Everything, on one sheet

The card

ax² + bx + c = a(x + b/2a)² − (b² − 4ac)/(4a). Vertex: x = −b/2a, height −(b² − 4ac)/(4a).
b² − 4ac: > 0 two distinct roots (crosses); = 0 one repeated root (touches); < 0 none (misses).
x = (−b ± √(b² − 4ac))/(2a) is that same completed square, solved in letters. Not in the booklet — memorise.
Parameters: can it be zero? Is the inequality strict? Never divide an inequality by the unknown.
Method mark, 3-term quadratic: factorisation, formula with values, or completing the square. Any one.

Why it works — Where $b^2 - 4ac$ actually comes from — it is not a rule, it is a leftover

Run the same procedure on letters instead of numbers, and the discriminant appears whether you were looking for it or not. Start with $ax^2 + bx + c$ and factor $a$ from the first two terms: $a\left(x^2 + \frac{b}{a}x\right) + c$. Halve the coefficient of $x$ inside the bracket: half of $\frac{b}{a}$ is $\frac{b}{2a}$, and $\left(x + \frac{b}{2a}\right)^2 = x^2 + \frac{b}{a}x + \frac{b^2}{4a^2}$, so $x^2 + \frac{b}{a}x = \left(x + \frac{b}{2a}\right)^2 - \frac{b^2}{4a^2}$. Multiply through by the $a$ outside and add the $c$ back: $ax^2 + bx + c = a\left(x + \frac{b}{2a}\right)^2 - \frac{b^2}{4a} + c$. Now combine those last two terms over the common denominator $4a$: $-\frac{b^2}{4a} + c = \frac{-b^2 + 4ac}{4a} = -\frac{b^2 - 4ac}{4a}$. So for every quadratic, without exception, $ax^2 + bx + c = a\left(x + \frac{b}{2a}\right)^2 - \frac{b^2 - 4ac}{4a}$. Nobody chose $b^2 - 4ac$; it is simply what was left over when the square was completed, and it is the only place the constant $c$ still appears. Read the identity twice and it tells you two separate things. First, the turning point of every quadratic is at $x = -\frac{b}{2a}$ with height $-\frac{b^2 - 4ac}{4a}$ — so for $a > 0$ the turning point lies below the x-axis exactly when $b^2 - 4ac > 0$, on it exactly when $b^2 - 4ac = 0$, and above it exactly when $b^2 - 4ac < 0$. Since an upward parabola's arms both go to $+\infty$, it crosses the axis twice, touches once, or misses entirely in precisely those three cases: the root count is not a separate rule bolted onto the discriminant, it is a restatement of where the vertex sits. Second, setting the whole thing equal to zero and solving gives $\left(x + \frac{b}{2a}\right)^2 = \frac{b^2 - 4ac}{4a^2}$. The denominator $4a^2$ is positive whatever $a$ is, so the right-hand side has the same sign as $b^2 - 4ac$ — and a real number squared can never be negative, so a real solution exists if and only if $b^2 - 4ac \geq 0$. That is the whole theorem, and it is one line of consequence from an identity you derived rather than memorised.

Traps — 6

discriminant-written-as-b-squared-plus-4ac

Confirmed directly on a real discriminant-with-parameter question: "a few used an incorrect expression b² + 4ac, gaining no marks" (Jan 2023, Q4). Note the severity — not a lost accuracy mark but no marks at all, because the method mark is for attempting the discriminant, and $b^2 + 4ac$ is not the discriminant. The derivation is the defence: the $-4ac$ comes from combining $-\frac{b^2}{4a}$ with $+c$ over a common denominator, so the minus sign is structural, not a convention to remember.

coefficient-of-x-not-squared-in-full

Confirmed on the same question, where the coefficient of $x$ was $6k$: "a small number did not square the 6, obtaining an upper limit of 10/3" (Jan 2023, Q4). Squaring $6k$ means squaring the whole thing — $36k^2$, not $6k^2$. The mis-squared version gives $6k^2 - 20k < 0$ and hence $0 < k < \frac{10}{3}$, exactly the wrong upper limit the report names, instead of the correct $0 < k < \frac{5}{9}$. The tell is that the wrong answer is a clean, plausible-looking fraction, so nothing about it looks wrong on the page.

parameter-zero-bound-dropped

The most common error on this question type, confirmed verbatim: "many candidates ignored the solution, k = 0, so the most common error seen was the absence of the lower bound 0 < k" (Jan 2023, Q4). The mechanism is almost always division: faced with $36k^2 - 20k < 0$, students divide through by $k$, which discards the root $k = 0$ and is in any case not a legal operation on an inequality whose divisor has unknown sign. Factorise instead and both critical values survive.

boundary-included-when-the-inequality-is-strict

Confirmed on the same question: "some gave their final answer as 0 ≤ k ≤ 5/9, losing the final mark" (Jan 2023, Q4). The distinction is exact, not stylistic. At $b^2 - 4ac = 0$ there is one repeated real root, which is a real root — so "no real roots" excludes the boundary and "two distinct real roots" excludes it too. Only "at least one real root" or "real roots" (unqualified) includes it. Read which of those phrasings the question used before choosing between $<$ and $\leq$.

discriminant-applied-before-the-quadratic-exists

Confirmed on a question where the quadratic had to be constructed first: "This question proved challenging for some candidates, who failed to realise that they needed to form a quadratic in x in order to use the discriminant" (Jan 2019, Q9). The same family of error runs in the other direction after a substitution — having formed and solved a quadratic in a substituted variable, students stop: "a significant number of responses did not go any further than finding solutions to the quadratic in p" (Jan 2025, Q4(b)), a trap confirmed independently across two separate series. Both are the same discipline failing at different ends: know which variable you are working in, and know which one the question asked about.

