition, trial number, measurement, unit and a short note for an invalid or unusual event.

92. Measure

To measure is to determine a quantity using a specified instrument, unit and procedure. Measuring differs from merely noticing that one object seems longer or heavier. A measurement allows the comparison to be checked numerically when the method is appropriate.

Measuring the wrong quantity precisely does not answer the intended question. If the research question concerns time but the procedure only counts rotations, the count may be interesting yet insufficient unless a valid relationship to time has been defined. Start with the question, then choose the measurement.

Use measure the dependent variable, measured in and measure from the same reference point. Kai Kai names the instrument and endpoints. For practice, identify the measurement hidden inside the phrase “stronger paper bridge” and rewrite it as a defined quantity.

93. Analyse

To analyse is to examine data systematically to identify relationships, summaries, variation and evidence relevant to the question. Analysis can include calculating means, comparing conditions, inspecting raw values, identifying anomalies and checking whether a trend is consistent across trials.

Analysis should not begin by deleting values that complicate the expected story. Nor is making a graph automatically analysis. The learner must interpret what the graph or calculation reveals and how it bears on the question. American English commonly uses analyze.

Use analyse the results, analysis of variation and the analysis compares. Alicia moves from raw data to a justified summary. For practice, list three analytical steps for a five-trial dataset and explain the question each step answers.

94. Interpret

To interpret is to explain what a result means in relation to the question, method and scientific context. A mean difference can be interpreted as evidence of a condition effect only after considering whether the groups were comparable and whether variation is small enough to make the pattern meaningful.

Interpretation is not permission to invent a cause. It is a reasoned reading of the evidence. A graph can show that one condition had a higher measured value; whether the mechanism is known is a separate question. Keep the observed relationship and explanatory model distinct when necessary.

Use interpret the result, the result means that within this test and an interpretation consistent with. Tricia writes the narrow meaning before the broader implication. For practice, add one sentence of interpretation to a numerical difference without using caused unless the design actually supports that word.

95. Verify

To verify is to check a value, calculation, record or method against an appropriate reference. Recalculating a mean verifies the arithmetic. Comparing a copied trial value with the notebook verifies the transcription. Checking a ruler against a reference length tests another layer entirely.

Verification has scope. Correct arithmetic cannot verify that the selected data came from a fair test. A matching transcription cannot verify that the original measurement was accurate. State what was checked and what remains a separate scientific question.

Use verify the calculation, verified against the source record and verification of the method requires. Kai Kai matches each possible error with a suitable check. For practice, explain why pressing the calculator twice may verify arithmetic but not the experiment’s causal interpretation.

96. Qualify

To qualify a scientific statement is to limit it so the wording matches the evidence. “Longer blades fall more slowly” may become “In this paper-spinner design, the longer-blade conditions had greater mean descent times over five trials.” The added conditions make the statement more precise, not weaker.

A good qualification identifies the real boundary. Adding perhaps to every sentence does not improve scientific quality. If the measured difference is secure, state it directly. Qualify the extension that remains uncertain, such as the generalisation to other materials or the proposed mechanism.

Use qualify the conclusion, qualified by the tested range and the claim applies under. Alicia removes always and adds the actual experimental conditions. For practice, repair an overgeneralised claim using the fewest words needed to make its scope accurate.

97. Revise

To revise is to change a question, method, analysis or conclusion after checking it against evidence. A pilot trial may reveal that the timing interval is too short to produce measurable differences, leading the learner to revise the procedure before collecting the main dataset.

Revision is not manipulation of the experiment to force a preferred outcome. A legitimate revision is documented and justified by a methodological reason. Data collected under different method versions should remain distinguishable rather than silently pooled as if nothing changed.

Use revise the procedure, revision prompted by and revised conclusion. Tricia writes what changed and why. For practice, compare a transparent pilot-based revision with an improper change made only after seeing which setting produces the desired result.

98. Substantiate

To substantiate is to support a claim with adequate relevant evidence. A repeated, fair comparison may substantiate a narrow statement about which tested condition produced the higher measured outcome. It does not automatically substantiate claims about every material, mechanism or practical application.

More technical vocabulary does not substantiate a weak claim. The evidence must bear on the statement. A paragraph containing reliability, validity and causation can still be unsupported if it never identifies the relevant measurements or design features.

Use substantiated by the measurements, insufficient to substantiate and substantiates the narrow claim. Kai Kai asks which exact evidence would have to disappear before the claim became unsupported. For practice, write a claim and the minimum evidence needed to substantiate it within a fictional paper test.

99. Communicate

To communicate scientifically is to present the question, method, evidence and conclusion so another person can understand and evaluate the work. Clear communication preserves units, conditions, uncertainty and evidence status. It helps a reader distinguish observation from inference and measured result from proposed explanation.

Communication does not improve by adding every technical term available. A simple sentence that states exactly what was compared can be stronger than a crowded sentence that uses advanced words incorrectly. Choose terminology when it clarifies a distinction the reader needs.

Use communicate the result, clear scientific communication and communicate the limitation. Alicia asks a partner to paraphrase the conclusion without seeing the raw notes. For practice, revise any sentence whose paraphrase changes the variable, unit or population.

100. Report

To report is to present what was done and found in a structured account. A scientific report can include the question, hypothesis, variables, method, results, analysis, conclusion and limitations. The exact format can vary; the essential requirement is that the evidence trail remains understandable.

Reporting should preserve the difference between recorded fact and later interpretation. “Three trials were excluded under the predefined rule” reports a methodological decision. “The material failed because it was badly made” adds an explanation that needs separate support. A report can be concise without hiding consequential choices.

Use report the results, scientific report and reported under the stated conditions. Tricia checks that the title does not promise more than the body establishes. For the final vocabulary task, write a one-paragraph report of a simple fictional fair test and underline the sentence that states the central conclusion.

Six original investigation laboratories with worked answers

Every investigation below is fictional teaching material. Where physical objects are mentioned, the values are supplied as complete records; learners do not need to recreate the experiments. This keeps the focus on experimental design and scientific language rather than equipment access. If an educator adapts any task into a real activity, appropriate school and adult safety requirements remain separate and must be followed.

Attempt each question before reading its worked answer. A good response does not need the model’s exact wording, but it should preserve the variables, units, design conditions and evidence limits. The repeated pattern is deliberate: identify the question, check the comparison, analyse the observations, and then write a conclusion that does not grow larger than the evidence.

