This Top 100 Secondary 1 Vocabulary List is an advanced science vocabulary collection for early-secondary learners who need more than definitions of scientific method, hypothesis, variables, experiments, measurement, data, evidence and conclusions. The language here is built for scientific investigation, experimental design, fair testing, independent and dependent variables, controls, accuracy, precision, uncertainty, graphs, patterns, claims, evidence, models, explanations and scientific evaluation. The aim is to help students read investigations accurately, write better Science answers and explain why evidence supports one conclusion more strongly than another.

Students searching for Secondary 1 science vocabulary, Grade 7 scientific investigation vocabulary, variables and experimental design vocabulary, science data and evidence vocabulary, scientific inquiry terms, measurement and graph vocabulary or advanced middle school science words often find short glossaries. This collection goes further. Every term is taught with a meaning, a semantic boundary, natural collocations, an original example and an application move so learners can distinguish a hypothesis from a prediction, accuracy from precision, repetition from replication, correlation from causation, and data from evidence.

This article extends eduKateSingapore’s advanced Secondary 1 collection while remaining distinct from the broader Scientific Inquiry & Evidence hub, the Secondary 2 Science Vocabulary collection and the Primary 6 Advanced Experimental Design collection. For a foundation-level Grade 7 route, students can also use the eduKateSG Secondary 1 Science, Observation and Investigation vocabulary guide. Here the challenge is deeper control of scientific language, evidence and reasoning.

How Maren, Iona and Leonie Use Advanced Science Vocabulary

Maren turns vague descriptions into structured explanations: what changed, what was measured, what was controlled and what mechanism could explain the result? Iona checks whether the evidence actually supports the claim: was the comparison fair, was the sample appropriate, was the measurement reliable, and could another explanation fit the data? Leonie checks execution: can the investigation be repeated, can the procedure be followed, can the graph be read, and can the conclusion be revised when the evidence changes? They are fictional learning companions used to make the reasoning visible.

Part I — Questions, Hypotheses and Experimental Design: Words 1–25

1. Phenomenon

Meaning: an observable event, pattern or process that can be investigated scientifically. Fence: a phenomenon is what is observed or needs explaining; the explanation is a separate scientific product. Collocations: natural phenomenon, observable phenomenon, explain a phenomenon. Example: “Condensation forming on the outside of a cold container is a phenomenon that invites explanation.” Science move: describe what occurs before jumping to why it occurs.

2. Inquiry

Meaning: systematic process of asking questions, gathering evidence and developing or testing explanations. Fence: inquiry is broader than one experiment; observation, modelling, field study and comparison can all contribute. Collocations: scientific inquiry, inquiry question, inquiry process. Example: “The class used field observations and measurements rather than a controlled laboratory experiment.” Science move: match the investigative approach to the question.

3. Research Question

Meaning: focused question that guides an investigation or study. Fence: a research question names what the investigation seeks to learn; it is not itself the predicted answer. Collocations: formulate a research question, answer the research question, focused question. Example: “How does water temperature affect the time required for a sugar cube to dissolve?” Science move: make the variables, population or phenomenon clear enough to guide data collection.

4. Testable Question

Meaning: question that can be investigated using observable or measurable evidence. Fence: an interesting question is not automatically testable with the available methods, time and tools. Collocations: testable question, investigate empirically, measurable outcome. Example: “Does increasing lamp distance change the measured light intensity at the sensor?” is testable with suitable equipment. Science move: identify what evidence could answer the question before collecting anything.

5. Hypothesis

Meaning: tentative, testable explanation for a phenomenon or relationship, usually grounded in prior observations or scientific ideas. Fence: a hypothesis is not merely an “educated guess,” and not every scientific investigation requires one. Collocations: formulate a hypothesis, test a hypothesis, hypothesis supported by evidence. Example: “The learner hypothesised that warmer water would increase dissolving rate because particles move and interact more rapidly.” Science move: connect the proposed explanation to a mechanism.

6. Prediction

Meaning: statement of an expected observable outcome under specified conditions. Fence: a prediction says what is expected to happen; a hypothesis proposes why. Collocations: make a prediction, predicted outcome, prediction based on a model. Example: “If the lamp moves farther from the sensor, the measured light intensity will decrease.” Science move: derive the expected result from the explanation or model rather than guessing independently.

7. Operational Definition

Meaning: precise statement of how a variable or concept will be measured, identified or produced in an investigation. Fence: an operational definition does not claim to capture every possible meaning of a concept; it specifies how this study handles it. Collocations: operationally define, operational definition, measurement rule. Example: “Seedling growth was operationally defined as change in stem height measured in millimetres.” Science move: make abstract terms measurable enough for another investigator to follow.

8. Variable

Meaning: factor or characteristic that can take different values or conditions in an investigation. Fence: not every feature in a setup is automatically a useful variable; the factor must be relevant to the question or analysis. Collocations: identify variables, measure a variable, variable value. Example: “Temperature and dissolving time are variables in the investigation.” Science move: state what can change and how each changing quantity enters the question.

9. Independent Variable

Meaning: variable deliberately changed or selected as the input whose relationship with an outcome is investigated. Fence: “independent” does not mean unrelated to everything else; it names the explanatory or manipulated variable in the design. Collocations: manipulate the independent variable, levels of the independent variable. Example: “Water temperature was the independent variable.” Science move: state its levels or range and why those values were chosen.

10. Dependent Variable

Meaning: measured outcome expected to change in relation to the independent variable. Fence: the word does not prove a causal relationship; it identifies the response measure in the design. Collocations: measure the dependent variable, response variable, outcome measure. Example: “Time taken for the sugar cube to dissolve was the dependent variable.” Science move: define the measurement method before testing.

11. Controlled Variable

Meaning: factor deliberately kept consistent so that it does not vary systematically across experimental conditions. Fence: controlled variables are not the same as a control group. One concerns factors kept stable; the other is a comparison condition or group. Collocations: control a variable, controlled variable, hold constant. Example: “Sugar-cube mass was controlled across trials.” Science move: explain how an uncontrolled change could distort the comparison.

12. Constant

Meaning: quantity or condition treated as unchanged within a specified investigation or model. Fence: constant describes the state of staying fixed; controlled variable emphasises a deliberate design decision to keep a relevant factor fixed. Collocations: keep constant, remain constant, constant condition. Example: “The volume of water remained constant at 200 millilitres.” Science move: state the actual fixed value rather than merely saying “same.”

13. Control Group

Meaning: comparison group or condition that does not receive the experimental treatment or receives a standard condition. Fence: not every investigation requires a control group; observational studies and many physical investigations use other comparisons. Collocations: control group, comparison group, untreated control. Example: “Plants receiving ordinary water formed the control group for comparison with fertiliser treatments.” Science move: explain what alternative outcome the control helps estimate.

14. Experimental Group

Meaning: group or condition receiving the treatment or manipulated condition of interest. Fence: experimental does not mean “unreliable” or “unproven”; it identifies the group exposed to the experimental condition. Collocations: experimental group, treatment group, experimental condition. Example: “The experimental groups received different fertiliser concentrations.” Science move: specify the treatment level for each group.

15. Treatment

Meaning: condition, intervention or exposure deliberately applied to an experimental unit. Fence: in science, treatment need not mean medical care; it can be light level, nutrient concentration, material type or another manipulated condition. Collocations: treatment level, treatment condition, apply a treatment. Example: “Each seedling tray received a different salt treatment.” Science move: describe the treatment precisely enough to reproduce it.

16. Condition

Meaning: specified set of circumstances under which an observation, measurement or test occurs. Fence: condition is broader than treatment because it can include naturally occurring or comparison settings not actively imposed. Collocations: experimental condition, under controlled conditions, environmental condition. Example: “Measurements were taken under three humidity conditions.” Science move: state the circumstances that differ and those that remain the same.

17. Trial

Meaning: one execution of a measurement or experimental procedure under specified conditions. Fence: a trial is one run; repeating trials does not automatically create independent replication of the whole study. Collocations: conduct a trial, repeated trials, trial result. Example: “The team performed five trials at each temperature.” Science move: keep trial conditions comparable and record every result, not only convenient ones.

18. Repetition

Meaning: repeated measurements or trials performed within the same investigation, often by the same team and method. Fence: repetition can reveal variation and reduce the influence of chance measurement fluctuation, but it is not identical to independent replication. Collocations: repeat measurements, repeated trials, within-study repetition. Example: “The group repeated each timing measurement five times.” Science move: use repetitions to examine spread, not merely to hunt for a preferred result.

19. Replication

Meaning: independent repetition of an investigation or key method to test whether a finding can be obtained again. Fence: repeating measurements inside one experiment improves one dataset; replication tests the finding across another execution, team, sample or setting. Collocations: independent replication, replicate a study, replication attempt. Example: “Another class replicated the investigation using the same protocol.” Science move: compare whether the main pattern survives independent repetition.

20. Procedure

Meaning: ordered set of actions followed during an investigation. Fence: a procedure tells what to do; it does not explain why the design answers the research question. Collocations: follow a procedure, step-by-step procedure, procedural detail. Example: “The procedure specified when the timer began and ended.” Science move: include enough operational detail for consistent execution without burying the scientific logic.

21. Method

Meaning: overall approach used to gather or analyse evidence. Fence: method is broader than procedure; it can describe experimentation, field observation, modelling, survey measurement or another approach. Collocations: research method, measurement method, choose a method. Example: “Remote sensing was the method used because direct sampling was impractical.” Science move: justify why the chosen method can answer the question.

22. Experimental Design

Meaning: planned structure of an experiment, including variables, conditions, controls, sampling, repetition and measurement. Fence: design is the logic linking the question to evidence; it is not just the equipment layout. Collocations: experimental design, revise the design, design limitation. Example: “The design compared four temperatures with five repeated trials each.” Science move: ask whether the design isolates the relationship the conclusion later claims.

23. Fair Test

Meaning: comparison designed so that relevant conditions are controlled and the tested factor can be compared meaningfully. Fence: “fair test” is useful introductory language, but advanced inquiry includes studies where variables cannot all be controlled experimentally. Collocations: fair comparison, fair test, controlled comparison. Example: “The battery test used the same device and workload for each brand.” Science move: explain what makes the comparison fair instead of using the label alone.

24. Confounding Variable

Meaning: factor associated with both the explanatory variable and outcome that can create or distort an apparent relationship. Fence: a confounder is more specific than any uncontrolled nuisance variable; it offers an alternative pathway explaining the observed association. Collocations: potential confounder, control for confounding, confounded comparison. Example: “Sunlight exposure could confound a comparison of fertiliser and plant growth if fertilised plants also receive more light.” Science move: identify plausible alternative pathways before making a causal claim.

25. Randomisation

Meaning: use of a chance-based process to assign units, order measurements or select samples in ways intended to reduce systematic bias. Fence: random does not mean careless or patternless; the randomisation procedure should be defined. Collocations: random assignment, random order, randomised design. Example: “Plant pots were randomly assigned to treatment groups.” Science move: state what was randomised and which source of bias the step is intended to reduce.

Part II — Sampling, Measurement and Data Representation: Words 26–50

26. Sample

Meaning: subset of units, organisms, observations or material selected for study from a larger set. Fence: a sample is what is measured; it is not automatically representative of the wider population. Collocations: sample size, representative sample, collect a sample. Example: “The class measured twenty leaves from the school garden as its sample.” Science move: ask how the sample was selected before generalising.

27. Population

Meaning: complete group about which an investigation aims to draw conclusions; in biology, it can also mean organisms of one species in a defined area. Fence: population is not always “all people”; the scientific population depends on the research question. Collocations: target population, study population, population of organisms. Example: “All bean plants in the greenhouse formed the target population.” Science move: state the population before deciding whether the sample supports a wider claim.

28. Sampling

Meaning: process used to select observations, organisms or material from a population. Fence: sampling is the selection process; sample is the result of that process. Collocations: random sampling, systematic sampling, sampling method, sampling bias. Example: “The class sampled every fifth plant along the transect.” Science move: explain why the method gives suitable coverage of the population or environment.

29. Bias

Meaning: systematic tendency in selection, measurement or interpretation that shifts results in a particular direction. Fence: bias is not the same as random variation; it creates patterned distortion. Collocations: sampling bias, measurement bias, observer bias, reduce bias. Example: “Sampling only the tallest plants would bias the estimate of average height.” Science move: identify the mechanism that pushes results away from a fair representation.

30. Measurement

Meaning: process of assigning a value to a quantity using a defined method, unit and instrument where appropriate. Fence: an observation can be descriptive without being a numerical measurement. Collocations: take a measurement, measurement method, repeated measurement. Example: “Stem height was measured from the soil surface to the highest growing point.” Science move: define exactly what quantity is being measured and how.

31. Instrument

Meaning: tool or device used to observe, detect or measure a quantity. Fence: an instrument does not guarantee a good measurement; calibration, resolution and correct use still matter. Collocations: measuring instrument, scientific instrument, instrument reading. Example: “A digital balance was the instrument used to measure mass.” Science move: match the instrument to the quantity and required precision.