answer-from-the-calculator-with-no-method-shown

WMA11 is a calculator paper with explicit no-calculator sub-parts, and quadratics are exactly where they land. The Jan 2023 paper carries "(Solutions relying on calculator technology are not acceptable.)" on several parts, and one instructs: "In this question you must show all stages of your working. Solutions relying on calculator technology are not acceptable." An examiner report on a solve-the-quadratic part records the consequence: "Quite a number used the formula or factorised to solve the quadratic, whilst others did not show a method and found their solution directly from a calculator... this approach could have been penalised" (Jan 2025, Q4(b)). Treat the warning phrase as a cue rather than boilerplate — it marks the parts where the algebra itself is the thing being examined.

The card

ax² + bx + c = a(x + b/2a)² − (b² − 4ac)/(4a). Vertex: x = −b/2a, height −(b² − 4ac)/(4a).
b² − 4ac: > 0 two distinct roots (crosses); = 0 one repeated root (touches); < 0 none (misses).
x = (−b ± √(b² − 4ac))/(2a) is that same completed square, solved in letters. Not in the booklet — memorise.
Parameters: can it be zero? Is the inequality strict? Never divide an inequality by the unknown.
Method mark, 3-term quadratic: factorisation, formula with values, or completing the square. Any one.

Why it works — Where $b^2 - 4ac$ actually comes from — it is not a rule, it is a leftover

Run the same procedure on letters instead of numbers, and the discriminant appears whether you were looking for it or not. Start with $ax^2 + bx + c$ and factor $a$ from the first two terms: $a\left(x^2 + \frac{b}{a}x\right) + c$. Halve the coefficient of $x$ inside the bracket: half of $\frac{b}{a}$ is $\frac{b}{2a}$, and $\left(x + \frac{b}{2a}\right)^2 = x^2 + \frac{b}{a}x + \frac{b^2}{4a^2}$, so $x^2 + \frac{b}{a}x = \left(x + \frac{b}{2a}\right)^2 - \frac{b^2}{4a^2}$. Multiply through by the $a$ outside and add the $c$ back: $ax^2 + bx + c = a\left(x + \frac{b}{2a}\right)^2 - \frac{b^2}{4a} + c$. Now combine those last two terms over the common denominator $4a$: $-\frac{b^2}{4a} + c = \frac{-b^2 + 4ac}{4a} = -\frac{b^2 - 4ac}{4a}$. So for every quadratic, without exception, $ax^2 + bx + c = a\left(x + \frac{b}{2a}\right)^2 - \frac{b^2 - 4ac}{4a}$. Nobody chose $b^2 - 4ac$; it is simply what was left over when the square was completed, and it is the only place the constant $c$ still appears. Read the identity twice and it tells you two separate things. First, the turning point of every quadratic is at $x = -\frac{b}{2a}$ with height $-\frac{b^2 - 4ac}{4a}$ — so for $a > 0$ the turning point lies below the x-axis exactly when $b^2 - 4ac > 0$, on it exactly when $b^2 - 4ac = 0$, and above it exactly when $b^2 - 4ac < 0$. Since an upward parabola's arms both go to $+\infty$, it crosses the axis twice, touches once, or misses entirely in precisely those three cases: the root count is not a separate rule bolted onto the discriminant, it is a restatement of where the vertex sits. Second, setting the whole thing equal to zero and solving gives $\left(x + \frac{b}{2a}\right)^2 = \frac{b^2 - 4ac}{4a^2}$. The denominator $4a^2$ is positive whatever $a$ is, so the right-hand side has the same sign as $b^2 - 4ac$ — and a real number squared can never be negative, so a real solution exists if and only if $b^2 - 4ac \geq 0$. That is the whole theorem, and it is one line of consequence from an identity you derived rather than memorised.

Traps — 6

discriminant-written-as-b-squared-plus-4ac

Confirmed directly on a real discriminant-with-parameter question: "a few used an incorrect expression b² + 4ac, gaining no marks" (Jan 2023, Q4). Note the severity — not a lost accuracy mark but no marks at all, because the method mark is for attempting the discriminant, and $b^2 + 4ac$ is not the discriminant. The derivation is the defence: the $-4ac$ comes from combining $-\frac{b^2}{4a}$ with $+c$ over a common denominator, so the minus sign is structural, not a convention to remember.

coefficient-of-x-not-squared-in-full

Confirmed on the same question, where the coefficient of $x$ was $6k$: "a small number did not square the 6, obtaining an upper limit of 10/3" (Jan 2023, Q4). Squaring $6k$ means squaring the whole thing — $36k^2$, not $6k^2$. The mis-squared version gives $6k^2 - 20k < 0$ and hence $0 < k < \frac{10}{3}$, exactly the wrong upper limit the report names, instead of the correct $0 < k < \frac{5}{9}$. The tell is that the wrong answer is a clean, plausible-looking fraction, so nothing about it looks wrong on the page.

parameter-zero-bound-dropped

The most common error on this question type, confirmed verbatim: "many candidates ignored the solution, k = 0, so the most common error seen was the absence of the lower bound 0 < k" (Jan 2023, Q4). The mechanism is almost always division: faced with $36k^2 - 20k < 0$, students divide through by $k$, which discards the root $k = 0$ and is in any case not a legal operation on an inequality whose divisor has unknown sign. Factorise instead and both critical values survive.