Laboratory A: Blade length and paper-spinner descent time

A1 — question and design. A fictional group asks: “How does blade length affect the descent time of this paper-spinner design when released from the same height?” Two spinner conditions are prepared from the same paper template. The short-blade spinner has blades eight centimetres long. The long-blade spinner has blades sixteen centimetres long. All other dimensions are identical.

A2 — controlled conditions. Every trial uses the same indoor location, paper type, clip mass, release height of 1.50 metres, release method with no push and the same timing protocol. One person releases the spinner and another operates the same timer. The procedure defines descent time from the instant the spinner is released until its lowest edge first touches the floor.

A3 — raw data. Short-blade times in seconds are 1.21, 1.25, 1.23, 1.20 and 1.26. Long-blade times are 1.78, 1.82, 1.79, 1.81 and 1.80. All ten trials are valid under the predefined procedure. The short-blade mean is 1.23 seconds. The long-blade mean is 1.80 seconds.

A4 — the proposed conclusion. Kai Kai writes, “Long blades make every falling object slower because the long spinner took 0.57 seconds longer.” Alicia argues that the result is strong within the model but the conclusion extends far beyond what was tested. Tricia asks whether the repeated trials support a real condition difference and which uncertainty still remains.

A5 — interpretation boundary. No other blade lengths, paper types, release heights or environmental conditions are included. The timing is manual. The experiment does not directly measure air forces or another mechanism; it measures descent time. It is therefore possible to support a relationship within the tested model without claiming that the mechanism has been directly observed.

Questions A

AQ1. Identify the independent variable, dependent variable and four controlled variables. Explain why blade length is a variable while eight centimetres is a value of that variable.

AQ2. Check the difference between the two means. What does 0.57 seconds represent, and what does it not represent?

AQ3. Evaluate the repeatability of the two conditions using the raw data qualitatively. Does the close clustering establish perfect accuracy?

AQ4. Write a conclusion that answers the research question without generalising to every falling object. Distinguish the measured result from a proposed mechanism.

AQ5. Name one measurement limitation and one generalisation limitation. Propose one follow-up condition that would extend the evidence without changing several variables at once.

Worked answers A

AQ1. Blade length is the independent variable because the group deliberately changes it from eight to sixteen centimetres. Descent time in seconds is the dependent variable because that is the measured outcome. Controlled variables include paper type, clip mass, release height, release method, location and timing protocol; any four correctly identified examples are sufficient. Blade length names the characteristic that can vary, whereas eight centimetres is one particular value or level of that variable. Confusing the variable with the value makes it difficult to describe more than one condition.

AQ2. The mean difference is 1.80 − 1.23 = 0.57 seconds. It represents the difference between the two condition means under the stated measurement procedure. It is not the time by which every individual long-blade trial exceeds every possible short-blade trial in all settings. It also does not directly measure air resistance or another force. The arithmetic compares recorded descent-time summaries; the physical explanation is a further interpretive step.

AQ3. Both sets are closely clustered: short-blade values span only 0.06 seconds and long-blade values 0.04 seconds. That suggests good repeatability for this manual method within each condition. Close agreement does not prove accuracy because the timer could have a consistent offset or the human start rule could be systematically late. Repeatability describes consistency among repeated measurements, while accuracy requires an appropriate reference or other check.

AQ4. A suitable conclusion is: “In this paper-spinner design, increasing blade length from eight to sixteen centimetres increased the mean descent time from 1.23 to 1.80 seconds across five trials per condition. The result supports a relationship between blade length and descent time under the controlled setup. A greater interaction with the surrounding air is a plausible mechanism, but that mechanism was not measured directly.” The sentence states the observed result confidently while labelling the mechanism appropriately.

AQ5. Manual timing creates measurement uncertainty because human reaction time affects the start and stop. The study also uses only one paper type and two blade lengths, so broad generalisation is limited. A sensible follow-up could add a twelve-centimetre blade condition while keeping the other variables and timing protocol the same. That extends the tested range and helps reveal whether the relationship between the two original endpoints continues smoothly.

Transfer A. Suppose a later check shows that the long-blade spinner was released from 1.70 metres while the short one used 1.50 metres. The original causal interpretation becomes confounded because release height changed with blade length. The recorded times remain real observations, but they no longer isolate blade length. Explain why repeating the same unequal-height procedure ten more times could improve repeatability while leaving the causal attribution unresolved.

Laboratory B: The absorbent sheet that was twice as large

B1 — question. A fictional design team wants to compare how much water two paper materials retain after the same controlled contact and draining procedure. The draft question says, “Which material is more absorbent?” The method defines the dependent variable as the increase in sample mass, measured in grams, after the stated procedure.

B2 — sample sizes. Material A is tested with 10 cm × 10 cm pieces, an area of 100 square centimetres. Material B is accidentally tested with 20 cm × 10 cm pieces, an area of 200 square centimetres. Thickness is the same. All samples use the same contact duration, same draining duration and same balance.

B3 — supplied results. Mass increases for A are 8, 9, 8, 9 and 8 grams, with a mean of 8.4 grams. Mass increases for B are 12, 13, 12, 13 and 12 grams, with a mean of 12.4 grams. Kai Kai immediately concludes that B is more absorbent because its mean mass increase is four grams larger.

B4 — the design problem. Alicia notices that the tested areas differ by a factor of two. The raw result “B retained more water in these tested pieces” is true. The intended material comparison, however, is confounded by sample area. Tricia proposes either standardising the sample area in a new experiment or, for a descriptive secondary analysis, comparing mass retained per unit area while clearly stating that this normalisation was not the original fair-test design.

B5 — derived comparison. Using the supplied means, A retained 8.4 grams per 100 square centimetres. B retained 12.4 grams per 200 square centimetres, equivalent to 6.2 grams per 100 square centimetres under a simple proportional normalisation. That derived rate reverses the ranking, illustrating why the denominator matters. The normalisation does not prove how new equal-sized samples would behave; a fair retest remains preferable.

Questions B

BQ1. Identify the intended independent and dependent variables. Which uncontrolled factor confounds the comparison?

BQ2. Explain why Kai Kai’s statement about the tested pieces is descriptively correct but insufficient to establish that Material B is intrinsically more absorbent.

BQ3. Calculate the normalised mean retention per 100 square centimetres for each material. Why should this be labelled a derived comparison rather than treated as a substitute for a proper equal-area experiment?

BQ4. Redesign the test as a fairer comparison. State four controlled variables and the operational definition of absorbency.