32. Calibration

Meaning: process of checking or adjusting an instrument against a known reference so its readings can be interpreted reliably. Fence: calibration does not remove all uncertainty; it checks the instrument against a standard. Collocations: calibrate an instrument, calibration check, calibration standard. Example: “The thermometer was checked against a known reference before use.” Science move: verify measurement tools before treating their readings as trustworthy.

33. Unit

Meaning: agreed standard quantity used to express measurement. Fence: a numerical value without its unit can be scientifically incomplete or ambiguous. Collocations: SI unit, unit of measurement, convert units. Example: “The length was 42 millimetres, not merely 42.” Science move: write the value and unit together and convert only with the correct relationship.

34. Scale

Meaning: system of marked values or range used for measurement or graphical representation. Fence: scale can refer to an instrument’s markings or the numerical spacing on a graph; context determines the sense. Collocations: measurement scale, graph scale, scale interval, logarithmic scale. Example: “The graph used intervals of five units on the vertical axis.” Science move: check the scale before interpreting visual differences.

35. Precision

Meaning: closeness of repeated measurements to one another or fineness with which a quantity is reported, depending on context. Fence: high precision does not guarantee accuracy; repeated values can cluster tightly around the wrong value. Collocations: measurement precision, precise reading, high precision. Example: “The measurements were precise but all slightly too high.” Science move: compare spread among repeated readings separately from closeness to the accepted value.

36. Accuracy

Meaning: closeness of a measurement or result to the relevant true, accepted or reference value. Fence: accuracy concerns correctness relative to a reference, while precision concerns repeatability or resolution of measurement. Collocations: measurement accuracy, accurate result, improve accuracy. Example: “After calibration, the balance gave values closer to the reference mass.” Science move: identify what reference makes an accuracy judgment possible.

37. Resolution

Meaning: smallest change or interval an instrument can distinguish or display. Fence: finer resolution can support more detailed readings but does not automatically make the measurement accurate. Collocations: instrument resolution, high-resolution sensor, smallest division. Example: “A ruler marked every millimetre has finer resolution than one marked every centimetre.” Science move: report measurements no more finely than the instrument supports.

38. Uncertainty

Meaning: quantified or described limitation on how exactly a measured or estimated value is known. Fence: uncertainty is not ignorance or failure; every measurement has limits. Collocations: measurement uncertainty, uncertainty range, quantify uncertainty. Example: “The length was reported with an uncertainty reflecting the ruler’s resolution.” Science move: communicate how much confidence the measurement method reasonably supports.

39. Error

Meaning: difference between a measured or estimated value and a reference or underlying value, or more broadly a problem affecting measurement. Fence: error in science does not always mean a careless mistake; it can arise from unavoidable measurement limitations. Collocations: measurement error, error source, reduce error. Example: “Small timing errors occurred because human reaction time affected stopwatch use.” Science move: identify the source and likely direction or variability of the error.

40. Systematic Error

Meaning: consistent measurement distortion that shifts results in the same or a predictable direction. Fence: repeating measurements does not remove systematic error if the same flawed instrument or method is used each time. Collocations: systematic error, calibration bias, consistent offset. Example: “A balance reading 0.5 grams too high for every sample creates systematic error.” Science move: use calibration, redesign or independent comparison to detect consistent bias.

41. Random Error

Meaning: unpredictable variation causing repeated measurements to scatter around a value. Fence: random error differs from systematic error because its direction varies rather than consistently shifting the result. Collocations: random measurement error, random variation, reduce random error. Example: “Small reaction-time differences produced slightly different stopwatch readings.” Science move: repeat measurements and examine spread rather than selecting a single convenient value.

42. Data

Meaning: recorded observations, measurements or values collected during an investigation. Fence: data become evidence only when interpreted in relation to a claim or question. Collocations: collect data, analyse data, raw data, data point. Example: “The temperature readings were data; they became evidence when used to test the prediction.” Science move: preserve raw records before summarising or interpreting them.

43. Observation

Meaning: information obtained by noticing, detecting or measuring a phenomenon using senses or instruments. Fence: observation reports what is detected; inference explains what the observation may mean. Collocations: direct observation, recorded observation, observational evidence. Example: “The solution changed from colourless to blue” is an observation. Science move: separate description from interpretation.

44. Qualitative Data

Meaning: descriptive, categorical or non-numerical information about qualities, types or characteristics. Fence: qualitative does not mean unscientific; such data can be systematic and highly informative. Collocations: qualitative observation, categorical data, descriptive data. Example: “Leaf colour was classified as pale, medium or dark green using a defined reference chart.” Science move: define categories clearly so different observers can apply them consistently.

45. Quantitative Data

Meaning: numerical information representing counts or measurements. Fence: quantitative does not automatically mean higher quality; poorly measured numbers can still mislead. Collocations: quantitative measurement, numerical data, quantitative result. Example: “Each leaf’s length was recorded in millimetres.” Science move: state units, measurement method and uncertainty where relevant.

46. Dataset

Meaning: organised collection of related data gathered for analysis. Fence: a dataset can contain raw observations, derived values and metadata; it is broader than one table or graph. Collocations: analyse a dataset, complete dataset, dataset structure. Example: “The dataset contained all plant heights, treatment labels and measurement dates.” Science move: preserve enough context to know what each value represents.

47. Table

Meaning: structured arrangement of data in rows and columns. Fence: a table preserves exact values more directly than many graphs, while graphs can make patterns easier to see. Collocations: data table, table heading, tabulate results. Example: “The table listed every trial rather than only the calculated average.” Science move: include units, clear headings and logically ordered variables.

48. Graph

Meaning: visual representation showing relationships, patterns or distributions in data. Fence: a graph is an interpretation-friendly representation, not the raw dataset itself. Collocations: plot a graph, line graph, scatter plot, graph data. Example: “A scatter plot showed the relationship between temperature and dissolving time.” Science move: choose a graph type that matches the variable types and scientific question.

49. Axis

Meaning: reference line on a graph along which values or categories are plotted. Fence: the plural is axes; axis labels must identify variable and unit where relevant. Collocations: horizontal axis, vertical axis, axis label, axis scale. Example: “Temperature was plotted on the horizontal axis and dissolving time on the vertical axis.” Science move: read labels and scale before interpreting the shape.

50. Trend

Meaning: overall direction or systematic pattern of change across ordered data. Fence: a trend is broader than one pair of points and does not prove causation. Collocations: upward trend, downward trend, long-term trend, trend line. Example: “Dissolving time showed a downward trend as temperature increased.” Science move: describe direction first, then discuss strength, exceptions and possible explanations.

Part III — Patterns, Claims, Models and Explanations: Words 51–75

51. Pattern

Meaning: repeated, organised or recognisable arrangement in observations or data. Fence: a pattern can exist without a known mechanism and can arise by chance in small datasets. Collocations: identify a pattern, recurring pattern, data pattern. Example: “Plant growth increased across the first three nutrient levels and then levelled off.” Science move: describe the pattern before explaining why it may occur.

52. Outlier

Meaning: observation or value lying unusually far from the main pattern of other data. Fence: an outlier is not automatically a mistake and should not be deleted merely because it is inconvenient. Collocations: identify an outlier, extreme value, investigate an outlier. Example: “One trial took twice as long as the others and was investigated as a possible outlier.” Science move: check measurement, method and plausible biological or physical explanations before deciding how to handle it.

53. Mean

Meaning: arithmetic average found by adding values and dividing by the number of values. Fence: the mean can be strongly affected by extreme values and is not always the best summary of a distribution. Collocations: calculate the mean, mean value, arithmetic mean. Example: “The five growth measurements had a mean of 24 millimetres.” Science move: inspect the spread and outliers before treating the mean as the complete story.

54. Median

Meaning: middle value when ordered data are arranged from lowest to highest, or the average of the two middle values when the count is even. Fence: median and mean answer different summary questions. Collocations: median value, calculate the median, median response. Example: “The median was used because one extreme measurement distorted the mean.” Science move: choose the summary statistic that fits the data distribution and question.

55. Range

Meaning: span from the smallest to largest value, or the set of values considered in a study. Fence: the numerical range describes extremes but not how values are distributed between them. Collocations: range of values, measurement range, calculate the range. Example: “The trial times ranged from 28 to 34 seconds.” Science move: use range as one simple measure of spread while recognising its limits.

56. Variation

Meaning: differences among measurements, organisms, trials or observations. Fence: variation is expected in many biological and measurement contexts; it does not automatically signal error. Collocations: natural variation, measurement variation, variation among samples. Example: “Seedlings showed variation in height even under the same treatment.” Science move: ask whether variation reflects biology, environment, measurement or several sources together.

57. Distribution

Meaning: way values are spread across possible outcomes or categories. Fence: two datasets can have the same mean while having very different distributions. Collocations: data distribution, frequency distribution, distribution of values. Example: “The class compared whether heights clustered tightly or spread widely around the mean.” Science move: inspect shape and spread instead of reducing a dataset to one average.

58. Correlation

Meaning: statistical association in which two variables vary together in a systematic way. Fence: correlation does not establish that one variable causes the other. Collocations: positive correlation, negative correlation, correlation between variables. Example: “There was a negative correlation between water temperature and dissolving time.” Science move: describe the association, then investigate mechanisms and alternative explanations separately.

59. Causation

Meaning: relationship in which changing one factor produces or contributes to a change in another. Fence: sequence, association or plausible mechanism alone does not automatically establish causation. Collocations: causal relationship, establish causation, causal mechanism. Example: “A controlled design can strengthen a causal claim by reducing alternative explanations.” Science move: ask what design and evidence would distinguish cause from mere association.

60. Relationship

Meaning: systematic connection between variables, quantities or phenomena. Fence: relationship is deliberately broad; it does not specify whether the connection is causal, correlational, proportional or qualitative. Collocations: relationship between variables, linear relationship, relationship with outcome. Example: “The graph suggested a relationship between lamp distance and measured light intensity.” Science move: name the type of relationship only when the data justify it.

61. Claim

Meaning: statement offered as an answer, conclusion or explanation that can be supported or challenged using evidence. Fence: a claim is not strengthened by confidence alone. Collocations: scientific claim, support a claim, claim about a phenomenon. Example: “The student claimed that higher temperature reduced dissolving time.” Science move: state the claim precisely enough that evidence can bear on it.

62. Evidence

Meaning: data or observations interpreted as support for or against a scientific claim. Fence: data are recorded values; evidence is data used in relation to a claim. Collocations: supporting evidence, empirical evidence, evidence for a claim. Example: “The repeated timing results became evidence because they directly tested the predicted relationship.” Science move: explain why the data are relevant and sufficient for the specific claim.

63. Reasoning

Meaning: logical connection explaining why evidence supports, weakens or qualifies a claim. Fence: evidence does not interpret itself; reasoning supplies the bridge from data to conclusion. Collocations: scientific reasoning, evidence-based reasoning, reasoning from data. Example: “Because all other measured conditions were controlled, the consistent timing difference was more plausibly linked to temperature.” Science move: make the inferential bridge explicit.

64. Conclusion

Meaning: reasoned statement summarising what an investigation’s evidence supports about the question or hypothesis. Fence: a conclusion should match the evidence scope and does not need to declare the hypothesis “proven.” Collocations: draw a conclusion, evidence supports the conclusion, qualified conclusion. Example: “Within the tested range, warmer water was associated with shorter dissolving time.” Science move: include conditions and limitations where they matter.

65. Interpretation

Meaning: explanation of what observed data, patterns or results mean. Fence: interpretation is constrained by the evidence and can differ from the raw observation. Collocations: interpret results, data interpretation, alternative interpretation. Example: “The plateau was interpreted as evidence that another factor became limiting.” Science move: distinguish the observed pattern from the explanation proposed for it.

66. Analysis

Meaning: systematic examination of data or evidence to identify patterns, differences, relationships and implications. Fence: analysis is broader than calculating an average; it includes selecting relevant comparisons and testing explanations. Collocations: data analysis, analyse results, analytical method. Example: “The analysis compared means, spread and the shape of the temperature relationship.” Science move: use methods that answer the research question rather than performing calculations without purpose.

67. Explanation

Meaning: account of how or why a phenomenon occurs, supported by evidence and scientific ideas. Fence: explanation goes beyond description; it connects observations to a mechanism or model. Collocations: scientific explanation, explanatory mechanism, explain a phenomenon. Example: “The explanation linked faster dissolving to more frequent particle interactions at higher temperature.” Science move: connect evidence with an appropriate scientific principle or mechanism.

68. Model

Meaning: simplified representation of a system, process or phenomenon used to explain, predict or explore behaviour. Fence: a model is not a miniature truth; every model simplifies and has a domain of usefulness. Collocations: scientific model, model prediction, revise a model. Example: “A particle model was used to explain dissolving at different temperatures.” Science move: state what the model represents well and what it leaves out.

69. Mechanism

Meaning: process or chain of interactions through which an effect is produced. Fence: mechanism is more specific than correlation; it proposes how one state leads to another. Collocations: underlying mechanism, causal mechanism, biological mechanism. Example: “The proposed mechanism involved increased particle motion and collision frequency.” Science move: identify intermediate steps instead of jumping from cause to outcome.

70. Theory

Meaning: in science, a broad explanatory framework supported by extensive evidence and capable of generating predictions. Fence: theory does not mean an unsupported guess, and theories do not “graduate” into laws. Collocations: scientific theory, theory explains, theory predicts. Example: “Atomic theory explains a wide range of observations about matter.” Science move: distinguish broad explanatory frameworks from individual hypotheses.

