Two measurements can define a line. They cannot prove the world is linear.
An experiment records a voltage of 2 V and current of 0.10 A. A second reading gives 6 V and 0.30 A. Draw those points and there is exactly one straight line through both. The calculation is straightforward. The judgement is harder: do the readings justify treating every value between and beyond them as linear? What would a third reading tell us? How should the units shape our interpretation? Mathematics begins here, at the point where a tidy answer needs an honest condition.
The previous Vol 0323 on current–potential-difference characteristics contrasted straight and curved graphs. This mathematics chapter has one separate owner: recovering a straight-line equation from two labelled readings, checking additional observations and using the model with care. We are not re-teaching circuit laws or treating a straight line as proof of a physical mechanism.
Begin with a general graph rather than a memorised electrical example. Horizontal axis x might be time, distance, quantity bought or applied voltage. Vertical axis y might be cost, temperature or measured current. Whatever the setting, a straight-line model uses the form y = mx + c, where m is the rate of change of y for one unit of x and c is the model’s value when x = 0. Neither symbol has meaning until we give it units and a context.
The SEC G3 Mathematics boundary
The 2027 official SEAB G3 syllabus index identifies Mathematics K310. Its official syllabus includes linear graphs and gradients, coordinates, algebraic manipulation, interpretation of graphs and problem solving. This chapter uses those specified capabilities to reason through a particular kind of unknown relationship.
For K310, Paper 1 and Paper 2 are each 2 hours 15 minutes, 90 marks and 50% of the grade. Paper 1 contains short-answer questions, while Paper 2 includes longer and more complex problems with a real-world-situation final question. Show essential working even when an answer looks easy on a calculator. The official instructions and the individual question decide the required accuracy; a common default is three significant figures for non-exact numerical answers when no different instruction is given.
The worked tasks below are original examples, not an official paper or a prediction about the exact 2027 questions. Some are richer than an ordinary single examination item because they train the checks students often omit under time pressure. The English and Physics vocabulary surrounding a graph is a context for the mathematics, not evidence of a combined EMS examination paper.
Why the gradient is a ratio of changes
For two points (x₁,y₁) and (x₂,y₂), gradient m = (y₂ − y₁)/(x₂ − x₁), provided x₂ differs from x₁. Numerator and denominator must follow the same order. If you subtract y₂ − y₁ above the line, use x₂ − x₁ below it. This rule is not a trick: it measures how much the vertical quantity changes for a specified horizontal change.
For (2,7) and (5,19), m = (19 − 7)/(5 − 2) = 12/3 = 4. The line rises 4 units vertically for each positive horizontal unit. If x is minutes and y is litres, the gradient is 4 litres per minute. If x is dollars and y is kilograms, the unit becomes kilograms per dollar. The same mathematical number would describe a different relationship.
A positive gradient means y grows as x increases. A negative gradient means y falls. Zero gradient means no change in y across x. These conclusions refer to a graph’s direction, not automatically to whether a real-world event is good, bad, faster or safer.
Find the intercept only after finding the gradient
Substitute either known point into y = mx + c. With (2,7) and (5,19), m = 4, so 7 = 4(2) + c, giving c = −1. The line is y = 4x − 1. Check the second point: when x = 5, y = 20 − 1 = 19. This substitution is a fast way to catch a wrong subtraction or sign.
One point and its gradient can identify a line. Two points with different x-coordinates can identify a line. But a single point with no gradient cannot identify one: infinitely many lines pass through it. Two identical points do not become independent pieces of information merely because a question prints them twice.
A common mistake is to set c = 0 because the chart begins at the origin. The bottom-left corner of a printed graph is a layout choice, not evidence that the line passes through (0,0). Calculate the intercept. Then decide whether x = 0 is part of the situation’s meaningful domain before interpreting it.
Read the x and y axes before you use a familiar formula
Suppose the graph puts time on the vertical axis and distance horizontally. Its gradient is seconds per metre, not metres per second. Reversing the labels changes the ratio. A student who automatically says ‘speed’ may have inverted the quantity even though the arithmetic is right. Every gradient is a rate with a named numerator and denominator.
If the vertical axis is current in amperes and the horizontal axis potential difference in volts, the straight-line gradient is in amperes per volt. Only in an appropriate ohmic model does its reciprocal correspond to resistance in ohms. If the graph uses voltage vertically, the gradient in volts per ampere can equal resistance. These connections rely on the named electrical model; a random straight line on a graph is not a resistor.