BQ5. Write a transparent report sentence that preserves both the original total-mass result and the design flaw.

Worked answers B

BQ1. The intended independent variable is material type. The dependent variable is increase in sample mass after the defined contact and draining procedure. Sample area is the confounding factor because it changes systematically with material type: every B sample has twice the area of every A sample. The design therefore cannot attribute the larger total retained mass solely to material type.

BQ2. B’s tested pieces do have a larger mean mass increase: 12.4 grams compared with 8.4. That statement describes the actual observations. The stronger claim “B is more absorbent” treats the difference as a material property while ignoring that the B pieces provide twice the area for retaining water. The experiment changes material and area together, leaving an alternative explanation. Good scientific writing preserves the valid descriptive result while withholding the unsupported single-cause attribution.

BQ3. A gives 8.4 grams per 100 cm². B gives 12.4/200 × 100 = 6.2 grams per 100 cm². The normalised figures suggest A retains more mass per unit area in these supplied results. However, the calculation assumes a simple proportional relationship between area and retained mass and uses data from unequal sample geometries. A direct equal-area retest would provide stronger evidence because the comparison would be built into the design rather than repaired statistically afterward.

BQ4. Use equal-area and equal-shape samples for A and B, keeping thickness, contact duration, draining duration, starting condition and balance procedure the same. Material type remains the independent variable. Absorbency can be operationally defined as the increase in mass in grams after the exact contact and draining procedure for a 100 cm² sample. Repeating several trials for each material would allow variation to be inspected rather than relying on one piece.

BQ5. A suitable sentence is: “Under the original procedure, the larger Material B pieces retained a mean of 12.4 grams compared with 8.4 grams for Material A, but the B samples had twice the area. The total-mass comparison is therefore confounded by sample area and does not isolate material type.” The statement does not erase the result; it explains exactly why the intended conclusion is not yet substantiated.

Transfer B. Imagine a new fair test uses equal 100 cm² pieces and obtains means of 8.3 grams for A and 6.4 for B with closely grouped trials. That new experiment directly supports a material difference under its procedure and is broadly consistent with the earlier per-area calculation. The earlier normalisation becomes supporting context, not a substitute for the new controlled comparison. Explain why agreement between the two analyses strengthens confidence without making the claim universal for every thickness or liquid.

Laboratory C: More insulation layers and the disappearing improvement

C1 — question and model. A fictional investigation asks how the number of identical paper insulation layers affects temperature decrease in a small model container over ten minutes. Four conditions use zero, one, two and three layers. Starting temperature, container, room, measurement duration, liquid volume in the sealed model and thermometer procedure are kept the same.

C2 — mean temperature decreases. The supplied condition means are: zero layers, 8.0°C decrease; one layer, 6.1°C; two layers, 4.7°C; three layers, 4.6°C. Each mean comes from five valid fictional trials. Smaller decrease means the model retained more heat over the ten-minute interval.

C3 — pattern. Adding the first layer reduces mean temperature decrease by 1.9°C relative to zero layers. Adding the second reduces it by a further 1.4°C relative to one layer. Adding the third changes the mean by only 0.1°C relative to two layers. The relationship shows diminishing additional change across the tested range.

C4 — competing headlines. Kai Kai writes, “Every added layer always makes insulation much better.” Alicia writes, “Temperature loss decreased as layers increased, but the difference between two and three layers was very small.” Tricia asks whether the data establish a plateau and whether any practical recommendation can be made without information about mass, cost, durability or measurement uncertainty.

C5 — scope. The model tests paper layers over one ten-minute duration in one container design. It does not test buildings, clothing or every insulation material. The data do not include a four-layer condition. No formal statistical significance test is supplied. The exercise therefore focuses on descriptive pattern, fair comparison and calibrated language.

Questions C

CQ1. Identify the independent and dependent variables and explain why four levels provide more information about pattern than only zero and three layers.

CQ2. Calculate the successive changes in mean temperature decrease. Which added layer produces the smallest additional change?

CQ3. Evaluate Kai Kai’s use of every, always and much. Which parts exceed the evidence?

CQ4. Write a conclusion about the tested range and a separate implication for a possible follow-up investigation.

CQ5. Why is “two layers are best” not yet a complete engineering recommendation even if two and three have nearly equal temperature results?

Worked answers C

CQ1. The independent variable is the number of paper insulation layers. The dependent variable is temperature decrease in degrees Celsius over the defined ten-minute period. Four levels show the shape of the relationship more clearly than two endpoints: the learner can see that the first two added layers produce substantial reductions while the change from two to three is much smaller. With only zero and three layers, that diminishing increment would remain hidden.

CQ2. Moving from zero to one layer reduces the mean decrease by 1.9°C. One to two reduces it by another 1.4°C. Two to three reduces it by 0.1°C. The third added layer therefore produces the smallest additional change in this dataset. That result is descriptive; the question of whether 0.1°C is distinguishable from measurement variation requires the trial-level data or uncertainty information not supplied in the summary.

CQ3. “Every” is unsupported as a universal statement because only three additions are tested. “Always” extends the conclusion to untested materials, durations and layer counts. “Much” is especially poor for the two-to-three comparison because the mean changes only 0.1°C. A better sentence is: “Across zero to three layers in this model, mean temperature decrease became smaller as layers increased, but the added change from two to three layers was slight.”

CQ4. Conclusion: “Within the tested model and ten-minute interval, additional paper layers were associated with smaller mean temperature decreases, with diminishing additional improvement across the tested range.” Follow-up implication: “A useful next investigation would repeat the two- and three-layer conditions with more trials and add a four-layer condition to examine whether the apparent levelling continues.” The implication proposes a test instead of announcing its future result.

CQ5. An engineering recommendation needs criteria beyond one temperature measure. Two layers may use less material than three while performing similarly, but the packet gives no mass, cost, structural durability, environmental impact or allowed tolerance for temperature change. “Best” depends on the objective and criteria. The experiment informs one criterion—temperature retention—without settling the full design decision.

Transfer C. Add a fictional four-layer mean decrease of 4.5°C under the same procedure. The sequence becomes 8.0, 6.1, 4.7, 4.6 and 4.5. The descriptive evidence for diminishing improvement becomes stronger across the extended range, but the learner still should not extrapolate indefinitely or claim a theoretical minimum without a model. Explain why an additional point can strengthen a pattern while leaving the mechanism and ultimate limit unresolved.