71. Scientific Law

Meaning: concise statement, often mathematical, describing a consistent relationship observed under specified conditions. Fence: a scientific law describes a regularity; a theory explains broader mechanisms and relationships. One does not simply become the other. Collocations: scientific law, law describes, law applies under conditions. Example: “A law may describe how variables relate without explaining the underlying mechanism.” Science move: separate description of regularity from explanation.

72. Assumption

Meaning: condition or proposition accepted as a basis for reasoning, modelling or analysis. Fence: an assumption can be reasonable and explicit without being established as fact. Collocations: underlying assumption, modelling assumption, test an assumption. Example: “The calculation assumed that heat loss to the surroundings was negligible.” Science move: identify which conclusions would change if the assumption failed.

73. Alternative Explanation

Meaning: another plausible account capable of explaining the same observation or result. Fence: an alternative explanation does not automatically refute the original one; evidence is needed to distinguish them. Collocations: consider alternatives, alternative explanation, competing explanation. Example: “Different light exposure was an alternative explanation for the apparent fertiliser effect.” Science move: design tests or collect evidence that discriminates among competing explanations.

74. Limitation

Meaning: feature of a design, dataset, method or model that restricts the strength, scope or applicability of a conclusion. Fence: naming a limitation does not make a study useless; it calibrates what the evidence can support. Collocations: study limitation, methodological limitation, acknowledge limitations. Example: “The narrow temperature range limited generalisation beyond the tested conditions.” Science move: connect each limitation to the specific claim it weakens.

75. Validity

Meaning: degree to which a method, measure or conclusion appropriately addresses what it is intended to address. Fence: validity differs from reliability; a method can produce consistent readings while consistently measuring the wrong thing. Collocations: valid measurement, internal validity, validity of conclusion. Example: “Timing how long a plant remained green would not be a valid measure of growth rate.” Science move: ask whether the design and measure actually answer the stated question.

Part IV — Reliability, Research Practice and Scientific Knowledge: Words 76–100

76. Reliability

Meaning: degree to which a method or measurement produces consistent results under comparable conditions. Fence: reliability does not guarantee validity or accuracy; a method can be consistently wrong. Collocations: reliable measurement, test reliability, improve reliability. Example: “Repeated measurements clustered closely, suggesting good reliability.” Science move: check consistency separately from whether the right quantity is being measured.

77. Reproducibility

Meaning: ability to obtain compatible findings when an investigation or analysis is independently repeated using sufficiently described methods or data. Fence: reproducibility is broader than repeating the same measurement yourself. Collocations: reproducible result, reproducibility crisis, reproduce an analysis. Example: “Another class used the documented method and recovered the same overall relationship.” Science move: preserve methods, data handling and analysis details so others can test the finding.

78. Repeatability

Meaning: consistency obtained when measurements are repeated under the same or closely matched conditions, often by the same operator and equipment. Fence: repeatability is a narrower within-setting idea than broad reproducibility across independent settings. Collocations: repeatable measurement, test repeatability, repeatable procedure. Example: “The sensor gave nearly identical readings across repeated measurements.” Science move: examine whether the same setup produces stable values before comparing laboratories.

79. Verification

Meaning: process of checking whether a result, record, calculation or claim agrees with evidence or an independent standard. Fence: verification checks correctness or support; it does not necessarily explain why the result occurs. Collocations: verify a result, independent verification, verification step. Example: “The calculated concentration was verified using a second method.” Science move: use an independent check where possible rather than repeating the same unchecked assumption.

80. Falsifiable

Meaning: capable in principle of being shown false by some possible observation or test. Fence: falsifiable does not mean false; it means the claim risks conflict with evidence. Collocations: falsifiable hypothesis, testable prediction, falsifiability criterion. Example: “A claim that predicts no possible observation capable of counting against it is difficult to test scientifically.” Science move: ask what result would make you revise the hypothesis.

81. Peer Review

Meaning: evaluation of scientific work by other knowledgeable researchers before or after publication, depending on the process. Fence: peer review is quality control, not a guarantee that a study is correct forever. Collocations: peer-reviewed article, peer-review process, reviewer comments. Example: “Reviewers questioned whether the sample supported the paper’s broad conclusion.” Science move: treat peer review as one layer of scrutiny within a larger evidence system.

82. Citation

Meaning: reference identifying the source of a claim, quotation, method, idea or evidence. Fence: citation improves traceability but does not automatically make the cited source strong or relevant. Collocations: cite a source, scientific citation, citation trail. Example: “Iona followed the citation to inspect the original experiment rather than relying on the summary.” Science move: use citations as paths back to evidence and method.

83. Source

Meaning: origin of information, data, material or evidence. Fence: the website where information was found may not be the original scientific source. Collocations: primary source, data source, credible source, source material. Example: “The news article linked to the published study that generated the original data.” Science move: trace scientific claims back through summaries to the underlying source where practical.

84. Research

Meaning: systematic investigation intended to develop, test or refine knowledge. Fence: casual searching is not automatically scientific research; research uses defined questions, methods and evidence. Collocations: conduct research, research question, research findings. Example: “The project combined laboratory measurements with published research.” Science move: identify question, method and evidential basis before calling an activity research.

85. Scientific Literature

Meaning: body of published scientific papers, reviews, reports and related scholarly material on a topic. Fence: “the literature” does not mean one article and does not imply that every source agrees. Collocations: review the literature, scientific literature, literature search. Example: “The literature contained several studies using different measurement methods.” Science move: compare patterns across sources rather than treating one paper as the whole field.

86. Methodology

Meaning: systematic framework of methods and reasoning used to design and conduct research. Fence: methodology is not simply a synonym for one procedure. Collocations: research methodology, methodological choice, methodological limitation. Example: “The paper explained why observational sampling was more appropriate than laboratory manipulation.” Science move: connect the chosen methods to the kind of evidence the question requires.

87. Protocol

Meaning: detailed, standardised instructions for carrying out a procedure, measurement or research process. Fence: protocol emphasises consistency and standardisation more strongly than the general word method. Collocations: laboratory protocol, follow a protocol, standard protocol. Example: “Every group followed the same protocol for preparing samples.” Science move: document critical steps that must remain consistent across operators or trials.

88. Ethics

Meaning: principles guiding responsible conduct in research, including treatment of people, animals, environments, data and colleagues. Fence: ethical approval does not prove a study’s scientific conclusion; it addresses whether the research conduct meets relevant standards. Collocations: research ethics, ethical consideration, ethical approval. Example: “The student study avoided collecting unnecessary personal information.” Science move: consider welfare, privacy and fairness before data collection begins.

89. Safety

Meaning: practices and conditions intended to reduce unacceptable risk of harm during scientific work. Fence: safety is risk reduction, not a promise of zero risk. Collocations: laboratory safety, safety procedure, safety equipment. Example: “Eye protection was required because splashes were a plausible hazard.” Science move: identify hazards before selecting controls and protective measures.

90. Risk

Meaning: combination of the likelihood and potential consequence of harm or an unwanted event. Fence: risk is not identical to hazard; hazard is the source of potential harm. Collocations: assess risk, reduce risk, risk level, risk control. Example: “A corrosive chemical is a hazard; the risk depends on concentration, exposure and controls.” Science move: consider both probability and severity when deciding precautions.

91. Hazard

Meaning: source, substance, condition or process with potential to cause harm. Fence: hazard exists even when exposure is tightly controlled; risk describes the chance and consequence under actual conditions. Collocations: chemical hazard, biological hazard, identify hazards. Example: “Open flame was the hazard considered before selecting the heating method.” Science move: name the hazard before choosing controls.

92. Consent

Meaning: informed agreement to participate in research or allow specified use of information, where consent is applicable. Fence: meaningful consent requires clarity about what participation involves and may be affected by age and institutional rules. Collocations: informed consent, obtain consent, consent process. Example: “The survey explained what data would be collected before participants agreed.” Science move: separate voluntary agreement from mere attendance or silence.

93. Data Integrity

Meaning: accuracy, completeness and trustworthy preservation of data throughout collection, storage, analysis and reporting. Fence: data integrity concerns whether records remain faithful to what was observed; it is broader than obtaining a statistically attractive result. Collocations: protect data integrity, data-integrity check, complete records. Example: “All trials were recorded, including the inconvenient outlier.” Science move: preserve original records and document any justified changes.

94. Transparency

Meaning: openness about methods, data processing, assumptions, conflicts and limitations so others can understand how a result was produced. Fence: transparency supports scrutiny but does not guarantee the conclusion is correct. Collocations: methodological transparency, transparent reporting, disclose assumptions. Example: “The report stated that two failed sensor readings were excluded and explained why.” Science move: make consequential analytical decisions visible.

95. Generalisation

Meaning: extension of a finding from the studied sample or conditions to a wider population, setting or range. Fence: generalisation must be earned by sampling, design and similarity of conditions; it is not automatic. Collocations: generalise findings, generalisation beyond the sample, limited generalisability. Example: “Results from one plant species were not generalised to all plants.” Science move: state exactly how far the evidence reasonably travels.

96. Scope

Meaning: boundaries of what a study, claim, model or conclusion covers. Fence: scope describes coverage, not importance. Collocations: study scope, scope of conclusion, outside the scope. Example: “The experiment covered temperatures from 10°C to 50°C, so claims outside that range required caution.” Science move: preserve population, variable range, setting and time limits in the conclusion.

97. Confidence

Meaning: degree of justified certainty placed in a scientific result or explanation based on evidence quality and consistency. Fence: scientific confidence is not the same as personal self-confidence. Collocations: confidence in a result, increase confidence, confidence level. Example: “Replication by independent groups increased confidence in the finding.” Science move: tie confidence to evidence strength rather than tone or authority alone.

98. Revision

Meaning: modification of a model, explanation, method or conclusion in response to new evidence or better reasoning. Fence: revision is not scientific failure; it is part of how knowledge becomes more accurate. Collocations: revise a model, revise a conclusion, evidence-driven revision. Example: “The team revised its explanation after a control test contradicted the first idea.” Science move: state what evidence changed and which part of the explanation was updated.

99. Scientific Knowledge

Meaning: organised body of explanations, models, observations, methods and findings developed through scientific inquiry and evaluation. Fence: scientific knowledge is reliable without being unchangeable; strength comes partly from openness to testing and revision. Collocations: scientific knowledge, knowledge base, develop knowledge. Example: “New measurements refined rather than erased earlier scientific knowledge.” Science move: distinguish provisional revision from the idea that ‘anything goes.’

100. Scientific Consensus

Meaning: broad agreement among relevant scientific experts that emerges when multiple independent lines of evidence support a conclusion. Fence: consensus is not a vote that creates truth, and it need not mean every scientist agrees on every detail. Collocations: scientific consensus, expert consensus, consensus view. Example: “Confidence grew as independent methods converged on the same explanation.” Science move: examine the evidence network and degree of expert convergence rather than counting isolated quotations.

The 100 Words as One Scientific Reasoning System

The collection begins with phenomena and ends with scientific consensus because scientific learning moves through a connected system. Questions define what needs to be known. Designs decide what evidence can be generated. Measurements produce data with uncertainty. Analysis identifies patterns and variation. Claims require evidence and reasoning. Models and mechanisms explain. Replication, peer review and transparency test whether findings survive scrutiny. Revision allows scientific knowledge to improve rather than freeze.

Part V — Advanced Science Laboratories: Design, Measurement and Evidence

The laboratories below are fictional. They are designed to make the vocabulary operational. Each case begins with a result that can be described quickly but requires deeper reasoning before a strong scientific conclusion is justified.

Laboratory 1 — Does Warmer Water Make Sugar Dissolve Faster?

A class investigates dissolving using identical sugar cubes placed into 200 millilitres of water at 10°C, 25°C, 40°C and 55°C. Students stir each beaker and record how long the visible sugar cube takes to disappear. The first set of results suggests shorter times at higher temperatures.

Maren begins with the variables. Water temperature is the independent variable. Dissolving time is the dependent variable. Water volume and sugar-cube mass should be controlled. But “stir each beaker” is not yet precise enough. Stirring speed and pattern can change dissolving rate, so the procedure needs an operational definition.

Iona asks what counts as “dissolved.” Does timing stop when the cube is no longer visible, when no crystals remain on the bottom, or when a sensor indicates uniform concentration? Different operational definitions can produce different results even when the underlying process is similar.

Leonie examines repetition. One trial at each temperature gives four data points but little information about random variation. Five trials at each temperature allow a mean, range and outlier check. Repetition cannot remove a systematic timing error, but it helps reveal scatter.

Suppose the five times at 40°C are 42, 41, 43, 42 and 68 seconds. The final value is an outlier relative to the cluster. The correct response is not to erase 68 automatically. Check whether stirring stopped, the sugar cube differed in mass, or the timer was started late. If no explanation is found, report how the analysis handles the value.

Maren then distinguishes observation from explanation. “Dissolving time decreased as water temperature increased” describes the pattern. “Higher temperature increases particle motion and collision frequency, contributing to faster dissolving” is an explanation using a particle model. The explanation should not be smuggled into the observation sentence as though it were directly seen.