Always convert milli-units first when the task expects standard units. If current is labelled mA and voltage V, the gradient may be calculated in mA/V, but it will be a thousand times the numerical value in A/V. Do not silently swap the units midway through a calculation.
A third reading tests the model, not the algebra
Two distinct points always determine a unique straight line. Three observations can reveal whether that line is an adequate description. Suppose measurements are (1,4), (3,8) and (5,13). The first two yield y = 2x + 2. The model predicts y = 12 when x = 5, but the third measurement is 13. At least one of these claims is then true: the readings have variation, there is an error or the underlying relationship is not exactly that line.
The difference between an observed y and the model’s predicted y is called a residual in this chapter. Observed minus predicted gives the sign: a positive residual means the observation lies above the line, a negative residual means below. Students can use this difference without learning a formal statistical regression method; it is simply a disciplined check.
A residual of 1 is not necessarily ‘small’. A difference of one litre may be large in a medicine measurement and negligible for a reservoir. Compare it with the graph scale, measurement resolution and decision being made. Likewise, a residual of zero at two construction points tells us only that the line was built to pass through them.
The temptation of making a claim from perfect-looking data
Students often write ‘there is a direct proportion’ when all they have shown is a straight line. A direct proportion of y to x requires y = kx, a straight line through the origin. A line with non-zero intercept, such as y = 2x + 5, is linear but not directly proportional. Doubling x does not simply double y because the fixed 5 remains.
More seriously, observations can fit a line even if the relationship changes outside the measured range. Interpolation estimates within the data range. Extrapolation predicts beyond it and carries greater risk when the model’s assumptions might fail. A filament lamp may look almost linear at very small voltages and become visibly curved as it heats. A three-point check is useful; it is not a guarantee of physical law.
A fair conclusion says, ‘The measurements are consistent with the straight-line model over this range’ or ‘The third reading differs by this amount from its prediction’. Those statements are stronger than confident but unsupported ‘the graph proves’.
Vertical lines, horizontal lines and data that do not determine a function
If x₁ = x₂ while the y-values differ, the gradient formula divides by zero. The points lie on a vertical line x = constant; they do not define a function y = mx + c with a finite m. Do not label the gradient ‘zero’. Zero gradient belongs to horizontal lines, where y remains constant while x changes.
Two readings with exactly the same coordinates do not identify a unique line. If data show (2,5) twice, any line through (2,5) fits both copies. The problem is lack of independent information, not a calculator failure.
Some word problems hide these cases by asking for a tariff or a conversion rate. Before solving, check that the horizontal values differ. If the values are identical, the information may be insufficient or inconsistent with a single-valued linear model. Saying so is mathematical reasoning.
A practical five-step working method
First, translate the two readings into ordered pairs and write their units. Second, find the gradient with a consistent subtraction order. Third, substitute one point into y = mx + c to recover the intercept. Fourth, substitute the other point as an independent algebra check. Fifth, use a new reading or a context condition to test whether the resulting model can be trusted.
Students who rush from step two to a final equation commonly omit the intercept. Students who know the formula but cannot label the units sometimes interpret the line backwards. A simple sentence beside the answer can repair both errors: ‘For every extra minute, the model adds 4 litres; at time zero, its extrapolated intercept is −1 litre.’ The second clause also exposes whether that extrapolated starting value is physically sensible.
Never force a line through the origin unless the question gives direct proportionality or the evidence supports it. Never use a scatter of measured points to claim an exact zero residual everywhere. The honest model starts with the data and ends with an appropriately limited conclusion.
From the PSLE number line to a secondary linear model
Primary Mathematics introduces a stable habit: state what the unknown represents and check that a result answers the question. The PSLE Mathematics word-problem bridge builds that habit. At secondary level, the unknown may be an entire relationship rather than one number: m and c together allow predictions at several inputs.
If negative values, coordinates or algebra remain shaky, revisit the Primary 6 Mathematics hub for familiar quantities, then work progressively on graph translation. The G3 Vol 0234 on intersecting lines answers a different question: where two models meet. Here we reconstruct one model and test whether it should be used.
The goal is not a new formula collection. It is the ability to build, test and limit a simple equation using observable evidence.