Laboratory D: When a model predicts well—until it does not

D1 — the model. A fictional mathematical model predicts mean descent time for one paper-spinner design from blade length. For eight-centimetre blades it predicts 1.20 seconds, for twelve centimetres 1.50 seconds, for sixteen centimetres 1.80 seconds and for twenty centimetres 2.10 seconds. The model is a simplified representation built from earlier measurements between eight and sixteen centimetres.

D2 — new observations. A later carefully standardised investigation obtains mean descent times of 1.23, 1.49, 1.80 and 1.92 seconds for the same four blade lengths. Each mean comes from five fictional trials using the same paper, clip, release height and timing protocol. The 8–16 cm results agree closely with the model, while the 20 cm result is 0.18 seconds below the prediction.

D3 — first interpretation. Alicia says the model works perfectly because three points are close. Kai Kai says the model is useless because one prediction misses. Tricia argues that both statements ignore where the model was originally supported and where it was extrapolated. The twenty-centimetre prediction lies beyond the earlier eight-to-sixteen-centimetre range used to build the simple relationship.

D4 — assumption. The model assumes the effect of each additional four centimetres of blade length continues at the same rate. The new twenty-centimetre result suggests that assumption may not hold across the extended range. The packet does not identify the mechanism of the deviation. It could reflect a nonlinear relationship, another unmodelled effect or measurement variation.

D5 — revision decision. The group wants to retain the useful predictive behaviour inside the earlier range while revising the model for larger blades. The exercise does not require fitting a new equation. Its purpose is to distinguish validation within a tested range from extrapolation beyond it and to show why a model can be useful without being universally correct.

Questions D

DQ1. Calculate the prediction errors for the four blade lengths as observed minus predicted. Which condition differs most?

DQ2. Explain why agreement from eight to sixteen centimetres supports the model within that range without proving its twenty-centimetre prediction.

DQ3. Identify the model assumption that the twenty-centimetre result challenges. Does the new observation by itself establish the mechanism causing the deviation?

DQ4. Evaluate Alicia’s “perfect” and Kai Kai’s “useless” descriptions. Write a more calibrated evaluation.

DQ5. Propose one follow-up measurement condition that would help locate where the relationship begins to depart from the simple model.

Worked answers D

DQ1. At eight centimetres, observed minus predicted is +0.03 seconds. At twelve it is −0.01. At sixteen it is 0.00. At twenty it is −0.18. The twenty-centimetre condition has the largest absolute discrepancy. The calculation identifies where the model fits less well; it does not explain why the mismatch occurs.

DQ2. The model was built from behaviour in the eight-to-sixteen-centimetre range, and the new observations in that same range agree closely with its predictions. That provides evidence that the simple relationship is useful there under comparable conditions. Twenty centimetres lies outside that original range. Predicting there is extrapolation, so the model depends on an additional assumption that the relationship continues unchanged beyond the tested interval.

DQ3. The challenged assumption is that each additional four centimetres produces the same increase in mean descent time across the extended range. The twenty-centimetre observation does not identify why that assumption fails. It could reflect a curved physical relationship or another omitted feature. The appropriate conclusion is that the simple continuation is less accurate there, not that one particular new mechanism has been proved.

DQ4. Alicia’s perfect is too strong because the twenty-centimetre prediction differs meaningfully from the observation and no model is established as universally exact. Kai Kai’s useless ignores the close agreement within the original tested range. A better evaluation is: “The model predicts the 8–16 cm conditions closely but overestimates the observed mean at 20 cm, so its simple linear extension is less suitable beyond the range used to construct it.”

DQ5. A sensible follow-up could test an eighteen-centimetre condition with the same protocol. That point lies between sixteen and twenty and can help show whether the departure appears gradually. Another follow-up could repeat the twenty-centimetre condition with more trials to check its stability. Each addresses a different question: location of the change versus reliability of the surprising endpoint.

Transfer D. Suppose the eighteen-centimetre observed mean is 1.88 seconds while the simple model predicts 1.95. Together with the 20 cm mismatch, the evidence suggests curvature beginning beyond sixteen centimetres. The group can revise the model class without claiming to know the final equation. Explain why a useful scientific response to model failure is often revision rather than choosing between total acceptance and total rejection.

Laboratory E: Light, water and the experiment that changed two things

E1 — intended question. A fictional plant-model investigation is described only through supplied records; it is not a set of growing instructions. The question asks whether light exposure affects stem growth over fourteen days. Group Bright receives a higher light condition. Group Shade receives a lower light condition. The plants are otherwise intended to be comparable.

E2 — hidden procedural difference. The method notes reveal that Group Bright also receives twenty millilitres of water per day while Group Shade receives ten millilitres. Pot size, starting plant height, plant type and observation duration are comparable. The final mean stem increases are twelve centimetres for Bright and seven centimetres for Shade.

E3 — initial report. Kai Kai writes, “More light caused five centimetres of extra growth.” Alicia identifies water amount as a second changing factor. Tricia notes that the five-centimetre difference is still a valid descriptive result between the two treatment combinations, but it cannot be attributed solely to light.

E4 — redesign. A follow-up design keeps daily water amount identical between the light conditions while preserving the same pot size, plant type, starting-height range and measurement period. The group may choose twenty millilitres for both or another school-approved amount, but the important design principle is that water no longer changes systematically with light.

E5 — separate causal questions. The original records can still support the statement that the combined Bright-plus-more-water condition had greater mean growth than the Shade-plus-less-water condition. They do not reveal whether light, water or their combination produced the difference. The redesigned experiment aims to isolate light; another experiment would be needed to isolate water.

Questions E

EQ1. Identify the intended independent variable, dependent variable and confounding variable.

EQ2. Explain why the five-centimetre difference is a valid result but not valid evidence for Kai Kai’s single-cause attribution.

EQ3. Write a transparent conclusion for the original design that preserves what can be said.

EQ4. Explain how the redesign isolates light better. Which statement about water should remain controlled rather than treated as another experimental outcome?

EQ5. If the redesigned experiment still finds greater mean growth in the higher-light condition, what stronger conclusion becomes available and what limitations remain?

Worked answers E

EQ1. Light exposure is the intended independent variable. Stem growth over fourteen days is the dependent variable. Daily water amount is a confounding variable because it changes systematically with light condition: Bright receives twice the water of Shade. The design therefore contains two potential causes changing together.

EQ2. The measured difference of twelve minus seven equals five centimetres, so the combined conditions genuinely differ by that amount in the supplied record. The problem occurs only when the writer attributes the difference to light alone. Because water also differs, both variables and their possible interaction remain alternative explanations. A correct numerical result can therefore sit inside an invalid causal sentence.