boundary-included-when-the-inequality-is-strict

Confirmed on the same question: "some gave their final answer as 0 ≤ k ≤ 5/9, losing the final mark" (Jan 2023, Q4). The distinction is exact, not stylistic. At $b^2 - 4ac = 0$ there is one repeated real root, which is a real root — so "no real roots" excludes the boundary and "two distinct real roots" excludes it too. Only "at least one real root" or "real roots" (unqualified) includes it. Read which of those phrasings the question used before choosing between $<$ and $\leq$.

discriminant-applied-before-the-quadratic-exists

Confirmed on a question where the quadratic had to be constructed first: "This question proved challenging for some candidates, who failed to realise that they needed to form a quadratic in x in order to use the discriminant" (Jan 2019, Q9). The same family of error runs in the other direction after a substitution — having formed and solved a quadratic in a substituted variable, students stop: "a significant number of responses did not go any further than finding solutions to the quadratic in p" (Jan 2025, Q4(b)), a trap confirmed independently across two separate series. Both are the same discipline failing at different ends: know which variable you are working in, and know which one the question asked about.

answer-from-the-calculator-with-no-method-shown

WMA11 is a calculator paper with explicit no-calculator sub-parts, and quadratics are exactly where they land. The Jan 2023 paper carries "(Solutions relying on calculator technology are not acceptable.)" on several parts, and one instructs: "In this question you must show all stages of your working. Solutions relying on calculator technology are not acceptable." An examiner report on a solve-the-quadratic part records the consequence: "Quite a number used the formula or factorised to solve the quadratic, whilst others did not show a method and found their solution directly from a calculator... this approach could have been penalised" (Jan 2025, Q4(b)). Treat the warning phrase as a cue rather than boilerplate — it marks the parts where the algebra itself is the thing being examined.

The card

Parameter in a, b or c turns b²−4ac from a number into an expression — solve FOR it, don't just evaluate it.
Squaring a term with the parameter inside it means squaring the WHOLE coefficient: (6k)² = 36k², not 6k².
Parameter multiplies a? Check it can't be zero — separately from the discriminant, even if the stem never restates it.
'No real roots' / 'two distinct roots' exclude b²−4ac=0. 'At least one root' / 'does not cross' include it.
Never divide an inequality by the parameter — factorise instead, or a root vanishes and the sign may flip unseen.

Why it works — The discriminant of the discriminant — solving "no real roots" as its own quadratic inequality

Once $a$, $b$ and $c$ are substituted in — possibly with the parameter inside one, two, or all three of them — $b^2 - 4ac$ becomes an algebraic expression in that parameter, and the condition supplied by the question's wording ($>0$, $=0$, $<0$, $\geq 0$, $\leq 0$) turns it into an equation or inequality to solve. When both $a$ and $b$ contain the parameter — the shape that produces every trap in this lesson at once, and the shape both worked examples below use — that expression is itself a QUADRATIC in the parameter: expand it out and there is a squared term, because $b$ was squared to get there. Solving a quadratic INEQUALITY needs one step beyond solving a quadratic EQUATION. The first part — finding the critical values — is exactly the same three-route problem as solving any three-term quadratic: factorise, use the formula, or complete the square, and the general principles for pure marking credit all three equally, whichever variable is actually being solved for. But a second question then has to be answered, one a plain equation never asks: which side of those critical values — or the region between them — actually satisfies the inequality. The safest way to answer it is the reasoning this course has already used once, for a quadratic in $x$: the expression, regarded as a function of the parameter, is itself an upward or downward parabola (upward here, always — a squared term coming from $b^2$ can never carry a negative coefficient), so it is positive outside its own roots and negative between them, and one substituted test value in each region confirms which. What must never happen, at any point in this process, is dividing the inequality through by the parameter to simplify it first — that move discards whichever root sits at parameter-equals-zero and is not valid in general, since dividing an inequality by a quantity of unknown sign can silently flip its direction. Every piece of this paragraph is the exact machinery the previous lesson used to find where a quadratic in $x$ crosses its axis, applied one level up: a quadratic inequality in the parameter, solved by finding where a second, related quadratic crosses ITS axis.

Traps — 7

discriminant-written-as-b-squared-plus-4ac

Confirmed directly on a real discriminant-with-parameter question: "a few used an incorrect expression b² + 4ac, gaining no marks" (Jan 2023, Q4). This is not a one-mark slip — the method mark itself is for attempting $b^2 - 4ac$, so a candidate who evaluates $b^2 + 4ac$ scores nothing on the line, however careful the rest of the working is. The sign is not a convention to remember: it falls out of combining $-\frac{b^2}{4a}$ with $+c$ when the square is completed (see the previous lesson's derivation), which is why getting it wrong here usually means the derivation was never actually understood, only the three-case outcome.

coefficient-not-squared-in-full

Confirmed on the same question, where the coefficient of $x$ was $6k$: "a small number did not square the 6, obtaining an upper limit of 10/3" (Jan 2023, Q4). Squaring $6k$ means squaring the whole product, $36k^2$, not $6k^2$. This lesson's own worked example shows the identical mechanism with different numbers: mis-squaring $12p$ to $12p^2$ instead of $144p^2$ produces the equally clean-looking wrong bound $p < 4$ in place of the correct $p < \frac{1}{3}$ — the tell is always that the wrong answer looks exactly as plausible as the right one, so checking the squaring is not optional the moment a parameter sits inside the $b$ term.