Forty worked line-reconstruction and evidence checks
These cases start with easy coordinates, then ask about units, extra readings, impossible values and model limits. Write the working before reading each answer. The aim is not memorising forty solutions; it is becoming able to recognise the same mathematical structure in unfamiliar settings.
1. Two points (2,6) and (5,15).
m = (15 − 6)/(5 − 2) = 3. From 6 = 3(2) + c, c = 0; therefore y = 3x. Substitute x = 5 to check that y = 15. Because the line passes through the origin, its two-coordinate model is directly proportional, though real-world validity still requires appropriate conditions.
2. Two points (1,7) and (3,15).
m = (15 − 7)/(3 − 1) = 4; c = 7 − 4 = 3. The model is y = 4x + 3. It is linear, not directly proportional. A doubling of x from 1 to 2 changes y from 7 to 11, not to 14. The intercept changes the scaling.
3. A line passes through (0,5) and (4,13).
The first point already supplies c = 5. m = (13 − 5)/4 = 2, giving y = 2x + 5. Checking x = 4 recovers 13. Do not calculate the gradient as 13/4 simply because the second point looks like a ratio; the first point says the line has a non-zero intercept.
4. The readings are (2,8) and (2,12).
Both have the same horizontal coordinate. The denominator x₂ − x₁ is zero, so there is no finite gradient for y = mx + c. The points lie on the vertical line x = 2. This is not a zero-gradient line, and it is not a single-valued y-function for x = 2.
5. The readings are (2,8) and (2,8).
Repeating the identical point supplies no new constraint. Infinitely many lines pass through it: y = 4x, y = x + 6 and y = 7x − 6 all do. The correct answer is insufficient information to identify a unique gradient and intercept.
6. Two points (2,7) and (6,−1).
m = (−1 − 7)/(6 − 2) = −8/4 = −2. From 7 = −2(2) + c, c = 11. Thus y = −2x + 11. The negative gradient means the predicted y falls by 2 units for each unit increase in x.
7. Two points (3,13) and (7,1).
m = (1 − 13)/(7 − 3) = −12/4 = −3. Substituting (3,13) gives c = 22; y = −3x + 22. The value −3 is not an error merely because x increases. It expresses a descending relationship.
8. The first point is (−2,5); the second is (2,13).
m = (13 − 5)/(2 − (−2)) = 8/4 = 2. Then c = 9, so y = 2x + 9. The subtraction of a negative number changes the denominator. Carry parentheses through the formula to avoid a denominator of zero or the wrong sign.
9. What if the coordinate order is reversed in the formula?
Using (5,19) and (2,7), m = (7 − 19)/(2 − 5) = −12/−3 = 4. The gradient is unchanged because both numerator and denominator reversed. The inconsistent method (7 − 19)/(5 − 2) would give −4 and reverse the line incorrectly.
10. A line passes through (4,9) with gradient 2.5.
Use 9 = 2.5(4) + c, so c = −1. The equation is y = 2.5x − 1. One point is enough when the gradient is known independently. Check that x = 4 gives 9 rather than confusing c with the supplied y-coordinate.
11. A line has gradient zero and includes (8,3).
The equation is y = 3. For any x on the line, y remains 3. The horizontal line is perfectly valid in y = mx + c with m = 0. Do not confuse it with a vertical line, whose gradient is undefined.
12. Does a negative intercept make a model invalid?
Not automatically. y = 4x − 1 can model measured quantities over x between 2 and 5 even though its extrapolated y-value at x = 0 is negative. If y represents a quantity that cannot be negative, the model may simply be inappropriate at x = 0. State the valid domain.
13. A taxi fare gives (2 km,$9) and (6 km,$17).
m = (17 − 9)/(6 − 2) = $2 per km. c = 9 − 2(2) = $5, so fare = 2(distance) + 5. The $5 intercept is a modelled fixed charge. Do not claim it is an officially announced fare; these are hypothetical values for algebra practice.
14. A reservoir data pair is (1 min,15 L), (4 min,39 L).
m = (39 − 15)/(4 − 1) = 8 L/min and c = 7 L. The equation V = 8t + 7 describes an initial modelled amount of 7 L and eight additional litres per minute. It remains conditional on a constant filling rate.
15. Voltage and current readings are (2 V,0.10 A) and (6 V,0.30 A).
On a current-vertical graph, m = (0.30 − 0.10)/(6 − 2) = 0.05 A/V, and c = 0 A. The mathematical model is I = 0.05V. For an ohmic resistor under unchanged conditions it corresponds to a resistance of 20 Ω, because V/I is the reciprocal slope.