EQ3. A transparent conclusion is: “Plants in the higher-light, higher-water condition had a mean stem increase five centimetres greater than plants in the lower-light, lower-water condition. Because both light exposure and daily water amount differed, the investigation does not isolate the effect of light.” This preserves the observation, names the confound and avoids pretending the dataset is useless.

EQ4. In the redesigned comparison, water amount is held constant while light exposure changes, so differences in growth can be attributed more specifically to the light contrast if other important factors remain comparable. Water belongs in the controlled-variable description: “Both conditions received the same daily water amount under the approved procedure.” The experiment is not testing the effect of water in this version.

EQ5. A repeated difference under the redesigned fairer test would provide stronger evidence that light exposure contributes to the growth difference under those tested conditions. The result would still be limited to the plant type, light levels, water level, duration and other conditions used. It would not establish the same effect for every plant species or every possible light intensity, and it would not reveal every mechanism involved in growth.

Transfer E. Imagine the redesigned test shows nearly equal mean growth between the two light conditions. That result would weaken the original light-only interpretation and increase interest in water amount or interaction as explanations of the first difference. It still would not prove water was the sole cause. Explain why one cleaner follow-up can change the balance of explanations without fully resolving every possible factor.

Laboratory F: Two groups, one spinner and two different stopping rules

F1 — replication attempt. Group North publishes a fictional paper-spinner method and reports a mean descent time of 1.80 seconds for a sixteen-centimetre blade condition. Five trials are 1.78, 1.82, 1.79, 1.81 and 1.80 seconds. Group South tries to reproduce the result using the same template, paper, clip, release height and timer model.

F2 — South’s result. South reports five times of 2.04, 2.07, 2.05, 2.03 and 2.06 seconds, with a mean of 2.05 seconds. Its values are closely grouped, so the result is repeatable within that group but substantially higher than North’s mean. Kai Kai says the experiment cannot be reproduced.

F3 — provenance check. Tricia asks both groups for their exact timing protocol. North stops timing when the spinner’s lowest edge first touches the floor. South stops timing only when the spinner has finished bouncing and lies still. The groups had copied the same phrase “time until landing” but interpreted landing differently.

F4 — standardised rerun. South repeats the five fictional trials using North’s first-touch stopping rule and obtains 1.80, 1.83, 1.79, 1.82 and 1.81 seconds, mean 1.81. The revised result is close to North’s 1.80. The earlier 2.05 mean remains an accurate record of South’s different operational definition; it should not be overwritten.

F5 — lesson. The case demonstrates why method provenance and operational definitions matter to reproducibility. Both groups had high within-group repeatability, yet their original results disagreed because they were measuring different endpoints. Standardising the definition made the measurements comparable.

Questions F

FQ1. Compare repeatability within North and South with reproducibility between their original results.

FQ2. Identify the operational-definition difference and explain why it creates a systematic rather than random shift.

FQ3. Why should South’s original 2.05-second result remain in the record after the protocol issue is discovered?

FQ4. What does the standardised rerun show, and what does it still not prove about every future group?

FQ5. Write a short method note that would make the stopping rule reproducible for a third group.

Worked answers F

FQ1. North’s five values are tightly clustered around 1.80 and South’s original five around 2.05, so each group shows good repeatability under its own procedure. The original between-group reproducibility is poor because their means differ by 0.25 seconds. The key insight is that consistency within two groups does not guarantee they measured the same operational outcome.

FQ2. North measures until first floor contact, while South includes the additional bouncing and settling time. South’s endpoint is consistently later, so the difference pushes its times upward in one direction. That creates a systematic method difference rather than unpredictable trial-to-trial scatter. Both groups can therefore be precise about different quantities.

FQ3. The 2.05-second mean accurately summarises South’s trials under its original “fully at rest” rule. Deleting it would erase evidence of how the disagreement arose. A transparent record keeps the original data, labels the method difference and adds the standardised rerun. Scientific correction should preserve history rather than make it appear that the first procedure never existed.

FQ4. South’s rerun mean of 1.81 seconds is close to North’s 1.80 and supports the explanation that the stopping-rule difference caused most of the original discrepancy. It also shows that South can obtain a similar result when using the same operational definition. It does not guarantee that every future group will produce exactly 1.80 or 1.81; measurement variation and other conditions remain possible.

FQ5. A useful note is: “Start the timer at the instant the spinner is released without a push. Stop it when the spinner’s lowest edge first touches the floor; do not include subsequent bouncing or settling time. Record the time in seconds using the same timer model.” The definition removes the ambiguous word landing and identifies the actual observed events that bound the measurement.

Transfer F. A third group follows the standardised first-touch rule but obtains a stable mean of 1.94 seconds. That would reopen the investigation: perhaps another controlled variable differs, the template dimensions were copied differently or an instrument offset exists. Do not immediately blame poor technique. Use provenance, apparatus records and method checks to locate which conditions are actually comparable before choosing an explanation.

What the six laboratories teach together

Laboratory A shows how a clean fair test can support a bounded causal conclusion. B shows how a larger sample can create a misleading material ranking when area changes with material. C shows why several levels reveal the shape of a relationship and why “more is always better” can exceed the data. D separates model validation from extrapolation. E shows how a correct difference becomes an incorrect single-cause explanation when two variables change. F shows that repeatability inside groups cannot substitute for a shared operational definition across groups.

The common skill is not memorising which fictional condition wins. It is protecting the identity of the comparison. Ask what changed, what was measured, what remained comparable, what the values actually show and what sentence can be defended. Once those relationships are stable, advanced vocabulary becomes useful compression. Without them, the same terminology can make a weak conclusion sound more scientific than it really is.

Original story: The Clip That Moved

The following story is fictional. Its paper spinners, class display, measurements and dialogue were created for this vocabulary manual. The scene is about scientific judgement and reporting, not a record of real pupils or a claim about an actual school experiment.

Kai Kai had already written the heading in black marker when Alicia found the fourth spinner.

“LONGER BLADES WIN EVERY TIME,” the poster said. The words filled the top strip so neatly that changing them would mean cutting a fresh sheet. Beneath the heading, three columns waited for the table, graph and conclusion. The class exhibition opened after lunch.