parameter-zero-bound-dropped

The most common error confirmed on this question, verbatim: "many candidates ignored the solution, k = 0, so the most common error seen was the absence of the lower bound 0 < k" (Jan 2023, Q4). The mechanism is almost always division: faced with an inequality like $144p^2 - 48p < 0$, it is tempting to divide through by $p$, which silently discards the root $p = 0$ and is in any case not a legal move on an inequality whose divisor has unknown sign. Factorise instead of dividing, and both critical values survive automatically.

boundary-included-when-the-inequality-is-strict

Confirmed on the same question: "some gave their final answer as 0 ≤ k ≤ 5/9, losing the final mark" (Jan 2023, Q4). The distinction is exact, not stylistic: at $b^2 - 4ac = 0$ there is one repeated real root, which is a real root, so "no real roots" and "two distinct real roots" both exclude that boundary — while "at least one real root", "real roots" unqualified, and "does not cross the x-axis" all include it. The marked-solution question above is built to show a boundary that is genuinely INCLUDED, in part (c) — reading which case is which matters in both directions, not only the strict one.

discriminant-applied-before-the-quadratic-exists

Confirmed on a question where the quadratic first had to be built: "This question proved challenging for some candidates, who failed to realise that they needed to form a quadratic in x in order to use the discriminant" (Jan 2019, Q9). The trap is not the discriminant technique itself — it is not recognising that a discriminant question is present at all. A term in $\frac{1}{x}$ cleared by multiplying through, a substitution, or a line set equal to a curve can all hide a three-term quadratic one algebraic step below the surface; the marked-solution question above is built around exactly this pattern, and part (a) exists specifically to make that hidden step visible before the discriminant is ever taken.

not-every-term-multiplied-through

Confirmed on the real question this lesson's marked-solution is modelled on, Jan 2019 Q9 (the equation $\frac{3}{x} + 5 = -2x + c$, "has no real roots, find the range of possible values of c"): "Examiners occasionally saw responses in which all terms except c were multiplied by x. This led to a discriminant which was linear in c and limited further progress could be made." The real mark scheme is explicit that this is a named risk, not a hypothetical one — it even states its own method mark "may be implied by later work" and explicitly "condone[s] one term not being multiplied by x for this mark but all four terms must be on one side," meaning the attempt still banks that one mark even when a term is missed, but every mark after it is lost, because an expression missing one multiplication is not actually the three-term quadratic the discriminant technique needs. This is the exact mechanism the marked-solution's part (a) drills: every term on both sides has to be multiplied by $x$ — not only the fractional one — or the "quadratic" that results is not genuinely quadratic in the variable being eliminated.

adjusted-working-to-fit-the-printed-answer

Part (a) above is an "ag" (answer given) mark — the target line $x^2 - cx + 4 = 0$ is printed on the page before any working is written. Confirmed on a different WMA paper's own "show that" question, Jan 2021 Q10: "The given answer here persuaded many candidates to 'adjust' their working following obvious mistakes." This is a different KIND of wrong from every other trap on this list: the final line can be completely, word-for-word correct and still earn nothing, because the mark is for a genuine forward derivation, not for a page that ends in the right place. The lesson's own warrant-check gate on part (a) exists specifically to make this distinction concrete rather than abstract — writing your own working BEFORE comparing it against the model line is what separates deriving $x^2 - cx + 4 = 0$ from reverse-engineering a plausible path to it.

The card

Rearrange the LINEAR equation first (always possible with no extra technique); substitute into the other equation.
Line meets curve: b² − 4ac of the resulting equation. >0 two points; =0 tangent (touches once); <0 never meet.
Two solves, not one: solve for x, then back-substitute into the simpler original equation for y. A coordinate needs both.
Any valid method for the resulting quadratic earns the method mark — factorisation, formula, or completing the square.
The simultaneous-equations route always works, but isn't always fastest — check for a geometric shortcut (e.g. a parallelogram's diagonals bisect each other) before committing.

Why it works — Why substitution turns two equations into one — and why the leftover discriminant counts how many times they meet

A simultaneous solution to two equations in $x$ and $y$ is a single pair $(x,y)$ that makes BOTH equations true at once — geometrically, a point that lies on both graphs. Suppose one of the two equations is linear, so it can always be rearranged into the form $y = (\text{an expression in } x)$ with no further complication. At any point that lies on BOTH graphs, that same expression for $y$ is also what the OTHER equation's own $y$ has to equal there — the two graphs meet exactly where their heights agree. Substituting the linear expression for $y$ directly into the other equation doesn't throw away any information about points that solve both: it simply forces the shared condition 'these two $y$'s are the same number' to become visible as a single equation in $x$ alone, because $y$ itself has been written out of the problem. Whatever value(s) of $x$ satisfy that equation are exactly the $x$-coordinates of every point where the two graphs genuinely meet — nothing more, nothing fewer. When the second equation is a quadratic curve, the equation left in $x$ alone is itself a quadratic (unless the leading terms happen to cancel, which is worth checking, since it changes which technique applies), and its discriminant does exactly the job it has done since spec 1.4: $b^2-4ac>0$ means two distinct real values of $x$, hence two genuine crossing points; $b^2-4ac=0$ means one repeated value of $x$, hence a single point where the line just touches the curve without passing through it — a tangent, and a genuinely different geometric fact from crossing; $b^2-4ac<0$ means no real value of $x$ at all, hence no point in common — the line and the curve simply never meet, however far the picture is drawn. Nothing new is being learned here about the discriminant itself; what's new is recognising that it still applies once you've done the extra step of building the quadratic through substitution, rather than being handed one directly.