16. A line through current readings has a non-zero current intercept.
An equation such as I = 0.05V + 0.02 A is a linear fit but not direct proportionality. It cannot automatically be read as an ordinary passive resistor law because the non-zero current at zero applied voltage needs a physical explanation. Keep mathematical fit separate from physical causation.
17. The readings are (1,4),(3,8),(5,13).
The first two give m = 2 and c = 2. At x = 5 the line predicts 12; the observed third value is 13, so the residual is +1. The third point contradicts an exact single straight line through all three points. Investigate whether measurement variation or a curved relationship is plausible.
18. The prediction is y = 2x + 1, but at x = 4 the observed value is 10.
The prediction is 9 and the residual observed minus predicted equals +1. A positive residual locates the observation above the model’s line. It does not by itself prove the model is useless; whether +1 matters depends on measurement accuracy and the decision.
19. A residual is −0.3. What does its sign tell us?
Observed minus predicted is negative, so the observation is 0.3 below the predicted vertical value, in whatever unit y uses. Do not call this a negative slope; residual describes a vertical departure at one x, while gradient describes the line’s overall rate.
20. Can a model have a zero residual at its two construction points?
Yes. The unique straight line through two points is defined to pass through them, so both calculated residuals are zero. This is not independent validation. Test a separate point, a new observation or a known physical constraint before making a wider claim.
21. Data x = 1 and 5 support a line. Predict at x = 3.
Since 3 lies within the observed x-range, the prediction is interpolation. It still assumes that a line is a suitable model between the measurements, but it requires a smaller extension of the assumption than predicting outside both points.
22. The same model predicts at x = 12.
This is extrapolation because 12 lies beyond the interval from 1 to 5. Even if substitution produces an exact algebraic number, the real-world forecast is not automatically exact. State a limit and check whether the process could change over the wider range.
23. The two y-values are 4 and 12, separated by 8 seconds.
If the two x-values represent times 2 s and 10 s, m = (12 − 4)/(10 − 2) = 1 unit per second. If instead the x interval is 4 s, the gradient would be 2 per second. Subtraction of actual coordinates, not a verbal shortcut, is essential.
24. A graph’s vertical values are in kilograms and horizontal values in hours.
The gradient units are kilograms per hour. If the modelled mass decreases, the gradient is negative kg/h. Report an amount of mass and a rate of mass change with different units. Confusing them can make a correct equation appear to describe the wrong quantity.
25. A graph marks 4 V per horizontal square and 0.2 A per vertical square.
One-square diagonal rise corresponds to 0.2 A/4 V = 0.05 A/V, not simply ‘gradient 1’. The numerical graph slope depends on axis scales. A correctly drawn straight line cannot rescue an answer that ignores the units printed on the grid.
26. The vertical axis shows milliamperes.
A gradient of 50 mA/V is 0.050 A/V, since 1000 mA = 1 A. Where an ohmic component has a straight-through-origin current–voltage graph, its corresponding resistance is 1/0.050 = 20 Ω. State the condition before giving the physical interpretation.
27. A line passes through (1.5,2.0) and (3.5,5.0).
m = (5.0 − 2.0)/(3.5 − 1.5) = 1.5. The intercept is c = 2.0 − 1.5(1.5) = −0.25. Thus y = 1.5x − 0.25. Check: at x = 3.5 the result is 5.0. Retaining decimals in the calculation helps avoid a premature-rounding error.
28. A model has y = −2x + 12. When is y zero?
Set 0 = −2x + 12 to get x = 6. If x represents time and y remaining stock, six time units is the modelled depletion time. The result is useful only if the negative linear trend persists to that point and quantities cannot become negative beyond it.
29. A sales graph uses x for packs and y for dollars.
An intercept of 18 dollars may represent a starting cost even when no packs are sold. A gradient of 4 dollars per pack represents a variable cost. These are distinct quantities. Reversing the labels would produce packs per dollar and change the question being answered.
30. A student assumes the first observed reading is the intercept.
An intercept belongs to x = 0, not to the smallest measured x unless that point is actually zero. For readings (2,9),(5,15), m = 2 and c = 5. The first measured y-value 9 is not the intercept; assuming it was would yield y = 2x + 9 and fail the given points.