Iona checks scope. The experiment tested sugar cubes over a limited temperature range with one stirring procedure. It does not establish that every solid dissolves faster in every liquid at every temperature. Generalisation must remain inside the tested system unless other evidence extends it.

Student task: write a conclusion with claim, evidence and reasoning. Include one limitation and one design improvement. A strong answer might say that within the tested range, higher water temperature was associated with shorter dissolving time across repeated trials, while noting that hand stirring introduced uncertainty.

Laboratory lesson: a simple experiment becomes advanced when definitions, variation, controls and scope stay visible.

Laboratory 2 — Fertiliser, Sunlight and a Confounded Plant Experiment

A student places ten bean plants on a sunny windowsill and gives them fertiliser. Another ten plants receive no fertiliser and are placed on a shelf farther from the window. After three weeks, the fertilised plants are taller. The student concludes that the fertiliser caused the difference.

Iona identifies the design problem immediately: sunlight exposure differs systematically between treatment groups. Light can affect plant growth, so it is a confounding variable. The comparison cannot isolate the fertiliser effect because treatment and light change together.

Maren redesigns the study. Twenty similar seedlings are randomly assigned to fertiliser and control groups. Pots are rotated across positions to reduce location bias. Water volume, pot size, soil type and measurement schedule are controlled. The treatment is defined as a specified fertiliser concentration applied on stated days.

Leonie checks the dependent variable. “Plant growth” can mean height, leaf number, dry mass or another measure. If growth is operationally defined as change in stem height, the method should specify where the ruler begins and which growing point is measured.

The redesigned study finds mean height increases of 8.4 cm in the fertiliser group and 6.9 cm in the control group. That difference is evidence of a treatment relationship under the experimental conditions. It does not tell the student whether the same fertiliser amount works optimally for every plant species or soil.

Maren considers alternative explanations that remain. Random assignment reduces systematic pre-existing differences, but plants still vary naturally. Measurement error, uneven soil moisture or unnoticed pest damage could contribute to variation. Good design reduces alternative explanations rather than making them metaphysically impossible.

Iona asks whether twenty plants are enough for broad claims. Sample size affects confidence and precision. A larger study across different environments may be needed before generalising strongly. Secondary 1 students do not need advanced statistics to understand that ten plants in one room represent a narrower evidence base than hundreds across several conditions.

Student task: draw the original and redesigned experiment. Label the independent variable, dependent variable, controlled variables, confounder, control group and randomisation step. Then write why the redesign strengthens validity.

Laboratory lesson: a causal conclusion becomes stronger when the design breaks the link between treatment and alternative explanations.

Laboratory 3 — Pendulum Timing and the Difference Between Accuracy and Precision

A class investigates how pendulum length affects period. One group times a single swing using a handheld stopwatch. Their repeated values are 1.21, 1.19, 1.20, 1.21 and 1.20 seconds. The measurements are tightly clustered, so the students declare them accurate.

Maren challenges the word choice. Tight clustering suggests precision or repeatability. Accuracy requires comparison with an appropriate reference or better measurement method. If the stopwatch method consistently starts late and stops late in a way that shifts results, the values can be precise but inaccurate.

Leonie improves the timing method by measuring twenty swings and dividing the total time by twenty. The same reaction-time uncertainty now affects a longer interval, reducing its relative influence on the calculated period. This does not eliminate reaction time, but it improves the measurement strategy.

Iona adds a light-gate sensor as an independent instrument check. The sensor reports a period slightly lower than the stopwatch method. That pattern suggests a possible systematic difference between methods rather than random scatter alone.

The class then examines resolution. The stopwatch displays hundredths of a second, but displaying two decimal places does not guarantee that human timing is accurate to one hundredth. Instrument resolution and overall measurement uncertainty are related but not identical.

Maren asks students to report their method transparently: pendulum length, release angle, number of swings timed, instrument, number of repetitions and calculation used. Another group should be able to reproduce the method well enough to compare findings.

Student task: explain why “1.20 seconds” can look precise without being highly accurate. Then propose one way to improve repeatability and one way to check accuracy.

Laboratory lesson: decimal places are not a substitute for understanding the measurement system.

Laboratory 4 — Cooling Water and the Temptation to Overread a Graph

Students pour equal volumes of hot water into two containers, one insulated and one uninsulated. They record temperature every two minutes for twenty minutes. Both temperatures fall, but the insulated container cools more slowly.

Iona begins with the graph. Time belongs on the horizontal axis because it is the ordered explanatory variable. Temperature belongs on the vertical axis. Both axes need units. If the vertical scale begins at 60°C rather than zero, the visual difference between curves may look dramatic; the scale must be read before interpreting magnitude.

Maren distinguishes trend from mechanism. The graph shows that temperature decreases over time and that the insulated condition remains warmer in the measured period. The graph does not directly display heat transfer mechanisms. A scientific explanation may invoke energy transfer to the surroundings and the insulation reducing the rate of transfer.

Leonie notices that the initial temperatures were 91°C and 87°C rather than identical. That baseline difference complicates direct comparison. A better design would begin with closely matched temperatures or analyse temperature change relative to each starting value.

One temperature reading jumps upward between minutes 12 and 14. Before calling it “proof of reheating,” the class checks whether the thermometer touched the container wall, the water was stirred inconsistently or the reading was copied incorrectly. Outliers deserve investigation before interpretation.

Iona also checks room conditions. If one container sits near a fan, airflow becomes a confounding factor. The experimental comparison should keep relevant environmental conditions similar unless airflow itself is being tested.

Student task: write three sentences: one observation from the graph, one mechanistic explanation and one limitation. Keep the three jobs separate.

Laboratory lesson: a graph is powerful because it reveals patterns, but those patterns still need design context and reasoning.

Part V — Advanced Science Laboratories: Sampling, Correlation and Model Revision

Laboratory 5 — Quadrat Sampling in a School Field

A class wants to estimate the abundance of clover across a large school field. Counting every plant is impractical, so students use square quadrats and record clover coverage in selected locations.

Maren begins by defining the population: all clover plants in the field during the survey period. The sample is the set of quadrat observations. If students choose only the greenest patches, the sample will be biased toward high clover abundance.

Iona compares sampling methods. Random locations can reduce deliberate selection bias. Systematic sampling along a transect can reveal spatial change. Neither method is automatically superior; the method should match the ecological question and field layout.

Leonie checks the operational definition of coverage. Are students counting individual clover plants, estimating percentage cover, or recording presence/absence? A plant with spreading stems may make “individual count” difficult, so percentage cover can be a more practical measure if the method is defined consistently.

The first team uses five quadrats and estimates mean cover at 32%. The second uses twenty quadrats and estimates 27%. The two means differ, but the larger sample may provide a more stable estimate because it captures more field variation. Sample size affects uncertainty and representativeness.

Maren also examines distribution. Clover is clustered near wetter ground. A single average can hide that spatial pattern. Mapping quadrat locations can reveal whether the field contains distinct zones rather than one uniform population.

Iona checks generalisation. The survey represents one field, season and measurement method. It should not be extended automatically to every school field or all months of the year.

Student task: design a sampling plan using ten quadrats. Explain location selection, measurement definition and one limitation. Then state what population your conclusion applies to.

Laboratory lesson: sampling is scientific reasoning about what can be measured and what can fairly be inferred from it.

Laboratory 6 — The Dataset With One Spectacular Outlier

Five groups measure the mass of the same sealed object using identical digital balances. Four readings lie between 49.8 g and 50.2 g. One group records 58.7 g. The class must decide whether the fifth value should remain in the analysis.

Iona begins with data integrity: every original reading should remain recorded. Removing a value from the summary is a separate analytical decision and should never erase the fact that the measurement occurred.

Maren checks the instrument and procedure. The fifth group discovers that a plastic tray remained on the balance and had not been tared. This provides a specific methodological explanation for the extreme value. The group documents the issue and repeats the measurement correctly.

Leonie compares mean and median before and after the outlier. With the erroneous reading included, the mean shifts substantially. The median remains close to 50 g because it is less sensitive to one extreme value. This does not make median automatically superior; it shows why summary choice depends on distribution and error understanding.

The class then considers a harder version: suppose no procedural error can be found. The value remains unusual but potentially real. Scientists should not delete it merely because it disagrees with expectation. A larger set of measurements, instrument cross-check and independent replication may be needed.

Maren also distinguishes measurement error from natural variation. Measuring one sealed object should not produce biological variation in true mass over a few minutes. That context makes an extreme value more suspicious than it would be in a naturally variable population of organisms.

Student task: write two analyses—one when the tray error is documented and one when no cause is found. Explain whether the outlier should be excluded from the summary and how the decision should be reported.

Laboratory lesson: outlier handling is an evidence decision, not a cosmetic step for making graphs look tidy.

Laboratory 7 — Screen Time and Sleep: Correlation Is Not Yet Causation

A school survey finds that students reporting more evening screen time also report fewer hours of sleep. A graph shows a negative correlation. One headline says, “Screens cause sleep loss.” The survey supports an association, but the causal claim needs more work.

Maren identifies possible alternative explanations. Students with heavier homework loads might stay up later and also use screens for schoolwork. Stress could influence both screen use and sleep. Some students may use screens because they are already unable to sleep. These possibilities do not prove the screen explanation false; they show that correlation alone does not isolate cause.

Iona checks measurement validity. “Screen time” is self-reported. Does it include schoolwork, messaging and video? “Sleep” is estimated by students rather than measured directly. The operational definitions affect what the variables actually represent.

Leonie considers sampling. If only students who enjoy technology complete the survey, the sample may not represent the wider school. Response bias can also appear if students estimate screen use inaccurately.

Maren rewrites the conclusion: “In this survey, greater reported evening screen time was associated with shorter reported sleep duration. The observational design does not establish how much screen use causes the difference.” This preserves the pattern without overclaiming.

Iona asks what further evidence could help: objective device-use logs, repeated sleep measurements, clearer time ordering or an experimental design where appropriate and ethical. Each adds information about one weakness in the original survey.

Student task: list three potential confounders and one possible reverse-causation pathway. Then write an accurate headline that reports the association without making a causal claim.

Laboratory lesson: correlation can be scientifically useful while remaining insufficient for strong causal attribution.

Laboratory 8 — A Particle Model Meets Unexpected Evidence

Students use a simple particle model to explain why a gas can be compressed. They draw particles far apart with empty space between them. The model successfully explains several observations, but it does not include particle attraction, energy distribution or molecular structure.

Maren asks what the model is for. At this level, the model represents spacing and motion well enough to explain compressibility and diffusion. It is not intended to represent every property of matter. A model is judged partly by purpose.

Iona presents new evidence: under some conditions, real gases do not behave exactly as the simplest model predicts. The correct scientific response is not “the model is useless.” The model’s scope is limited, and a more detailed model may be needed when interactions become important.

Leonie distinguishes revision from replacement. Sometimes a model is extended by adding assumptions or mechanisms. Sometimes a different model is used for a different scale. Scientific knowledge can contain several models suited to different purposes.

The class also examines prediction. The particle model predicts that increasing temperature generally increases average particle motion. If an observed result conflicts, students should first check the experiment and operational definitions before immediately discarding the model.

Maren writes a model statement with three parts: what the model represents, what it predicts and where it is limited. This structure prevents diagrams from becoming decorative pictures with no explanatory role.

Iona connects revision to scientific knowledge. A model gains strength when it survives tests, explains multiple phenomena and integrates with other evidence. It remains revisable if new observations reveal systematic limitations.

Student task: choose a familiar science model—the particle model, ray model of light, simple circuit model or food-web model. Write what it represents, one prediction, one limitation and one kind of evidence that might require revision.

Laboratory lesson: models are tools for reasoning, not pictures students must defend unchanged.

What the Eight Laboratories Reveal

The eight cases share one discipline: define the question, design the comparison, measure consistently, preserve the data, analyse variation, distinguish pattern from cause, connect evidence to a claim and keep the conclusion inside the scope of the investigation. Scientific vocabulary matters because each term protects one link in that chain.

Part VI — Precision Clinics: Scientific Terms That Must Not Collapse Into One Another

Clinic 1 — Hypothesis vs Prediction

A hypothesis proposes a testable explanation; a prediction states an expected observable result if the explanation or model is correct under specified conditions. “Warmer water increases particle motion, which should speed dissolving” contains explanatory reasoning. “The 50°C trial will dissolve faster than the 20°C trial” is the predicted outcome. A prediction can be tested without being the explanation itself.

Clinic 2 — Fair Test vs Scientific Inquiry

A fair test is one useful experimental structure in which relevant conditions are controlled so a comparison can isolate a factor. Scientific inquiry is broader. Astronomers may rely on observation, ecologists may sample field populations and scientists may use models or historical records when manipulating the phenomenon is impossible or inappropriate. Do not teach science as one compulsory linear recipe.

Clinic 3 — Controlled Variable vs Control Group

A controlled variable is a factor deliberately kept consistent across conditions, such as water volume. A control group is a comparison group not receiving the treatment of interest or receiving a standard condition. One is a variable-management decision; the other is a comparison condition. An experiment can have many controlled variables and one control group—or no control group at all, depending on the question.