Traps — 6

solving-simultaneously-when-a-faster-route-exists

The one directly-evidenced trap in this lesson, and the reason the method-comparison block above exists at all. A real WMA11 examiner report on a rectangle fourth-vertex question — solvable by finding two side equations and solving them simultaneously, by using the shared midpoint of the diagonals, or by a direct vector between two known vertices — records all three routes performing very differently in practice: candidates "who attempted to find equations for RS and PS and solving simultaneously rarely produced enough work to score full marks," the diagonal-midpoint route was "a more unusual but often successful approach," and "candidates who worked by counting units between points or used vector methods were most confident in reaching the correct, or partially correct, answer" of all three (Jan 2023, Q2). None of the three methods is wrong, and a mark scheme credits all equally — the difference is entirely in how much has to go right, unbroken, before an answer appears. Before committing to the simultaneous-equations route on a coordinate-geometry question, check the shape for a geometric shortcut first.

stops-at-the-x-coordinate

Now directly confirmed on a genuine spec 1.6 line-meets-curve substitution question, not just by analogy to a different topic: Jan 2023 Q7(d) — a line meeting a reciprocal curve, re-extracted for this lesson's own mark-scheme-bullet coverage audit — records that 'a small number forgot to substitute their value for x to find the y coordinate,' the exact failure this trap names, on the exact technique this lesson teaches. The same discipline is independently confirmed a second time on a different WMA11 topic entirely, this course's own hidden-quadratic-substitution lesson: 'a significant number of responses did not go any further than finding solutions to the quadratic in [the substituted variable]' (Jan 2025, Q4(b) — the same trap independently confirmed on the Jan 2023 Q5 equation that lesson's marked-solution is built on, though that exact quoted sentence is Jan 2025's own). There, the substituted letter isn't $x$; here, the equation substitution produces genuinely IS solved for $x$ — but $x$ still isn't the whole answer, because the question asked for point(s) of intersection or coordinates, and a coordinate needs a $y$ too. Two independent series, two different question types, the same discipline: solve the equation the substitution produced, then don't stop.

over-complicating-a-known-value-substitution

Confirmed directly on a real spec 1.6 substitution question: Jan 2023 Q7(c) gave a line and a curve meeting at two points, named one point's x-coordinate, and asked only for the constant k — the fast route is to substitute the known x-value straight into either equation and solve the single resulting equation in k. The examiner report records many candidates taking a slower, more error-prone route instead: 'a lot of students made this more complicated by equating the expressions for y and rearranging into a quadratic before substituting in x = −4. More errors in the algebra were found when taking this approach.' None of this lesson's own worked examples hand you a known coordinate this way, but the discipline generalises: when a question already gives you one variable's value, substitute it in immediately, rather than doing algebra that isn't needed to use it.

reaching-for-the-discriminant-when-a-value-is-already-known

The mechanism block above trains one reflex hard — compute the discriminant of the equation substitution produces to find out how many times a line and curve meet. A real WMA11 examiner report on the one genuine line-meets-curve substitution question this course's research has located (Jan 2023 Q7(d), a line meeting a reciprocal curve at two named points) records candidates over-applying that exact reflex where it doesn't belong: 'some did not use their value for k and tried to use the discriminant, with little success.' The question had already stated the line meets the curve at two points and had already supplied enough information to find k directly — the discriminant answers 'how many intersections exist,' not 'what is this specific unknown constant,' and reaching for it once the number of intersections is already settled is wasted, often unproductive work. The two tools solve genuinely different questions: use the discriminant to count intersections when the equations are otherwise complete; substitute a known value directly when the question hands you one.

sign-error-when-collecting-terms-after-substitution

Mechanically motivated, not itself quoted for this exact question type in the research pass this lesson draws on. Substituting a linear expression like $y=mx+c$ into a quadratic and rearranging into $=0$ form requires moving BOTH the $mx$ term and the $c$ term across the equals sign — and it is exactly as easy to flip only one of their signs as it is to flip neither. The check that catches it costs one line: substitute your two claimed $x$-values back into the ORIGINAL, unrearranged linear and quadratic equations, not into your own rearranged quadratic — a sign error made during rearrangement doesn't show up when you check against the very equation the error is hiding inside.

assumes-a-line-and-a-curve-always-meet-twice

Mechanically motivated rather than directly quoted for this specific question type, and the natural mirror of the same over-generalisation the quadratic-functions lesson already names for a curve meeting the x-axis: a positive discriminant is not guaranteed just because the two original equations 'looked like' they should cross. The equation substitution produces can just as easily have $b^2-4ac$ equal to zero (the line is a tangent — one point, not two) or negative (no real intersection at all) as it can be positive. Compute the discriminant of what substitution actually produces before assuming the number of intersection points — the same discipline spec 1.4 already teaches, applied here to a new equation you had to build first.

The card

ax² + bx + c = a(x + b/2a)² − (b² − 4ac)/(4a). Vertex: x = −b/2a, height −(b² − 4ac)/(4a).
b² − 4ac: > 0 two distinct roots (crosses); = 0 one repeated root (touches); < 0 none (misses).
x = (−b ± √(b² − 4ac))/(2a) is that same completed square, solved in letters. Not in the booklet — memorise.
Parameters: can it be zero? Is the inequality strict? Never divide an inequality by the unknown.
Method mark, 3-term quadratic: factorisation, formula with values, or completing the square. Any one.