Alicia was sorting the paper spinners into labelled envelopes: eight centimetres, twelve centimetres, sixteen centimetres. Each design had been timed five times the previous afternoon. The values formed the pattern they had expected. Longer blades, greater mean descent time. Tricia had checked the arithmetic twice. Kai Kai had made a graph whose points climbed almost as smoothly as the staircase outside the room.

The fourth spinner was not part of the final dataset. It was a spare sixteen-centimetre copy used during practice. Alicia turned it over and saw the paper clip at the lower corner instead of the centre fold.

She reached into the sixteen-centimetre envelope. The first spinner’s clip sat at the centre. The second did too. The third did not.

“Tricia.”

Tricia looked up from the graph. Alicia placed the third spinner beside the spare. Both clips were near the corner.

“Were these used?” Tricia asked.

Alicia opened the trial notebook. The sheets had been labelled by number, not by clip position. Spinner 16-C appeared in all five trials for the sixteen-centimetre condition. The notebook said nothing about the clip being moved.

Kai Kai read the note over their shoulders. “But the times are good.”

“Good for what?” Alicia asked.

He pointed to the graph. “For the pattern.”

Tricia put one finger on the sixteen-centimetre point. “We said the only important thing changing was blade length.”

Kai Kai lifted the spinner and let it hang from the clip. It tilted. “Maybe the clip doesn’t matter.”

“Maybe,” Alicia said.

That word sat between them like a fourth column on the poster.

The class had not tested clip position. They did not know whether moving the small metal mass changed stability, rotation or descent time. It might have made almost no difference. It might have mattered enough to shift the result. The problem was not that they had discovered the true cause of their pattern. They had discovered that the method no longer isolated the cause they wanted to claim.

Kai Kai read the heading again. “Longer blades still had the longest times.”

“In our data,” Tricia said.

“So the result is real.”

“The timings are real,” Alicia replied. “The explanation is the problem.”

They opened the notebook at the original planning page. Controlled variables: paper type, paper size except blade length, release height, clip mass, clip position, release method. Clip position was there, underlined in blue. Somewhere between planning and construction, one of the control rules had been lost.

Kai Kai stopped defending the heading. “Do we throw away the sixteen-centimetre result?”

Tricia shook her head. “We record what happened.”

“And then?”

Alicia turned the poster over and wrote a new sentence on the back: “The original sixteen-centimetre condition included a spinner with a different clip position, so the comparison did not fully control mass placement.”

Kai Kai frowned. “That sounds like we failed.”

“It sounds like we found a limitation.”

Tricia drew a line under the sentence. “And we can test it.”

They had enough time for a short follow-up using supplied spare templates. Instead of pretending the original condition had never happened, they created a clean sixteen-centimetre set with every clip centred according to the written rule. The follow-up times remained longer than the twelve-centimetre condition, though the mean shifted slightly from the first version.

The pattern survived the correction. That made Kai Kai smile for half a second, then he looked back at the poster.

“Can I keep every time?”

“No,” Alicia said. “Not every time. We didn’t test every time.”

They replaced the heading with a smaller one: “Longer Blades Increased Mean Descent Time in Our Controlled Follow-Up.” It took two lines. Kai Kai disliked the empty space at the right, so he placed a short note there explaining the original clip-position problem and linking the old trials to the revised procedure.

The graph now showed two sixteen-centimetre points: original procedure and controlled follow-up. The first was not hidden. The second was not presented as if it had always been there. A visitor could see why the group had changed the method and what the change did to the result.

Just before lunch, the teacher stopped at their table. She read the new heading, then the method note.

“Which result do you trust?” she asked.

Kai Kai looked at the two points. Yesterday he would have answered with the one that made the cleanest line.

“The follow-up isolates blade length better,” he said. “But the original tells us why we changed the procedure.”

“And what did the moved clip prove?”

He nearly said that it changed the fall. Instead he stopped.

“It proved our first comparison had another changing factor. It didn’t prove what that factor did.”

Alicia closed the notebook. That was the sentence she wanted on the poster, but there was no space left. It did not matter. The sentence had moved somewhere more useful: into the way they were now reading their own evidence.

Reading workshop: evidence, motive and revision in the story

Question 1. Why is Kai Kai’s first defence—“the times are good”—insufficient? Worked answer: The recorded timings can be internally consistent while the causal comparison is confounded. The issue is not whether the numbers look smooth; it is whether blade length is sufficiently isolated from clip position. Good-looking data do not repair a design problem.

Question 2. What does the moved clip establish, and what does it not establish? Worked answer: It establishes that a planned controlled variable was not kept the same in the original sixteen-centimetre condition. That weakens attribution of the result solely to blade length. It does not establish that the moved clip definitely increased or decreased descent time. Its actual effect remains a separate question.

Question 3. Why do Alicia and Tricia keep the original trials instead of deleting them? Worked answer: The original measurements are part of the investigation’s provenance. Keeping them makes the revision transparent and shows why the method changed. Deleting the earlier record would erase evidence of the design problem and make the final report look cleaner than the actual process was.

Question 4. How does the new heading improve scientific communication? Worked answer: It changes a universal claim—“every time”—into a bounded result about the group’s controlled follow-up. It also names the measured outcome as mean descent time rather than the vague word win. The longer heading is scientifically stronger because its nouns, scope and evidence status are clearer.

Question 5. Does the story portray revision as evidence of failure? Worked answer: No. The characters initially worry that admitting a flaw will weaken the project, but the final process shows revision as part of improving the design. They preserve the old record, isolate the problem better and update the conclusion. Scientific strength comes partly from responding honestly when a method no longer supports the intended claim.

Writing transfer. Rewrite the final conversation from the teacher’s viewpoint without stating the pupils’ private thoughts. The teacher can observe the changed heading, two sixteen-centimetre points and Kai Kai’s answers. Keep the experimental facts unchanged. This task combines literary viewpoint control with scientific evidence control: the narrator should not claim access to information that has not been shown.

Scientific writing clinic: claim, evidence and reasoning without overclaiming

A strong scientific paragraph does not begin by searching for the longest vocabulary word. It begins by identifying the question and the exact result that answers it. Then it selects evidence, explains why that evidence matters and keeps the conclusion inside the conditions that were tested. Technical terms are useful when they preserve those relationships.

Build the claim from the variable relationship

“The sixteen-centimetre spinner was better” is too vague because better has no criterion. “The sixteen-centimetre condition had a greater mean descent time than the eight-centimetre condition” names the measured quantity. If the design was fair, the writer can then add that increasing blade length was associated with, or supported as a cause of, greater descent time under the tested conditions. The verb should follow the design strength.