Clinic 4 — Accuracy vs Precision

Accuracy concerns closeness to a relevant reference or true value. Precision concerns closeness among repeated measurements or fineness of measurement, depending on context. A miscalibrated balance can return 51.0, 51.0 and 51.1 grams for a 50.0-gram reference: highly precise, poorly accurate. Students should not use the words as interchangeable compliments.

Clinic 5 — Repeatability vs Reproducibility

Repeatability asks whether closely matched repeats within the same setting give consistent results. Reproducibility asks whether compatible findings can be obtained more independently—another operator, dataset, laboratory or implementation. Repeating one measurement ten times can improve confidence in repeatability without testing whether another team can recover the same finding.

Clinic 6 — Data vs Evidence

Data are recorded observations or measurements. Evidence is data interpreted in relation to a claim. A table of temperatures and times is data. When the student uses those times to test the claim that dissolving becomes faster as temperature rises, selected aspects of the data function as evidence. This distinction forces the student to explain relevance rather than merely attach a table to a conclusion.

Clinic 7 — Observation vs Inference

An observation records what is detected: “the liquid turned blue.” An inference is a conclusion drawn from evidence: “a reaction probably produced the substance responsible for the blue colour.” Strong science writing allows readers to see where direct observation ends and interpretation begins.

Clinic 8 — Correlation vs Causation

Correlation describes association between variables. Causation claims that changing one factor produces or contributes to another change. A correlation can motivate a causal investigation and can be scientifically important, but it does not automatically eliminate confounding, reverse causation or coincidence. The stronger verb requires the stronger design.

Clinic 9 — Theory vs Scientific Law

A scientific theory is a broad explanatory framework supported by extensive evidence. A scientific law describes a regular relationship, often concisely or mathematically, under specified conditions. A theory does not become a law after receiving enough evidence. Explanation and description are different scientific jobs.

Clinic 10 — Validity vs Reliability

Validity asks whether a measure, design or conclusion addresses what it claims to address. Reliability asks whether results are consistent. A bathroom scale stuck five kilograms high can be reliable across repeated readings while inaccurate. A questionnaire can be consistently scored while still failing to measure the intended construct. Consistency is not enough if the wrong thing is being measured.

Clinic 11 — Error vs Uncertainty

Error often describes a difference from a reference or a source of measurement distortion. Uncertainty describes how exactly a value is known. An investigator may not know the exact error in a measurement but can still estimate an uncertainty range based on instrument resolution, method and repeat variation. Reporting uncertainty is not admitting that the measurement is worthless.

Clinic 12 — Peer Review vs Scientific Consensus

Peer review examines individual scientific work through expert scrutiny. Scientific consensus emerges across a wider body of evidence when relevant experts broadly converge on a conclusion. A single peer-reviewed paper does not create consensus, and consensus does not mean every expert agrees on every detail. The scale of evidence differs.

A 30-Day Secondary 1 Advanced Science Vocabulary Route

This is a flexible teaching route, not a promise that thirty days automatically produces mastery. Each day combines retrieval with a scientific reasoning task. Learners who need more time should repeat the weak distinction rather than rush to the next set.

Days 1–5 — Questions and Experimental Structure

Day 1: retrieve phenomenon, inquiry, research question, testable question and hypothesis. Turn three vague curiosities into investigable questions.

Day 2: distinguish hypothesis and prediction. Write one mechanism-based hypothesis and two observable predictions that follow from it.

Day 3: retrieve variable, independent variable, dependent variable, controlled variable and constant. Label a simple investigation and explain one consequence of failing to control a relevant factor.

Day 4: study control group, experimental group, treatment and condition. Design a comparison in which the control provides a meaningful alternative outcome.

Day 5: work with trial, repetition, replication, procedure and method. Explain why five repeated trials by one group are not five independent replications.

Days 6–10 — Design Quality and Sampling

Day 6: retrieve experimental design, fair test, confounding variable and randomisation. Repair the fertiliser experiment from Laboratory 2.

Day 7: study sample, population and sampling. Define the population before choosing a sampling method.

Day 8: retrieve bias. Create one sampling-bias example, one measurement-bias example and one interpretation-bias example.

Day 9: build a field-sampling plan using a quadrat or transect. Explain why the sample can support some generalisation but not unlimited generalisation.

Day 10: mixed retrieval: explain ten terms without notes, then correct every explanation whose boundary is too broad.

Days 11–15 — Measurement Quality

Day 11: retrieve measurement, instrument, calibration, unit and scale. Write one complete measurement record with method and unit.

Day 12: distinguish precision, accuracy and resolution using three imaginary instruments.

Day 13: work with uncertainty, error, systematic error and random error. For each error source, state whether repetition alone would help.

Day 14: analyse the pendulum laboratory and write one paragraph explaining why many decimal places do not prove accuracy.

Day 15: design a measurement protocol that another student could follow consistently.

Days 16–20 — Data and Pattern

Day 16: retrieve data, observation, qualitative data, quantitative data and dataset. Convert one vague description into a structured data record.

Day 17: study table, graph, axis, trend and pattern. Choose a graph type for three different variable combinations.

Day 18: work with outlier, mean, median, range, variation and distribution. Analyse a dataset where mean and median tell different stories.

Day 19: retrieve correlation, causation and relationship. Rewrite five causal headlines as accurate correlational statements where the design is observational.

Day 20: complete the screen-time laboratory and list the evidence needed to make a stronger causal argument.

Days 21–25 — Claim, Evidence and Explanation

Day 21: retrieve claim, evidence and reasoning. Write a CER paragraph from a new dataset.

Day 22: distinguish conclusion, interpretation and analysis. Annotate a sample report so each sentence has one primary job.

Day 23: study explanation, model and mechanism. Take a simple graph and write first a description, then a mechanism-based explanation.

Day 24: retrieve theory, scientific law, assumption and alternative explanation. Correct the misconception that theories become laws.

Day 25: work with limitation and validity. For one investigation, write a limitation only if you can explain which conclusion it weakens.

Days 26–30 — Research Quality and Scientific Knowledge

Day 26: retrieve reliability, repeatability, reproducibility and verification. Design two checks that operate at different levels.

Day 27: study falsifiable, peer review, citation, source and scientific literature. Trace one scientific claim from a summary to its underlying source.

Day 28: retrieve methodology, protocol, ethics, safety, risk, hazard and consent. Write a safe, ethical plan for a fictional student survey or laboratory activity.

Day 29: work with data integrity, transparency, generalisation, scope and confidence. Audit a report that hides an excluded outlier or overextends its sample.

Day 30: finish with revision, scientific knowledge and scientific consensus. Explain how science can be reliable while remaining open to change.

Five Levels of Advanced Science Vocabulary Mastery

Level 1 — Recognition: the learner recognises the term in a science passage and can select a reasonable broad meaning.

Level 2 — Retrieval: the learner explains the term without notes and supplies an original scientific example.

Level 3 — Distinction: the learner rejects a convincing near-miss such as precision for accuracy, prediction for hypothesis or repetition for replication.

Level 4 — Application: the learner uses the vocabulary to diagnose a design, graph, measurement or conclusion.

Level 5 — Transfer: the learner can apply the same distinction in an unfamiliar science domain and revise the explanation when new evidence appears.

Part VII — The Secondary 1 Scientific Reasoning Operating Manual

The operating manual converts the vocabulary into a repeatable scientific workflow: Phenomenon → Question → Design → Measurement → Data → Pattern → Claim → Explanation → Evaluation → Revision. Students do not need to force every science task through identical steps. They need to recognise what kind of evidence the question requires and which link in the reasoning chain is weakest.

Module A — Turn a Phenomenon Into a Researchable Question

Science often begins with something noticed: leaves near a lamp bend toward light, a metal spoon feels colder than a wooden spoon in the same room, one solution changes colour faster than another, or some regions of a field contain more clover. The first task is to describe the phenomenon before trying to explain it.

Maren writes the observation without causal language. “Plants nearest the window were taller” is different from “more sunlight caused the plants to grow taller.” The first records a pattern. The second adds a causal explanation that still needs evidence.

Iona then turns the observation into a research question. Good questions name a measurable relationship or phenomenon. “Why are plants weird near windows?” is too vague. “How does daily light exposure relate to stem growth in bean seedlings over fourteen days?” defines variables, organism and time.

Leonie checks whether the question is testable with available methods. Some questions are scientifically meaningful but impractical for a school laboratory. If direct manipulation is impossible, observation or modelling may be more appropriate than forcing a controlled experiment.

Maren distinguishes question from hypothesis. A research question can exist without a hypothesis. If prior knowledge supports a tentative explanation, the learner can formulate one. If the purpose is descriptive—such as mapping clover abundance across a field—the study may proceed without a causal hypothesis.

Iona makes the hypothesis mechanistic where possible. “Plants grow more with more light” is a prediction-like statement. “Greater light availability increases photosynthetic energy capture, which can support greater biomass accumulation when other resources are adequate” contains explanatory structure, though the exact level of detail should match the learner’s knowledge.

Leonie derives a prediction that can be observed. The prediction should follow from the hypothesis rather than be a separate guess. If the hypothesis predicts greater growth with more light, the measurable outcome might be larger mean height gain or dry mass in the higher-light condition.

Operating drill: choose four everyday phenomena. For each, write an observation, a research question, an optional hypothesis and one prediction. Mark any question that is not practically testable in the classroom and propose another scientific method.

Module B — Design the Comparison Before Collecting Data

A weak investigation often collects many numbers before deciding what comparison matters. Strong design works backward from the claim the student hopes to test. If the question concerns temperature, the design must vary temperature while managing other factors that could alter the outcome.

Maren identifies the independent and dependent variables. She then lists plausible controlled variables rather than writing “keep everything the same,” which is impossible and unhelpful. Relevant controls are factors likely to influence the dependent variable or make trials incomparable.

Iona asks whether a control group is needed. In a fertiliser experiment, untreated plants create a useful comparison. In a pendulum investigation comparing several lengths, there may be no separate “control group”; each length is simply one condition in a structured relationship. Advanced vocabulary should not force a control group into every design.

Leonie checks randomisation and order. If all low-temperature trials occur first and all high-temperature trials occur later, an unnoticed time-related change can become entangled with temperature. Randomising or alternating order can reduce such systematic bias.

Maren makes the procedure operational. “Heat the water,” “stir normally” and “measure growth” are not reproducible instructions. State target temperature, stirring method, timing rule, measurement point and instrument.

Iona anticipates confounding. Ask what other factor could vary with the independent variable and affect the outcome. If fertilised plants sit closer to the window, sunlight competes as an explanation. If one material is tested with a thicker sample than another, thickness becomes part of the comparison.

Leonie plans repetition before seeing the first result. Deciding to repeat only when a result looks inconvenient creates bias. The design should specify how many trials or samples will be used and what rule will handle failed measurements.

Operating drill: design an investigation of one factor affecting dissolving, cooling, friction or plant growth. Include variables, controls, number of trials, order, operational definitions and one potential confounder.

Module C — Build Measurement Quality Into the Method

Measurement is not a neutral window onto reality. Instruments have limits, observers make choices and operational definitions shape what is recorded. Students should therefore design measurement quality before collecting data rather than apologise for weaknesses afterward.

Maren chooses an instrument appropriate to the quantity. A ruler can measure centimetre-scale length but cannot resolve microscopic change. A kitchen timer can time minutes but may be unsuitable for millisecond events. The instrument should match the expected range and required resolution.

Iona checks calibration or reference performance. A balance can display many digits while remaining systematically offset. A thermometer can respond slowly. A colour sensor may depend on ambient light. Calibration and validation against known references reveal whether the readings deserve confidence.

Leonie writes units into the data sheet before the first trial. Missing units create ambiguity later. She also decides the reasonable number of decimal places based on instrument resolution rather than copying every digit from a calculator.

Maren distinguishes systematic and random error. Repeating a measurement can reduce the influence of random scatter on an average. It cannot fix a ruler whose zero mark has been cut off unless the offset is identified and corrected.

Iona considers observer judgment. When leaf colour is classified by eye, categories should be defined with a reference chart. When a timer stops at “fully dissolved,” the stopping rule must be explicit. Operational definitions convert subjective judgments into more consistent procedures.

Leonie records uncertainty honestly. A scientific report becomes stronger, not weaker, when it states the realistic limit of a method. False precision—such as reporting 12.347891 cm from a millimetre ruler—creates the appearance of knowledge the instrument never provided.

Operating drill: choose three instruments and write quantity, range, resolution, calibration check and likely error source for each. Then explain one way a precise result could still be inaccurate.

Module D — Preserve Raw Data Before Summarising It

Students often jump directly from experiment to graph. Strong science preserves the raw record first. A table of every trial allows later checking, reanalysis and identification of outliers. A single average can hide the structure of the data.

Maren separates raw and derived values. Raw measurements might be 41, 42, 44 and 43 seconds. The mean of 42.5 seconds is a derived value. Both belong in the analytical record, but they perform different roles.

Iona checks data integrity. Failed trials, missing measurements and excluded values should be documented. A student should not quietly remove an inconvenient point after seeing that it weakens the preferred conclusion.