Why it works — Where $b^2 - 4ac$ actually comes from — it is not a rule, it is a leftover

Run the same procedure on letters instead of numbers, and the discriminant appears whether you were looking for it or not. Start with $ax^2 + bx + c$ and factor $a$ from the first two terms: $a\left(x^2 + \frac{b}{a}x\right) + c$. Halve the coefficient of $x$ inside the bracket: half of $\frac{b}{a}$ is $\frac{b}{2a}$, and $\left(x + \frac{b}{2a}\right)^2 = x^2 + \frac{b}{a}x + \frac{b^2}{4a^2}$, so $x^2 + \frac{b}{a}x = \left(x + \frac{b}{2a}\right)^2 - \frac{b^2}{4a^2}$. Multiply through by the $a$ outside and add the $c$ back: $ax^2 + bx + c = a\left(x + \frac{b}{2a}\right)^2 - \frac{b^2}{4a} + c$. Now combine those last two terms over the common denominator $4a$: $-\frac{b^2}{4a} + c = \frac{-b^2 + 4ac}{4a} = -\frac{b^2 - 4ac}{4a}$. So for every quadratic, without exception, $ax^2 + bx + c = a\left(x + \frac{b}{2a}\right)^2 - \frac{b^2 - 4ac}{4a}$. Nobody chose $b^2 - 4ac$; it is simply what was left over when the square was completed, and it is the only place the constant $c$ still appears. Read the identity twice and it tells you two separate things. First, the turning point of every quadratic is at $x = -\frac{b}{2a}$ with height $-\frac{b^2 - 4ac}{4a}$ — so for $a > 0$ the turning point lies below the x-axis exactly when $b^2 - 4ac > 0$, on it exactly when $b^2 - 4ac = 0$, and above it exactly when $b^2 - 4ac < 0$. Since an upward parabola's arms both go to $+\infty$, it crosses the axis twice, touches once, or misses entirely in precisely those three cases: the root count is not a separate rule bolted onto the discriminant, it is a restatement of where the vertex sits. Second, setting the whole thing equal to zero and solving gives $\left(x + \frac{b}{2a}\right)^2 = \frac{b^2 - 4ac}{4a^2}$. The denominator $4a^2$ is positive whatever $a$ is, so the right-hand side has the same sign as $b^2 - 4ac$ — and a real number squared can never be negative, so a real solution exists if and only if $b^2 - 4ac \geq 0$. That is the whole theorem, and it is one line of consequence from an identity you derived rather than memorised.

Traps — 6

discriminant-written-as-b-squared-plus-4ac

Confirmed directly on a real discriminant-with-parameter question: "a few used an incorrect expression b² + 4ac, gaining no marks" (Jan 2023, Q4). Note the severity — not a lost accuracy mark but no marks at all, because the method mark is for attempting the discriminant, and $b^2 + 4ac$ is not the discriminant. The derivation is the defence: the $-4ac$ comes from combining $-\frac{b^2}{4a}$ with $+c$ over a common denominator, so the minus sign is structural, not a convention to remember.

coefficient-of-x-not-squared-in-full

Confirmed on the same question, where the coefficient of $x$ was $6k$: "a small number did not square the 6, obtaining an upper limit of 10/3" (Jan 2023, Q4). Squaring $6k$ means squaring the whole thing — $36k^2$, not $6k^2$. The mis-squared version gives $6k^2 - 20k < 0$ and hence $0 < k < \frac{10}{3}$, exactly the wrong upper limit the report names, instead of the correct $0 < k < \frac{5}{9}$. The tell is that the wrong answer is a clean, plausible-looking fraction, so nothing about it looks wrong on the page.

parameter-zero-bound-dropped

The most common error on this question type, confirmed verbatim: "many candidates ignored the solution, k = 0, so the most common error seen was the absence of the lower bound 0 < k" (Jan 2023, Q4). The mechanism is almost always division: faced with $36k^2 - 20k < 0$, students divide through by $k$, which discards the root $k = 0$ and is in any case not a legal operation on an inequality whose divisor has unknown sign. Factorise instead and both critical values survive.

boundary-included-when-the-inequality-is-strict

Confirmed on the same question: "some gave their final answer as 0 ≤ k ≤ 5/9, losing the final mark" (Jan 2023, Q4). The distinction is exact, not stylistic. At $b^2 - 4ac = 0$ there is one repeated real root, which is a real root — so "no real roots" excludes the boundary and "two distinct real roots" excludes it too. Only "at least one real root" or "real roots" (unqualified) includes it. Read which of those phrasings the question used before choosing between $<$ and $\leq$.

discriminant-applied-before-the-quadratic-exists

Confirmed on a question where the quadratic had to be constructed first: "This question proved challenging for some candidates, who failed to realise that they needed to form a quadratic in x in order to use the discriminant" (Jan 2019, Q9). The same family of error runs in the other direction after a substitution — having formed and solved a quadratic in a substituted variable, students stop: "a significant number of responses did not go any further than finding solutions to the quadratic in p" (Jan 2025, Q4(b)), a trap confirmed independently across two separate series. Both are the same discipline failing at different ends: know which variable you are working in, and know which one the question asked about.

answer-from-the-calculator-with-no-method-shown

WMA11 is a calculator paper with explicit no-calculator sub-parts, and quadratics are exactly where they land. The Jan 2023 paper carries "(Solutions relying on calculator technology are not acceptable.)" on several parts, and one instructs: "In this question you must show all stages of your working. Solutions relying on calculator technology are not acceptable." An examiner report on a solve-the-quadratic part records the consequence: "Quite a number used the formula or factorised to solve the quadratic, whilst others did not show a method and found their solution directly from a calculator... this approach could have been penalised" (Jan 2025, Q4(b)). Treat the warning phrase as a cue rather than boilerplate — it marks the parts where the algebra itself is the thing being examined.