A claim about a controlled experiment can be stronger than a claim from an observational comparison, but it still needs scope. “Blade length affected mean descent time in this paper-spinner model” is narrower than “blade length controls how everything falls.” A scientific sentence is not weaker because it names its model. The explicit boundary makes it possible for another reader to know exactly what was established.

Select evidence that actually bears on the claim

If the claim concerns mean descent time, relevant evidence includes the condition means and the repeated trial values that show their consistency. The colour of the paper or who drew the graph is not supporting evidence unless those details affect the measurement or question. A long results paragraph can still be weak when it lists many facts but never identifies the comparison that matters.

Evidence should remain traceable to the raw record. When a mean is reported, keep the original trial values available. When an anomalous value is excluded, state the rule. When a procedure changes after a pilot, distinguish the old and new datasets. These practices make the final claim easier to audit and reduce the temptation to remember only the results that fit the expected story.

Reasoning explains why the evidence supports the claim

Reasoning is not another name for the conclusion. In a fair-test spinner example, the reasoning can state that blade length was the intended changing factor while release height, paper, clip and timing method were controlled, so a consistent difference in descent time can be attributed more specifically to blade length than in a confounded comparison. That reasoning uses the design to connect the numbers to the causal claim.

If a mechanism is proposed, label its evidence status. “Greater blade area may alter interaction with the surrounding air” is an explanatory model. Unless the experiment measures the forces directly, do not write that the force change was observed merely because descent time changed. A good CER paragraph can include a mechanism while still distinguishing measured evidence from theoretical explanation.

Protect the small words that change the science

“Increased to 1.80 seconds” gives an endpoint; “increased by 0.57 seconds” gives a difference; “increased from 1.23 to 1.80” gives both. “Support” is not “prove”. “Associated with” is not automatically “caused”. “Under these conditions” is not decoration: it names the scope of the evidence. Many of the most consequential improvements in scientific writing come from ordinary prepositions and qualifiers.

Twelve precision repairs with explained answers

Each first sentence is an original practice error. Rewrite it before reading the model repair. More than one version can be acceptable if the evidence relationship is preserved.

1. “The hypothesis was correct because the graph went up.” Repair: “The increasing pattern supports the hypothesis within the tested conditions, provided the comparison was fairly controlled.” A graph shape is evidence only in relation to the method and variables. The word correct overstates what one investigation establishes.

2. “We controlled the independent variable.” Repair: “We deliberately changed the independent variable and kept the relevant controlled variables constant.” In a fair-test design, the independent variable is the intended difference. Calling it controlled can reverse the roles of the variables.

3. “Five trials prove the measurement is accurate.” Repair: “Five closely grouped trials can show repeatability, but accuracy needs an appropriate reference or calibration check.” Repetition helps reveal random variation; it does not automatically detect a systematic offset.

4. “The outlier is wrong, so we removed it.” Repair: “The unusual value was checked against the source record and predefined validity rule before any exclusion decision.” Outlier status describes extremeness, not falsity. A confirmed valid extreme result may be scientifically important.

5. “The material retained more water, so it is more absorbent.” Repair for Laboratory B: “The larger tested pieces retained more total water, but unequal sample area confounds the material comparison.” The stronger property claim requires an equal-area design or another justified operational comparison.

6. “The long blades caused slower falling in all spinners.” Repair: “In this controlled paper-spinner model, the longer-blade condition had a greater mean descent time.” A causal interpretation may be reasonable within the fair test, but all spinners generalises beyond the tested design and materials.

7. “The experiment was unreliable because the result changed after we fixed the method.” Repair: “The revised method controlled an additional factor, so the new result belongs to a better-isolated comparison.” A changed result after a justified design correction can improve validity rather than demonstrate that revision is unscientific.

8. “South could not reproduce North’s result because its mean was 2.05 seconds.” Repair: “The groups initially used different stopping rules, so the original means measured different operational endpoints.” Reproducibility can be assessed fairly only after the critical method definitions are made comparable.

9. “The model is false because one prediction missed.” Repair: “The model fitted the original range closely but overestimated the extended condition, indicating that its simple extrapolation is limited.” A model can remain useful within one range while needing revision elsewhere.

10. “No significant difference means the conditions are identical.” Repair: “The supplied evidence does not establish a clear difference under the stated analysis; it does not prove the conditions are identical.” Absence of convincing evidence for a difference is not automatically evidence of exact equality. In this manual, formal significance testing should only be discussed when an appropriate statistical method is actually supplied.

11. “The control group is the thing we keep the same.” Repair: “Controlled variables are factors kept constant; a control condition is a baseline comparison when the design needs one.” The similar wording hides two different experimental roles.

12. “The conclusion is weak because it says ‘under these conditions’.” Repair: “The phrase ‘under these conditions’ makes the conclusion appropriately scoped.” Scientific strength comes from fit between evidence and claim, not from pretending the tested conditions do not matter.

A six-session teaching sequence

This is a suggested sequence, not a validated fixed programme or a promise of improvement after six lessons. A session can be divided or repeated. Choose a small active vocabulary set that addresses the learner’s present reasoning difficulty, and record whether success was independent or supported.

Session 1: question, variable and operational definition

Use entries 1–20 and Laboratory A. Ask the learner to turn a vague question into a measurable one, then identify independent, dependent and controlled variables. Finish by defining one outcome operationally. Do not begin with memorising abbreviations such as IV and DV if the learner cannot yet explain what is changed and measured in ordinary English.

Session 2: procedure, measurement and fair-test design

Use entries 21–40 and Laboratory B. Compare the original unequal-area material test with the redesigned equal-area version. Ask which control matters and why. Then distinguish repeatability, accuracy, validity and reliability with one concrete example each. The learner should be able to say why repeating the wrong comparison does not make it fair.

Session 3: results, anomalies and uncertainty

Use entries 41–50 and Laboratory C. Move from raw values to a mean, then back to the raw values. Ask what a mean hides. Introduce anomaly and outlier through verification rather than automatic deletion. Write one secure result and one remaining uncertainty. The lesson is complete when the learner can retain confidence in the measured pattern without claiming perfect certainty.

Session 4: association, causation and alternative explanations

Use entries 51–60 and Laboratory E. Have the learner mark every factor that changes between Bright and Shade. Ask why the five-centimetre difference remains real while the light-only cause is unsupported. Redesign the test to isolate one variable. Then ask what stronger conclusion becomes possible and which generalisations still remain outside the evidence.