Leonie chooses summary statistics after inspecting distribution. Mean is useful when values are reasonably symmetric and no extreme value dominates. Median can be more robust when a few extreme values pull the mean. Range provides a simple view of spread but uses only the extremes.

Maren selects a graph matched to the variables. A scatter plot is useful for two quantitative variables. A line graph can show ordered change over time or another continuous scale. A bar chart is often suitable for comparing categories. The graph type should emerge from the data structure.

Iona checks axes, units and scale. Truncated axes can exaggerate small differences visually. Unequal intervals can distort interpretation. Missing units can make values unusable. The graph should make the data easier to understand without changing their meaning.

Leonie annotates unusual points rather than hiding them. If an outlier is excluded from a summary because a documented procedural failure occurred, the report should say so and preserve the original value in the record.

Operating drill: create a seven-trial dataset containing one outlier. Calculate mean, median and range. Decide whether the outlier should remain in the summary under two different scenarios: documented instrument failure and no known cause.

Part VII — Scientific Reasoning Operating Manual: Analysis, Explanation and Research Quality

Module E — Move From Pattern to Claim Without Overclaiming

Once data are organised, the next danger is linguistic: a pattern becomes a conclusion stronger than the design permits. “A increased when B increased” can become “B caused A” in one careless sentence. Advanced science vocabulary exists partly to prevent that jump.

Maren begins with the most descriptive statement the data directly support. If two variables move together, she describes a positive or negative relationship. If a controlled experiment changed one factor and managed plausible alternatives, stronger causal language may become appropriate.

Iona asks whether the sample and range support the scope of the claim. A relationship observed from 10°C to 50°C should not become “temperature always has this effect.” A study of one species does not automatically support a statement about all organisms.

Leonie checks variation and uncertainty before choosing verbs. A weak, noisy trend may justify “suggests” or “is associated with.” A strong, replicated experimental result may justify firmer language. Qualification should follow evidence strength rather than personal caution level.

Maren distinguishes statistical description from mechanism. A graph can show that values rise together. It cannot by itself reveal what microscopic or biological process produces the relationship. That explanation comes from scientific reasoning, prior knowledge and additional evidence.

Iona adds alternative explanations explicitly when the design leaves them open. “The observed difference could reflect treatment, baseline group differences or measurement bias.” This does not make the study worthless. It calibrates the causal claim.

Leonie then writes a conclusion with four parts: result, evidence summary, scientific interpretation and limitation. The structure is especially useful for open-ended Science answers because each part performs a distinct reasoning job.

Operating drill: take one scatter plot and one controlled experiment. Write the strongest justified claim for each, then underline any word that would overstate causation or generalisation.

Module F — Build Explanations With Models and Mechanisms

Description tells what happened. Explanation tells how or why. Scientific explanations become powerful when they connect evidence to a model or mechanism without pretending that the model contains every detail of reality.

Maren uses the pattern Observation → Scientific Idea → Mechanism → Outcome. If an insulated container cools more slowly, the explanation connects energy transfer, material properties and rate of heat loss. The graph supplies the observation; the mechanism supplies the why.

Iona checks whether each mechanism step is actually relevant. Students sometimes add correct science facts that do not explain the result. Saying “particles are always moving” is true but incomplete unless the explanation connects particle behaviour with the measured change.

Leonie makes predictions from the model. A good model should do more than explain past data; it should help predict what might happen under a new condition. If the model predicts slower cooling with thicker insulation, a new test can challenge or support it.

Maren also states model limits. A simple particle diagram may ignore intermolecular forces because they are unnecessary for the current question. That simplification is acceptable when it serves the explanatory purpose and does not contradict the intended phenomenon.

Iona distinguishes theory and law in reading passages. A law may summarise a regular quantitative relationship. A theory can explain a wider set of phenomena through mechanisms and principles. Neither label represents a rung on a ladder where one automatically becomes the other.

Leonie treats unexpected evidence as diagnostic. First check the measurement and procedure. Then test alternative explanations. If the unexpected pattern survives scrutiny, the model or its assumptions may need revision.

Operating drill: take a familiar model and produce one explanation, one prediction, one assumption and one known limitation. Then state what observation would cause you to revise the model.

Module G — Evaluate Reliability, Validity and Reproducibility Separately

Students often call an investigation “reliable” when they mean accurate, fair, repeated or believable. Advanced scientific evaluation separates these dimensions so recommendations target the actual weakness.

Maren asks about reliability first: do repeated measurements or repeated applications of the method produce consistent results? If values vary wildly under nominally identical conditions, the method may lack repeatability or the system may contain uncontrolled variation.

Iona asks about validity: does the measure capture the intended variable, and does the design answer the research question? A consistently scored questionnaire can be reliable while measuring a different concept from the one claimed.

Leonie then asks about reproducibility. Can another group understand and repeat the method? Are units, instruments, exclusions and analysis rules described? Reproducibility requires more than saying “we did the same experiment again.”

Maren checks verification through independent routes. A second instrument, another observer, a calibration standard or independent dataset can reveal errors that internal repetition misses.

Iona distinguishes design improvement from cosmetic improvement. Adding more decimal places does not improve validity. Increasing the sample size does not fix a biased sampling rule. Repeating a systematically miscalibrated measurement does not fix accuracy.

Leonie writes recommendations that correspond to diagnoses: calibrate the balance for accuracy, define categories for observer consistency, randomise group assignment to reduce confounding, increase sample coverage to improve generalisation, and provide a full protocol to improve reproducibility.

Operating drill: diagnose five fictional weaknesses and match each to one improvement. Award no credit for a generic “repeat more” answer unless repetition addresses the actual problem.

Module H — Treat Scientific Knowledge as a Revisable Evidence Network

Science is sometimes presented as a book of final answers. A stronger picture is a network of claims connected to experiments, observations, models, methods, citations, replications and critical review. Some parts of the network are extremely well supported; others remain active areas of uncertainty.

Maren traces a claim backward. What study or dataset supports it? What method generated the data? What assumptions entered the analysis? Which later studies replicated, refined or challenged the finding? Citation becomes a route through the evidence network rather than decoration at the end of a paragraph.

Iona checks convergence. Independent methods pointing toward the same conclusion can increase confidence because they do not share every weakness. One instrument may have one source of error, another a different one. Agreement across independent approaches is therefore more informative than repeated quotation of one study.

Leonie treats peer review as one checkpoint, not a magic stamp. Reviewers can miss errors, studies can be underpowered and interpretations can later change. Replication, reanalysis and new evidence continue after publication.

Maren distinguishes scientific consensus from popularity. Consensus emerges when relevant experts broadly converge because multiple evidence streams support a conclusion. It is not produced by counting social-media posts or asking whether everyone agrees on every detail.

Iona asks what would change confidence. A falsifiable scientific claim should encounter possible evidence that could count against it. If no conceivable result can weaken the claim, it is difficult to test scientifically.

Leonie finishes with revision. When a better model explains more evidence with fewer unresolved contradictions, scientific knowledge changes. Revision is not proof that science is unreliable. It is part of the system that makes scientific knowledge self-correcting over time.

Operating drill: choose one scientific claim from a textbook. Build a simple evidence network containing source, method, data, model, replication and one limitation. Then explain what new evidence would strengthen or weaken confidence.

The Scientific Reasoning Operating Manual in One Page

• Phenomenon: describe what is observed before explaining it.
• Question: define the relationship or process you want to investigate.
• Design: choose a method capable of producing relevant evidence.
• Variables: identify what changes, what is measured and what must be managed.
• Measurement: define instruments, units, resolution and operational rules.
• Data: preserve raw observations before summarising.
• Pattern: describe trends, spread and outliers before explaining them.
• Claim: state only what the design and data support.
• Explanation: connect evidence with models, mechanisms and scientific principles.
• Evaluation: separate reliability, validity, bias, uncertainty and scope.
• Verification: seek independent checks, replication and transparent reporting.
• Revision: change the conclusion or model when stronger evidence requires it.

Closing Principle — Scientific Vocabulary Is a System for Controlling Claims

The most advanced student is not the one who says hypothesis, validity, confounder and reproducibility most often. It is the student who knows which word changes the reasoning. A hypothesis needs a mechanism. A fair comparison needs controlled alternatives. A measurement carries uncertainty. Data become evidence only in relation to a claim. A model has a scope. A conclusion must remain revisable.

Part VIII — Four Integrated Scientific Investigation Cases

These cases require several vocabulary systems at once. The data and organisations are fictional. Work through the questions before reading the analyses. The objective is not to guess the teacher’s preferred conclusion; it is to show that every conclusion remains connected to design, measurement and evidence.

Integrated Case A — Which Battery Lasts Longest?

A consumer-science club compares three AA battery brands in identical battery-powered lamps. Each lamp uses the same bulb model. Students install fresh batteries and record the time until the lamp’s measured light output falls below 50% of its initial value. Brand A is tested on Monday, Brand B on Tuesday and Brand C on Wednesday.

The first dataset contains three trials per brand. Brand A lasts 412, 419 and 416 minutes. Brand B lasts 448, 452 and 451 minutes. Brand C lasts 463, 470 and 466 minutes. A simple comparison suggests C lasts longest. But Iona notices that room temperature increased across the three days because the building’s air-conditioning failed.

Design diagnosis: day and brand are confounded. If temperature affects battery performance or lamp electronics, the experiment cannot cleanly separate brand from day. The data remain real, but the causal interpretation “Brand C caused the longest life” is weakened.

Maren redesigns the experiment so all three brands are tested each day, with lamp positions randomised. Each brand receives six trials distributed across days. The lamps are rotated among table positions. Room temperature is recorded. The stopping rule—light output below half the initial calibrated reading—is operationally defined before testing begins.

The revised means are A = 415 minutes, B = 449 minutes and C = 451 minutes. C remains slightly higher than B, but the difference is much smaller than in the first design. Trial ranges overlap substantially. The evidence now supports a strong difference between A and the other brands but much weaker evidence that C genuinely lasts longer than B.

Leonie asks whether “longest battery” is even the full consumer question. Price, shelf life, device compatibility and environmental impact could matter. The experiment validly addresses lamp runtime under the tested conditions, not overall product value.

Iona also checks replication. One school club using one lamp model does not establish universal battery performance. Another class using the same protocol with a different batch and device could test reproducibility.

Student questions: What was the confounding variable in the first design? Why does testing all brands on each day improve validity? What do the revised data support about A, B and C? Which conclusion would overgeneralise beyond the experiment?

Worked analysis: the improved design breaks the systematic link between brand and day. The revised means still suggest A performs worse in this lamp setup. B and C are close enough that the available sample does not justify a strong claim that C is superior in all settings. A careful conclusion would name the device, stopping rule and test conditions.

Case lesson: better design can change not only confidence but the apparent size of an effect.

Integrated Case B — Which Material Is the Best Insulator?

Students compare cotton, bubble wrap, aluminium foil and felt by wrapping identical cups containing hot water. They record temperature after twenty minutes and choose the material leaving the water hottest. The first conclusion says, “Felt is the best insulator.”

Maren checks the design and finds that material thickness differs. The felt layer is 8 mm thick, cotton 5 mm, bubble wrap 4 mm and foil less than 1 mm. The experiment therefore compares complete wrapping conditions, not material type alone. Thickness is part of the treatment.

Iona asks what “best” means. Highest final temperature? Smallest temperature drop per millimetre of material? Lowest mass? Lowest cost? A scientific comparison needs a criterion. Without one, “best” hides a value judgment inside an experimental result.

Leonie redesigns the question: “How does equal-thickness wrapping material affect temperature loss from 200 mL of water over twenty minutes?” Equal thickness makes material type the intended independent variable. Cup type, water volume, initial temperature, lid use and room position are controlled.

The class records initial temperature rather than assuming each cup begins identically. Temperature loss is calculated for each trial. This matters because one cup starting at 91°C and another at 86°C should not be compared using final temperature alone.

Suppose felt shows a mean loss of 14°C, bubble wrap 13°C, cotton 18°C and foil 22°C. Bubble wrap now gives the smallest mean temperature loss, although one felt trial overlaps its range. The revised result differs from the first because the question and controls changed.

Maren develops the explanation using mechanisms. Trapped air and material structure can reduce heat transfer, but the experiment itself does not directly measure every transfer mechanism. The explanation should connect with accepted models of conduction, convection and radiation while recognising what the data actually measured.

Iona notes a remaining limitation: equal thickness may require different masses of material. If engineering design cares about weight or cost, another investigation is needed. One experiment answers one defined comparison well; it does not settle every design criterion.

Student questions: Why was the first comparison confounded? Why is temperature change a more valid outcome than final temperature when starting values differ? What criterion is used in the redesigned study? How could another criterion produce a different “best” material?

Case lesson: experimental quality improves when value words such as best are converted into measurable criteria.

Integrated Case C — Is the Pond Water Becoming More Turbid?

A school ecology group measures water turbidity in a pond once each month. A handheld sensor gives values in nephelometric turbidity units. The group observes higher readings during the rainy season and concludes that construction upstream is polluting the pond.

Iona separates the measured pattern from the causal claim. The data show turbidity varies over time and appears higher during rainy months. Turbidity can increase through suspended sediment, algal growth, runoff and other processes. The sensor does not identify the source of particles.