The card

u = a^x: a^(x+n) = a^n · u, a^(x-n) = u ÷ a^n, a^(nx) = u^n. Every term becomes a power of u — that's spec 1.1, not new rules.
Two solves, not one: solve the quadratic for u, THEN solve a^x = u for x. The question asked for x.
a^x > 0 for every real x, any a > 0. A negative value of u is rejected, with a reason — never reported as an answer.
No logs needed: match indices directly (1/9 = 3^-2). WMA11 substitution questions are built this way, since P1 doesn't teach logs.
'Show all stages' questions penalise a calculator-only route even when the value of x is right.

Why it works — Why the substitution always works — and where it stops working

Take the general case algebraically rather than trusting that it works because it happened to work on one example. Suppose every term in an equation has the same base $a>0$, each raised to an exponent of the form $x+k$ for some rational constant $k$ (a plain $x$ is the case $k=0$; a coefficient like $2x$ is handled separately below). By the addition law, $a^{x+k}=a^k \cdot a^x$. Set $u=a^x$: every such term becomes $a^k u$, a constant multiple of $u$ — never a higher power, because the exponent's VARIABLE part is always exactly $x$, contributing exactly one factor of $u$. A term with exponent $2x$ is different in exactly one respect: by the power law, $a^{2x}=(a^x)^2=u^2$, contributing two factors of $u$ instead of one. Collect every term of the equation as a multiple of $u^2$, $u$, or a constant (a term with exponent $0\cdot x + k$, i.e. no $x$ at all, is already a plain number), and what remains is, by construction, a polynomial in $u$ of degree at most 2 — a quadratic, exactly because the highest power of $x$ appearing in any exponent was $2x$ and nothing higher. Change that highest exponent to $3x$ and the same argument produces a CUBIC in $u$, not a quadratic; WMA11's hidden-quadratic questions are hidden quadratics specifically because every exponent stays at or below $2x$, not because substitution itself is limited to quadratics.

Traps — 6

stops-at-the-substituted-variable

The named trap this lesson exists to fix, confirmed independently across two series. A real WMA11 examiner report states it directly: "a significant number of responses did not go any further than finding solutions to the quadratic in [the substituted variable]" (Jan 2025, Q4(b)) — the same failure the Jan 2023 Q5 equation this lesson's marked-solution is built on documents in its own report. The substituted letter is never the thing the question asked for; it is scaffolding on the way to it.

exponent-shift-mapped-as-a-power-not-a-multiple

Confirmed verbatim on the real question this lesson's marked-solution reproduces: "Common misconceptions included, $3^{x+2}=p^2$ and $3^{x-1}=p^{-1}$" (Jan 2023, Q5). Both errors treat a SHIFTED exponent as though it behaved like a raised power of the substituted letter — $3^{x+2}$ becomes $p^2$ (which is what $3^{2x}$ would give, not $3^{x+2}$), and $3^{x-1}$ becomes $p^{-1}$ (which is what $3^{-x}$ would give). The correct laws are additive on the base, not on the letter after substitution: $a^{x+k}=a^k\cdot u$, never $u^k$.

negative-value-of-u-not-rejected

Not itself a phrase the research bank's examiner-report extracts quote by name — this is mathematical necessity, not a documented misconception, and is named here because leaving it out would understate what a full solution requires. $u=a^x$ for $a>0$ can never be negative, so a value of $u$ that comes out negative when solving the quadratic is not a second solution waiting to be converted — it is not a solution at all, and the correct response states the rejection with a reason rather than omitting it silently. The real Jan 2023 Q5 equation has exactly this shape (p=-3 is one of its two roots).

answer-from-the-calculator-with-no-method-shown

The real Jan 2023 Q5 question this lesson's marked-solution is built on carries the instruction "In this question you must show all stages of your working. Solutions relying on calculator technology are not acceptable" (verbatim, from the question paper itself), and the official mark scheme for this exact question backs the instruction with a direct, unambiguous rule of its own in part (b)'s general guidance: "Answers just written down scores M0" (WMA11_01_MS_2301, Q5). Re-checking a related Jan 2025 quote against its own report during this pass found it does NOT transfer cleanly to this trap: that series' Q4(b) comment on calculator-only responses ends "...this approach could have been penalised on a different occasion" — the report's own general comments confirm the explicit show-your-working warning that series was printed only on Q2(a) and 5(b), not Q4, so a prior pass's use of that quote here overstated what actually happened on that specific question. The Jan 2023 Q5 mark scheme's own M0 rule is the correct, unhedged source for this trap. Treat the printed warning as a cue that the algebra itself, not just the final value, is what earns the marks.

quadratic-not-yet-formed-before-a-quadratic-technique-is-applied

A related but distinct discipline failure, confirmed on a different real question (not a substitution question, but the same underlying skill): "This question proved challenging for some candidates, who failed to realise that they needed to form a quadratic in x in order to use the discriminant" (Jan 2019, Q9). The substitution trap and this one are mirror images — one is about recognising you HAVE built a quadratic and still needing to finish the job; this one is about failing to recognise a quadratic needs building in the first place, before any quadratic technique (discriminant, factorising, the formula) has anything to act on.

hence-ignored-original-equation-attacked-directly

A genuinely distinct failure from every trap above, confirmed on the real Jan 2023 Q5 question this lesson's marked-solution is built on, and found only by re-reading the full examiner report during this pass rather than the excerpt an earlier pass had already pulled: "In part (b), some candidates ignored the “Hence” and attempted to solve the original equation in terms of x and could score no marks in this part" (WMA11_01_ER_2301). Part (b) says "Hence solve" specifically because it is built to be answered by using part (a)'s own quadratic, 9p²+26p−3=0 — not by any other route that happens to reach the right x. A response that abandons the substitution and attacks $3\times9^x+3^{x+2}=1+3^{x-1}$ by some other method, even one that is mathematically valid in principle, scores zero on part (b): the mark scheme credits the substitution-and-solve method specifically, not "any way of getting the right x." "Hence" is an instruction about which method to use, not a stylistic connective — treat it as one every time it appears attached to a part-mark question.