Session 5: models, predictions and revision

Use entries 61–80 and Laboratory D. Compare prediction with observation across the original and extrapolated ranges. Ask the learner to explain why one mismatch can lead to model revision rather than total rejection. Finish with a claim–evidence–reasoning paragraph that clearly labels the model assumption.

Session 6: reproducibility, transparency and reporting

Use entries 81–100, Laboratory F and The Clip That Moved. Give the learner two similar datasets produced under different operational rules and ask whether they can be merged. Then write a final report that preserves method changes and old data. A partner should be able to reconstruct what was measured, which version produced which result and why the final conclusion changed.

Assess the scientific language problem, not a vague idea of ability

Record specific evidence such as “identifies the independent variable but confuses controlled variable with control condition” or “states a fair-test conclusion but generalises beyond the tested range.” These observations suggest different next lessons. A single total score can hide whether the difficulty lies in vocabulary meaning, experimental design, calculation or sentence scope.

Use three informal categories when helpful: demonstrated independently, demonstrated with support and not yet demonstrated in this task. Note the support supplied. Choosing the correct variable from a visible list and generating it independently are different performances. Both can represent useful progress when recorded accurately.

Include fresh transfer. Change one consequential condition: move the clip, alter sample area, add another model point or change the stopping rule. Ask which conclusion must change and which observations remain valid. A learner who updates only the affected statement is showing deeper control than one who memorises a complete model answer and rewrites everything.

Questions about Primary 6 experimental-design vocabulary

Must a Primary 6 learner use all 100 terms? No. This is an advanced collection. Terms such as independent variable, controlled variable, fair test, evidence and conclusion may be central in many upper-primary contexts, while provenance, reproducibility and operational definition can remain extension vocabulary. Understanding a concept in ordinary language is valuable before independent technical use.

Is there one universal scientific method with fixed steps? No. The NOAA resource used here describes one way of organising science, and modern science education also emphasises a broader set of practices such as asking questions, planning investigations, analysing data, modelling and arguing from evidence. Teach the logic of inquiry rather than treating a mnemonic as a law of science.

Does a fair test prove causation? A well-designed controlled experiment can provide strong causal evidence for the manipulated factor under the tested conditions, especially when alternative explanations are reduced. The conclusion should still reflect measurement quality, replication and scope. “Supports a causal effect in this setup” is different from “proves the same effect everywhere.”

Why include statistics words if Volume 6 already covers data? Volume 6 owns charts, tables and data interpretation. This volume uses a smaller overlap only where experimental design needs it, such as repeated trials, uncertainty and results. Its distinct ownership is the logic connecting question → variables → fair test → evidence → causal conclusion.

Can a child learn the vocabulary without doing experiments? Yes. Complete fictional packets, diagrams, videos, simulations and supplied datasets can teach much of the language and reasoning. Physical investigation adds practical skills such as handling instruments, but should be appropriate to the learner and setting. The safest adequate learning format should be preferred.

What should be corrected first in a scientific answer? Correct an error that changes the meaning: wrong variable role, wrong unit, confounded causal claim, invented mechanism or unsupported generalisation. Grammar and style matter, but an elegant sentence stating the wrong experimental relationship needs conceptual repair before polishing.

Reference notes and further reading

The NOAA/NESDIS scientific-method resource is written for Grades 5–8 and presents one accessible inquiry sequence from wondering and defining through experimentation, analysis and conclusion. It supports the broad process language in this article without implying that every scientific investigation follows one rigid order.

The NGSS Science and Engineering Practices describe planning and carrying out investigations, analysing and interpreting data, constructing explanations and engaging in argument from evidence across grade bands. Their Grades 3–5 examples explicitly connect controlled variables, observations and measurements with evidence, while middle-school examples extend investigation and modelling into more complex variable relationships.

The Science Buddies guides to variables and fair tests provide practical student-facing explanations of independent, dependent and controlled variables. They distinguish controlled variables from a control group and emphasise keeping relevant conditions comparable when testing one factor. These external resources do not endorse this article or its fictional exercises.

All six laboratories, numerical datasets, the story The Clip That Moved, precision repairs and teaching suggestions are original material created for this manual. Their conclusions follow only from the stated fictional records and assumptions. They are not observations of real students or claims about tuition effectiveness.

Continue through the Primary 6 Advanced Vocabulary Collection

Return to the Primary 6 Advanced Vocabulary Collection or the Vocabulary Article Directory. For the preceding application, open Volume 6: Charts, Tables and Data Interpretation. It develops the numerical-reading language used when experimental results move into tables, averages and graphs.

For broader Grade 6 science-investigation vocabulary across the eduKate ecosystem, use Top 100 PSLE Vocabulary List | Science Investigation & Experiment Words for Grade 6 on eduKateSG. That owner provides the broad science-investigation word list. The present eduKateSingapore volume remains the advanced experimental-design and scientific-conclusion application manual rather than duplicating that broader owner.

For other advanced applications, use Volume 1: Literary Analysis, Tone and Subtext, Volume 2: Argument, Evidence and Critical Reading, Volume 3: Explanatory Writing, Processes and Systems, Volume 4: Historical Timelines, Continuity and Change and Volume 5: Map Descriptions, Scale and Spatial Relationships.

The final principle: the conclusion inherits the design

A fair conclusion cannot be stronger than the comparison that produced it. If two relevant factors changed, the causal sentence must reflect that. If the instrument is limited, the reported precision should reflect that. If a model has only been checked inside one range, the prediction outside that range remains an extrapolation. Scientific vocabulary helps because it names these relationships explicitly.

Ask the question clearly. Define what will change and what will be measured. Keep relevant conditions comparable. Record the observations faithfully. Analyse the pattern, inspect the exceptions and preserve uncertainty where it belongs. Then write the strongest conclusion the evidence supports—and stop at the point where the evidence stops.

Continue to Volume 8: Probability, Uncertainty, Risk and Decision-Making

Continue with Primary 6 Advanced Vocabulary Collection | Probability, Uncertainty, Risk and Decision-Making for Grade 6. Volume 8 develops 100 terms for outcomes, sample spaces, theoretical and experimental probability, dependence, simulation, forecasts, risk, trade-offs and defensible decisions under uncertainty.

Use Volume 7 when the question is whether an experiment fairly isolates a variable and supports a scientific conclusion. Use Volume 8 when the outcome itself remains uncertain and the learner must explain likelihood, long-run behaviour, consequences or a choice made before the result is known.