31. The graph scale does not begin at zero.
A truncated axis can make a gentle difference look steep. If coordinates are labelled accurately, the gradient formula still works. Use numeric differences from printed values, not visual ‘rise in squares’ without the scale. A graphic’s dramatic appearance is not a mathematical result.
32. A model predicts −4 chairs for a small venue.
An algebraic line can yield −4 when extrapolated, but a negative number of physical chairs is not feasible. Do not round to −4 or pretend the interpretation makes sense. State that the linear model cannot represent this situation at the chosen input, then find the domain where predictions are meaningful.
33. Observed changes are unequal for equal x-steps.
If x goes 1,2,3,4 while y goes 2,4,7,11, the successive changes are 2,3,4. No single constant gradient describes all four readings exactly. Two chosen points still define a line, but fitting those two cannot establish an exact linear relationship for all four.
34. A line fitted to a heated lamp’s I–V data appears adequate near the origin.
The local approximation may be useful over a narrow range, but heating can change resistance as voltage rises. Do not extrapolate the same current-to-voltage gradient across the whole characteristic. A mathematical line is a description with a domain, not a licence to ignore the Physics from Vol 0323.
35. A question gives y = 3x + c and one point (−2,4).
Substitute: 4 = 3(−2) + c, hence c = 10. The model is y = 3x + 10. The negative x-coordinate should be kept inside parentheses. The fact that it is negative does not by itself make the point invalid in a coordinate-geometry problem.
36. Two lines share the same gradient but have different intercepts.
They are parallel and distinct. If y = 2x + 3 and y = 2x − 1, both gain two units vertically for each unit horizontally, but one is four units above the other. This helps compare proposed calibration lines without solving an intersection problem again.
37. A model is derived from measurements to one decimal place.
Keep sufficient precision during the algebra, then report an answer appropriate to the given precision and question instruction. Do not round the gradient repeatedly before calculating the intercept and prediction, because small errors can compound. Exact fractional expressions are often safer mid-solution.
38. A reader asks whether the slope is the ‘speed’.
Ask what y and x represent. If y is distance and x time, the gradient can be speed for a straight distance–time segment. If y is cost and x time, it is a cost increase per time. The mathematical calculation alone does not assign a physical name.
39. Why isn’t an exact equation automatically a good prediction?
A line can reproduce its two construction points perfectly while failing independent readings or physical constraints. An equation is exact within its algebra; the model is an approximation of a particular situation. Check another point, the input range, and the reason a constant gradient might hold.
40. What is the final three-part examination answer?
Give the equation y = mx + c with a transparent gradient and intercept calculation. Verify at least one given point. Then interpret the numerical gradient and intercept using their units and, if the task is applied, explain the limits of prediction. A bare equation earns no confidence unless its meaning survives these checks.
A full model answer: school display test
A student has hypothetical observations for a small educational display: at input settings x = 2 and x = 6, measured output readings y = 7 and y = 19. The questions are to reconstruct a straight-line candidate, predict y at x = 4, and assess a third observation y = 12 at x = 4. Work systematically: m = (19 − 7)/(6 − 2) = 3; c = 7 − 3(2) = 1; hence y = 3x + 1.
The model predicts y = 13 at x = 4, whereas the separate observation is 12. The residual is 12 − 13 = −1. That third observation is evidence that the three readings do not lie exactly on the proposed line. It does not automatically tell us whether the data are mistaken, the process is noisy or the relationship is curved. We have shown exactly what the mathematics supports and not more.
If this were a Science experiment, measurement resolution and physical mechanisms would matter. If it were a Mathematics exam question, the problem might only ask for the line and residual. Read the command word to decide how far to go. A good solution is complete for its task, not unnecessarily long.
Use cross-subject practice without mixing up examination ownership
The G3 Mathematics learning route is the owner for coordinate and algebraic problem solving. The G3 Science route is the owner for interpreting what a physical reading means. The Vol 0320 on reconstructing quadratics asks a deliberately different problem: how to recover a curved equation from sufficient conditions. Here two distinct points determine a straight-line equation, but independent evidence decides whether the line is an appropriate model.
The discipline that survives to the SEC examination is straightforward: preserve the labels, show the algebra, test the fit and explain the limitation when the question asks for one. Numbers are useful because they can be checked. They become misleading only when we forget which assumptions gave them meaning.