Maren examines sampling. Measurements are taken only from the pond edge near the school gate because it is convenient. That location may not represent the whole pond. A better sampling design includes several points—near inflow, centre, outflow and a reference area—using consistent depth and timing.

Leonie checks calibration. The sensor is calibrated using standard solutions before each sampling day. Without calibration, apparent month-to-month change could partly reflect instrument drift. Replicate readings at each location help assess local repeatability.

The group adds rainfall records and upstream sampling. Turbidity rises at both the upstream reference stream and the pond after heavy rainfall, including periods when construction activity is minimal. That evidence weakens the simple construction-only explanation and supports runoff as an alternative mechanism.

However, turbidity near the construction-side inflow remains consistently higher than the upstream reference after similar rainfall events. The pattern is compatible with an additional local sediment source. The evidence network now supports a more nuanced conclusion than either “construction caused everything” or “construction had no effect.”

Maren writes: “Pond turbidity increased during rainy periods across sites, indicating a strong rainfall-related component. Readings near the construction-side inflow were additionally elevated under comparable rainfall, suggesting a possible local sediment contribution that warrants further investigation.”

Iona checks scope. The study measures turbidity, not chemical pollution or overall ecosystem health. The word pollution can be too broad unless the specific contaminant or ecological effect is measured.

Student questions: What did the original sensor data actually measure? How did broader sampling change interpretation? Why is rainfall an alternative explanation? What additional evidence would strengthen attribution to construction?

Case lesson: field science often builds causation through converging observations rather than one perfectly controlled experiment.

Integrated Case D — A Science Headline Says “Study Proves…”

A news headline reads: “Study Proves Background Music Makes Students Smarter.” The article describes a study in which 60 volunteers complete a short reasoning test twice: once in silence and once with instrumental music. Average scores are slightly higher in the music condition.

Maren rewrites the claim. The study measures performance on one short reasoning test, not intelligence as a whole. “Makes students smarter” extends both the outcome and causal scope far beyond the measurement.

Iona checks design details. If every participant completes the silent test first and music test second, order or practice effects could confound the result. Counterbalancing order across participants would reduce that problem.

Leonie examines the sample. Volunteers may differ from the wider student population. Sixty participants can provide useful evidence, but broad generalisation requires attention to age, recruitment and context.

The article reports only average scores. Iona wants variation, individual responses and uncertainty. A two-point mean difference could coexist with substantial overlap between conditions. The size and consistency of the effect matter.

Maren also asks whether instrumental music was standardised. Volume, tempo and familiarity could influence performance. “Music” is not one uniform treatment unless operationally defined.

The student follows the citation to the original paper. The authors themselves write that the result is preliminary and should not be interpreted as evidence of increased intelligence. The headline therefore overstates the source it claims to summarise.

Leonie writes a corrected headline: “Small Study Finds Slightly Higher Reasoning-Test Scores With Instrumental Music Under Tested Conditions.” The new headline is less dramatic but much closer to the evidence.

Student questions: Which words in the original headline overstate outcome, causation and certainty? What order effect could occur? Why does following the citation improve evaluation? What would replication add?

Case lesson: scientific literacy includes reading how evidence changes as it travels from research paper to public headline.

Part IX — Advanced Science Vocabulary Mastery Assessment

This original assessment samples the language and reasoning taught in the article. It is not an official examination, curriculum standard or predictor of a school grade. Complete the tasks before reading the worked guidance. The assessment rewards precision of scientific reasoning, not merely the number of target terms included.

Section A — Ten Distinction Checks

1. A student says, “My five readings were 10.1, 10.1, 10.2, 10.1 and 10.2, so they are accurate.” Which word is directly supported by the clustering—accuracy or precision? Explain what additional reference is needed before accuracy can be judged.

2. Another class repeats your entire investigation using your documented protocol and obtains the same overall relationship. Is this best described as repetition or replication? Explain why.

3. A table contains recorded temperature and time values. The student says, “This table is evidence.” Under what condition do those data function as evidence rather than only as recorded data?

4. Plants receiving fertiliser also receive more sunlight. Which term identifies the design problem: random error, confounding variable or outlier? Explain the alternative explanation created.

5. A claim says, “If a gas is heated, pressure will increase in this sealed fixed-volume container.” Is this a hypothesis, prediction or law? State what makes the sentence the category you choose.

6. A thermometer reads 1°C too high at every temperature. Which kind of error is this? Why would taking twenty readings with the same thermometer fail to solve the basic problem?

7. A study of one pond reports high turbidity after rainfall. Why would the claim “rain always increases turbidity in all ponds” be a generalisation problem even if the local measurements are correct?

8. A scientific model explains several observations but fails under extreme conditions. Does that automatically make the model useless? Use scope and revision in your answer.

9. An article has been peer reviewed. Why is it still reasonable to inspect methods, sample size and later replication?

10. A survey shows screen time and sleep duration are correlated. Write one statement the survey supports and one stronger statement it does not establish by itself.

Section B — New Investigation Case

A fictional class tests whether ramp angle affects the time a toy car takes to travel one metre. They use angles of 10°, 20°, 30° and 40°. Each angle is tested four times. The same car and ramp are used throughout. Timing is done by hand with a stopwatch.

The recorded mean times are 3.8 s, 2.9 s, 2.3 s and 2.0 s. At 40°, the individual trials are 1.6, 1.8, 2.1 and 2.5 s, showing more variation than at the other angles. The group concludes: “Increasing ramp angle causes every toy car to travel faster, and 40° is the optimal angle.”

Question A: identify the independent variable, dependent variable and two controlled variables. Question B: describe the trend supported by the mean times. Question C: identify two reasons the conclusion is too broad. Question D: explain what the variation at 40° suggests about measurement or control. Question E: propose one measurement improvement and one scope correction.

Worked Guidance for Section A

1 — Precision. The readings are close to one another, which supports repeatability or precision. Accuracy requires a reference value or calibrated standard. Tight clustering around the wrong value would still be inaccurate.

2 — Replication. Another class independently repeating the investigation tests whether the finding survives another execution. Repetition more often refers to repeated measurements or trials within the same study.

3 — Claim relationship. The values become evidence when used to support, challenge or qualify a defined claim or research question. A table by itself is data; the evidential role comes from interpretation.

4 — Confounding. Sunlight changes systematically with fertiliser and can affect plant growth. The observed difference therefore has at least two plausible causal pathways.

5 — Prediction. It states an expected observable relationship under specified conditions. A hypothesis would usually propose an explanatory mechanism; a law is a broader established descriptive relationship, not one student’s single prediction.

6 — Systematic error. Every measurement is shifted in the same direction by the calibration offset. Repetition can reveal consistency but will reproduce the same offset unless calibration or correction changes.

7 — Scope. One pond, one rainfall pattern and one measurement period do not represent all ponds or all climates. The local finding can remain valid while the universal generalisation is unsupported.

8 — No. Models have purposes and domains. A model can remain useful within its valid scope while requiring refinement or replacement for conditions it was not designed to represent.

9 — Peer review is one checkpoint. Review can improve quality but does not make errors impossible. Methods, sample size, assumptions, replication and later evidence remain relevant to confidence.

10 — Supported: students reporting more evening screen time also reported shorter sleep duration in the surveyed sample. Not established: increasing screen time necessarily causes each student to sleep less. Confounding, reverse causation and measurement limitations remain possible.

Worked Guidance for Section B

A: ramp angle is the independent variable; travel time is the dependent variable. Controlled variables include car, ramp surface, travel distance and release method. The answer should identify actual controlled factors, not simply say “everything else.”

B: mean travel time decreases as ramp angle increases from 10° to 40° in this setup. The statement describes the measured relationship and preserves the tested range.

C: “every toy car” generalises beyond the one car tested. “Optimal angle” requires a criterion and a wider range of angles; the study only shows that 40° has the shortest mean time among the four tested conditions.

D: the wider spread at 40° suggests reduced repeatability or greater sensitivity to release/timing variation. It does not by itself identify the cause. Hand timing and release consistency should be checked.

E: a light gate or video timing system could reduce reaction-time uncertainty. A scope correction would say “for this car and ramp over the angles tested” instead of “every toy car.”

Section C — Write an Evidence-Controlled Conclusion

Write 130–170 words explaining the ramp experiment to another Secondary 1 student. Include at least four target terms accurately. Your paragraph should describe the trend, evaluate one measurement weakness, qualify the scope and propose one improvement. Do not claim that the investigation proves a universal law.

Suggested marking dimensions: scientific vocabulary precision, evidence use, design evaluation, scope control and clarity. A learner who uses ordinary language accurately should outperform a learner who inserts many technical terms incorrectly.

Model Conclusion

The experiment found a clear trend in this setup: mean travel time decreased as ramp angle increased from 10° to 40°. This supports the claim that, for the tested car and ramp, steeper angles were associated with faster travel over one metre. However, the 40° trials showed greater variation, so the repeatability of the timing was weaker at that condition. Hand timing may have introduced random error, and a light gate could reduce this uncertainty. The conclusion should not be generalised to every toy car or every ramp because only one car and surface were tested. Nor can 40° be called the optimal angle without defining an optimisation criterion and testing additional angles. A stronger follow-up would standardise the release mechanism, use electronic timing and replicate the relationship with another car.

Teacher and Parent Guide — Teach the Scientific Distinction, Not Just the Definition

This collection is deliberately larger than a list a learner should memorise in one sitting. Select the vocabulary that matches the scientific reasoning problem currently appearing in the student’s work. If a learner confuses hypothesis and prediction, teach that pair through one investigation. If conclusions overreach the sample, teach scope and generalisation. If repeated measurements are called accurate simply because they cluster, teach precision, accuracy and calibration together.

Start with the learner’s existing explanation before supplying the advanced term. Ask, “What exactly changed?” “What did you measure?” “What else could explain the result?” “How do you know the measurement is trustworthy?” A student who can answer those questions in ordinary language is ready to attach the specialist vocabulary. A student who cannot answer them gains little from copying the technical noun.

A Three-Student Lesson: One Investigation, Three Scientific Jobs

Give three learners the same short investigation: “A metal ball rolls down ramps set at 10°, 20° and 30°. The time to travel one metre is measured three times at each angle. The same ball and ramp surface are used.” Before discussion, each learner writes a private first response to three questions: what is being changed, what is being measured and what conclusion could the data support?

Maren’s role is structure. She labels independent variable, dependent variable and controlled variables. Iona’s role is evidence. She asks whether three trials are enough, whether hand timing introduces uncertainty and whether the conclusion applies to all balls and ramps. Leonie’s role is execution. She checks release method, timing rule, units and whether another group could follow the procedure.

Rotate the roles on a second problem. A student who could label variables when prompted should now diagnose the measurement weakness. A student who noticed uncertainty should now write the conclusion. This rotation checks whether the vocabulary belongs to the learner or only to the role they just practised.

End by changing one fact: the 30° trials were timed using a different stopwatch. Ask which conclusion now weakens and why. This small alteration tests whether the learner understands comparability rather than memorising the original answer.

Science Writing Workshop — From Data Table to Controlled Explanation

Workshop 1 — Replace “Proves” With the Strongest Justified Verb

Students often write “the experiment proves…” because the phrase sounds scientific. Replace it only after deciding what the evidence actually does. A controlled experiment may support a causal conclusion within the tested conditions. An observational study may show association. A small pilot may suggest a pattern. A single unexpected result may challenge a prediction without overturning an entire theory.

Weak: “The experiment proves that hotter water always dissolves sugar faster.” Controlled: “Across the tested temperatures, higher water temperature was associated with shorter dissolving time in repeated trials.” If the design controls plausible alternatives well enough, the next sentence can discuss why the temperature change is likely causal in this setup.

The advanced skill is not replacing proves with suggests automatically. It is matching the verb to design strength, sample, uncertainty and scope.

Workshop 2 — Build a Claim–Evidence–Reasoning Paragraph

Use three explicit jobs. Claim: answer the question. Evidence: name the relevant pattern or comparison. Reasoning: explain why that evidence supports the claim using scientific ideas and design logic.

Model: “Within the tested range, increasing water temperature reduced dissolving time. At 10°C the mean time was 96 seconds, while at 50°C it was 38 seconds, and intermediate temperatures showed the same downward trend. Because sugar-cube mass, water volume and stirring procedure were controlled, temperature is a plausible explanation for the difference. A particle model predicts more rapid molecular motion at higher temperature, providing a mechanism consistent with the observed pattern.”

Notice what the paragraph does not claim. It does not say every solute behaves identically or that the relationship continues without limit beyond the tested range. The strongest writing includes enough boundary information that the reader knows exactly what was established.

Workshop 3 — Write a Limitation That Actually Changes the Conclusion

“Human error may have occurred” is rarely useful because it names no mechanism. A stronger limitation states the source, likely effect and affected conclusion. “The endpoint was judged by eye, so different observers may have stopped timing at different levels of remaining crystal; this increases uncertainty in the dissolving-time comparison.”

Do not list every imaginable imperfection. Prioritise limitations that materially affect validity, reliability, bias or generalisation. A limitation about handwriting quality is irrelevant if handwriting could not change the measurement.

Workshop 4 — Turn “Make It More Accurate” Into a Real Improvement

An improvement should target the diagnosed weakness. If reaction time creates random timing variation, use electronic timing or measure a longer interval. If an instrument has systematic offset, calibrate it. If sampling is biased, change the sampling method. If a treatment is confounded with location, randomise or balance locations.

Weak: “Repeat more for accuracy.” Better: “Repeat each condition five times and calculate a mean to reduce the influence of random timing variation; calibrate the sensor separately because repetition does not remove systematic offset.” The second answer distinguishes two error mechanisms and two remedies.

Workshop 5 — Explain an Unexpected Result Without Hiding It

Unexpected data are opportunities for scientific reasoning. Begin with verification: was the value copied correctly, was the instrument functioning, and was the protocol followed? Then examine whether the result might be real. If so, consider whether the original model omitted an important factor.

A strong report might say: “Trial 4 produced a value substantially above the other repetitions. The original record and instrument were checked and no procedural failure was identified, so the value was retained. The mean is therefore reported alongside the median and range, and additional trials are recommended.” Transparency is stronger than silently removing the point.

Frequently Asked Questions About Advanced Secondary 1 Science Vocabulary

Is this an official Secondary 1 science word list?

No. It is an eduKateSingapore advanced enrichment collection. Curriculum terminology differs across Singapore, international schools and other countries. The terms are selected because they support scientific inquiry, evidence and technical reading across early-secondary contexts. Teachers should align exact syllabus expectations with the relevant school or examination authority.

Why not just teach “the scientific method” as seven fixed steps?

Real scientific inquiry uses multiple methods. Controlled experiments are important, but observational studies, field sampling, modelling, remote sensing and analysis of existing data also produce scientific evidence. Middle-school learners benefit from seeing that the question determines the method. The National Science Teaching Association’s “Doing Science” resource explicitly warns against teaching one rigid linear method as though every scientific investigation follows it.

Do all investigations need a hypothesis?

No. Hypotheses are useful when prior knowledge supports a tentative explanation that can be tested. Descriptive investigations can instead begin with a question and collect evidence about a phenomenon or pattern. The NSTA resource on hypotheses notes that middle-school students should learn what hypotheses are and when they are used, without assuming every investigation must contain one.

Why distinguish data from evidence?

Because a table does not explain itself. Data are observations or measurements. They function as evidence when a learner connects them to a scientific question or claim and explains why they are relevant. This is central to stronger Science answers: the student should not merely quote a number but show how the number supports, weakens or qualifies the conclusion.

Why distinguish accuracy and precision so carefully?

The distinction prevents a common mistake: assuming closely clustered readings must be correct. Precision can be high even when every value is shifted by the same calibration problem. The repair differs too. More repeats can help characterise random variation, while calibration or method correction is needed for systematic offset.

What science vocabulary should become productive first?

Prioritise terms the learner needs to explain current school tasks: variable, evidence, control, pattern, accuracy, conclusion, limitation and mechanism often provide high utility. Terms such as falsifiable, methodology or scientific consensus may remain secure reading vocabulary before they become natural words in student writing. Productive use should follow conceptual control.

How do I know whether a student really understands a term?

Ask for a contrast and an application. “What is precision?” tests a definition. “Can readings be precise but inaccurate?” tests the boundary. “Here are five readings and a calibration record—what can you conclude?” tests application. Later, change the scientific domain and see whether the distinction transfers.

Does peer review mean a study is definitely correct?

No. Peer review is an important quality-control process, but scientific claims continue to be tested after publication through replication, reanalysis, new data and wider comparison with the literature. Reliable science is not defined by never changing; it is defined partly by methods that allow claims to be scrutinised and revised.

What should a student do when two studies disagree?

Compare population, sample size, measurement method, treatment, outcome definition, uncertainty and analysis. The studies may not have tested exactly the same question. If they did, disagreement can identify uncertainty worth investigating. Students do not need to force an immediate winner when the evidence remains mixed.

Authoritative Reference Route

For middle-school investigation practice, the NSTA overview of planning and carrying out investigations describes identifying independent and dependent variables, controls, tools, measurements and sufficient data in Grades 6–8. The NSTA overview of asking questions and defining problems emphasises questions that require appropriate empirical evidence. These frameworks support the article’s emphasis on question-driven investigation rather than a single memorised recipe.

The NSTA Grade 7 resource on variables provides a direct middle-school example of control-of-variables reasoning. For broader middle-school science expectations, the Next Generation Science Standards middle-school evidence statements illustrate how evidence, models and investigations appear across scientific domains. These are external reference frameworks, not claims that this eduKate list is an official NGSS or NSTA vocabulary syllabus.

Continue the Advanced Secondary 1 Collection

Use the general Advanced Secondary 1 Vocabulary collection for broader English vocabulary and the Vocabulary Article Directory for other thematic routes. For the underlying science system, continue to Scientific Inquiry & Evidence | How Science Knows What It Knows. This article remains the advanced vocabulary owner for scientific investigation, measurement, data and evidence rather than duplicating the deeper Science World explanations.

Final Principle — Good Science Writing Makes the Evidence Boundary Visible

A strong Secondary 1 scientist does not merely know difficult words. The learner can show which variable changed, which measurement was taken, how uncertainty entered, what the data pattern supports, which alternative explanation remains and how far the conclusion can travel. That is the point of the advanced collection: vocabulary should make scientific reasoning easier to inspect, challenge, reproduce and improve.

Final Evidence-Network Clinic — From a Scientific Headline Back to the Investigation

A student sees the headline, “New Material Cuts Heat Loss by 40%.” It sounds simple, numerical and authoritative. Advanced scientific reading asks what chain of evidence sits behind the sentence. The headline is the final public claim. Beneath it should be a source, a study, a method, measurements, analysis and an interpretation. If one link is missing or distorted, confidence should change.

Step 1 — Rewrite the Headline as a Precise Claim

Maren removes the promotional wording and asks what “cuts heat loss by 40%” means. Compared with what material? Over what time? Under which temperature difference? Was heat loss measured directly, or estimated from temperature change? Does 40% describe a relative difference in one laboratory setup or a universal property of the material?

The rewritten claim becomes: “In a laboratory cup-cooling test, containers wrapped with Material X lost 40% less thermal energy over twenty minutes than containers wrapped with the comparison material, under the stated test conditions.” This sentence is less memorable but much more useful. It identifies comparison, outcome, period and setting.

Step 2 — Find the Original Source

Iona follows the article’s citation to a conference paper. The news article is a secondary source; the conference paper is closer to the investigation. She checks authors, institution, publication date and whether a later peer-reviewed version exists. Citation is not a magic badge—the source still requires evaluation—but traceability is already stronger than relying on the headline alone.

The paper reports that Material X was compared with felt. Both were wrapped around identical metal containers containing 250 mL of water. Initial temperature was approximately 80°C. Temperature was recorded every minute for twenty minutes. The “40%” figure refers to the calculated rate of energy loss over a selected ten-minute interval, not the entire twenty-minute test.

Step 3 — Inspect the Operational Definitions

Leonie asks how heat loss was operationally defined. The researchers did not measure thermal energy leaving the cup directly. They measured water temperature and used a calculation based on mass, specific heat capacity and temperature change. That method can be valid, but the claim “heat loss” depends on assumptions about where the energy goes and whether other components of the system matter.

The paper also defines insulation thickness as five millimetres for both materials. This matters because comparing a thick sample with a thin sample would mix material properties with geometry. Operational definitions protect the comparison from hidden ambiguity.

Step 4 — Evaluate Design and Measurement

Maren checks variables. Material type is the independent variable. Temperature change is the principal measured outcome used in the energy-loss calculation. Water volume, starting temperature, container type, material thickness and room position are controlled as far as practical.

Iona checks measurement quality. Temperature sensors were calibrated against a reference thermometer. Each condition was repeated eight times. Sensors recorded to 0.1°C resolution. The authors report that one trial was lost because a sensor failed and the missing measurement was documented rather than silently replaced.

Leonie notices that all Material X trials were performed in the morning and all felt trials in the afternoon. The room temperature differed slightly between sessions. This introduces a possible systematic condition difference. The paper attempts to correct for ambient temperature mathematically, but randomising material order would have created a cleaner design.

Step 5 — Read the Data, Not Only the Percentage

The mean calculated energy-loss rate for felt is 12.5 units per minute; Material X averages 7.5. The difference is 5 units, which is 40% of the felt value. This explains the headline figure. However, individual trials vary. Felt ranges from 11.2 to 13.8; Material X ranges from 6.9 to 8.4.

The groups are clearly separated in this dataset, increasing confidence that the observed difference is not produced by one outlier. Yet the percentage still belongs to this comparison and method. If Material X were compared with another insulation, the relative difference would change. “40% better insulation” without naming the baseline is incomplete.

Step 6 — Separate Result From Mechanism

The study shows a performance difference in the test. The authors propose that Material X’s microscopic pore structure traps air and reduces conductive and convective transfer. Microscopy images support the presence of pores, but the experiment does not directly measure every heat-transfer pathway.

Maren writes two sentences instead of one: “Material X reduced the measured cooling rate relative to felt in the test setup. Its porous structure provides a plausible mechanism because trapped air can reduce heat transfer.” The first sentence states the empirical result; the second connects it to an explanatory model.

Step 7 — Inspect Scope and Generalisation

Iona asks what the study did not test. It used small metal containers, one material thickness, indoor still-air conditions and a short time period. It did not test building walls, outdoor weather, fire behaviour, long-term durability, manufacturing cost or environmental impact.

Therefore, “Material X is the best building insulation” would be a major overgeneralisation. The study answers a narrower question about thermal performance in one laboratory model. Engineering adoption would require additional criteria and evidence.

Step 8 — Check Replication and the Wider Literature

Leonie searches the scientific literature and finds two later studies using different apparatus. One also reports lower heat transfer for Material X than felt. Another finds a smaller advantage when humidity is high. The findings broadly converge while revealing an important condition.

This is how scientific confidence becomes an evidence network rather than dependence on one dramatic result. Independent studies can share a pattern while refining the scope. Reproducibility does not require identical numerical results under different conditions; it asks whether the underlying finding remains compatible when the method changes appropriately.

Step 9 — Rewrite the Public Claim

After following the evidence chain, Iona rewrites the headline: “Porous Material Reduced Cooling Rate by About 40% Versus Felt in Laboratory Cup Tests; Humidity and Real-World Performance Need Further Study.” The headline remains interesting while preserving baseline, setting and uncertainty.

Leonie then writes the student summary: “The study provides evidence that Material X reduced heat transfer more than felt under the tested conditions. The comparison was repeated and produced a clear difference, although testing order created a possible room-temperature confounder. Later studies broadly support the performance advantage but suggest humidity matters. The evidence is therefore promising for the tested property, not proof that the material is optimal for every engineering use.”

Step 10 — Identify What Would Change the Conclusion

A strong scientific conclusion includes revision conditions. Confidence would increase if independent laboratories using randomised test order, wider temperatures and different humidity levels recovered the effect. Confidence would fall if calibrated replication showed no difference, if the original calculation contained an error, or if the apparent advantage disappeared when ambient conditions were controlled.

This final step matters because scientific confidence should be earned and revisable. A conclusion that cannot state what evidence would count against it risks becoming protected from testing rather than strengthened by testing.

The Evidence-Network Checklist

• Claim: what exactly is being asserted?
• Source: where did the claim originate?
• Operational definition: how were the important variables measured?
• Design: what comparison and controls were used?
• Measurement: what instruments, calibration and uncertainty matter?
• Data: what do the raw values and variation show?
• Mechanism: what scientific model explains the result?
• Scope: which populations, conditions and ranges were actually tested?
• Replication: do independent methods recover compatible findings?
• Revision: what new evidence would change confidence?

When students can move through this checklist independently, advanced science vocabulary stops being a memorisation exercise. The terms become a map for deciding what a scientific claim means, how strongly it is supported and what still needs to be tested.

Final Calibration Note — Match Confidence to the Strength of the Evidence

Scientific language becomes most useful when it controls confidence. A single observation can justify reporting what was seen. Repeated measurements can increase confidence that a pattern is not one accidental reading. A controlled comparison can strengthen a causal interpretation. Independent replication can strengthen confidence that a finding survives a new setting. Converging results across methods and studies can support a broader scientific consensus. Each step adds something different; none should be skipped in the wording.

Maren therefore asks students to choose conclusion verbs deliberately. Observed describes a direct result. Associated with describes a relationship without automatically assigning cause. Supports connects evidence with a claim while leaving room for revision. Suggests can be appropriate where evidence is limited or preliminary. Causes requires stronger design and reasoning. The verb should reflect the evidence, not how strongly the student wants the answer to sound.

Iona asks what would reduce confidence: a calibration failure, biased sampling, an uncontrolled confounder, poor repeatability, incompatible replication or a better alternative explanation. Leonie asks what would increase confidence: clearer operational definitions, stronger controls, larger or more representative samples, independent verification and transparent analysis.

The final habit is simple: before writing a conclusion, state what the evidence definitely establishes, what it makes plausible and what it still does not show. That three-level distinction prevents both overconfidence and unnecessary scepticism. It also gives students a practical way to revise scientific writing when new evidence appears.

Precise scientific vocabulary helps students make claims that are accurate, testable, transparent and appropriately limited by evidence.