The card

s = rθ, A = ½r²θ — θ MUST be in radians. Neither formula is in the booklet — memorise both.
Major angle = 2π − minor angle. Check which arc/sector the question actually wants before substituting.
Sector perimeter = arc + 2r. The arc alone is not the whole boundary — don't forget the two radii.
A chord is not a diameter unless it passes through O at angle π. Check before assuming.
Inexact answers: 3 s.f., unless an exact form (e.g. 'in terms of π') is requested.

Why it works — Why s = rθ — arc length is proportional to the angle, and the radius is the constant

Arc length and the angle it subtends are directly proportional: double the angle at the centre, and you sweep out exactly double the arc, because the curve being traced is the same circle throughout — turning through the angle twice runs over the same shape twice. So $s = k\theta$ for some constant $k$ that depends only on the circle, not on which particular angle is chosen. To find $k$, use the one case that is already known without any formula at all: a full turn, $\theta = 2\pi$, traces the entire circumference, $s = 2\pi r$. Substituting into $s = k\theta$: $2\pi r = k(2\pi)$, so $k = r$. That is the whole derivation — the radius is not an arbitrary ingredient bolted onto the angle, it is the specific number that makes the full-turn case come out right, and the formula $s = r\theta$ then works for every other angle for free, by the same proportionality. It is also exactly why $\theta$ has to be in radians: the constant $k=r$ was derived from the fact that a full turn is $2\pi$ of whatever unit $\theta$ is measured in. Using degrees would mean redoing the derivation with $360$ in place of $2\pi$, which changes the constant from $r$ to $\frac{\pi r}{180}$ — an extra factor the radian version simply doesn't carry.

Traps — 5

sector-area-half-omitted

Confirmed directly on a real sector-area question: "The most common error was to omit the ½ in the area formula." The mechanism block above shows why the ½ is not an arbitrary constant: it is what falls out of dividing a full circle's area, $\pi r^2$, by the full angle that produces it, $2\pi$. Dropping it always exactly doubles the answer — a useful check on any sector-area result that looks suspiciously like double another value in the same question.

minor-angle-used-for-major-arc-or-sector

Confirmed directly and specifically: "The most common mistake was to use 1.64 as the angle, finding the minor, rather than the major, arc length." The facts this lesson is built from also state, as a general summary across the richest series on this topic, that this same major/minor confusion "confirmed across two of" the three most fully-worked series checked — though only this one instance comes with a directly quotable figure attached, and that distinction is preserved here rather than inventing a second citation to match the summary's count. The defence is always the same: name which angle — the given one, or $2\pi$ minus it — belongs to the region actually asked for, in words, before any formula is touched.

composite-baseline-mistaken-for-diameter

Confirmed directly on a real composite-shape question: "a small number made the mistake of taking AOD as a diameter and so gave the answer 6.25 m" — instead of the correct radius, 5.77 m, obtainable only from the sector's own area formula. AOD was the straight base line of the whole composite design, passing through the centre O, but its own two ends sat on the two triangles attached to the sector, not on the sector's own arc — so it was never a length the circle itself actually reached, and not even a chord of the relevant circle, let alone its diameter. A length through the centre is a diameter only when the circle reaches both of its ends; a straight line elsewhere in a composite figure that merely happens to pass through O is a separate, unrelated quantity that has to be found from the given data (an angle and an area, or an angle and an arc length), never read off a diagram by assumption.

composite-perimeter-parts-missed

Confirmed directly on a genuine composite-perimeter question: "the most common errors were forgetting to double [a repeated] length, missing out [a base length], or [forgetting] to include the lengths of the radii." All three are the same underlying failure — going straight to a formula for one piece of the shape without first listing, in order, every edge the perimeter is actually made of. A sector alone already has three edges (arc, radius, radius); a composite shape built from a sector plus other pieces has more, and each has to be accounted for once, and only once, per its role in the boundary.

composite-shape-assumed-a-single-sector

Confirmed directly on a genuine composite-shape question: "a few candidates incorrectly assumed that the entire shape was a sector of a circle." A single sector has exactly two straight edges, both radii, both meeting at the centre, and one curved edge. A shape with a straight edge that does not pass through the centre, or with more than two straight edges, is not one sector — it has to be decomposed into a sector plus (or minus) whatever else is there before any formula is applied to any part of it.

The card

u = a^x: a^(x+n) = a^n · u, a^(x-n) = u ÷ a^n, a^(nx) = u^n. Every term becomes a power of u — that's spec 1.1, not new rules.
Two solves, not one: solve the quadratic for u, THEN solve a^x = u for x. The question asked for x.
a^x > 0 for every real x, any a > 0. A negative value of u is rejected, with a reason — never reported as an answer.
No logs needed: match indices directly (1/9 = 3^-2). WMA11 substitution questions are built this way, since P1 doesn't teach logs.
'Show all stages' questions penalise a calculator-only route even when the value of x is right.

Why it works — Why the substitution always works — and where it stops working

Take the general case algebraically rather than trusting that it works because it happened to work on one example. Suppose every term in an equation has the same base $a>0$, each raised to an exponent of the form $x+k$ for some rational constant $k$ (a plain $x$ is the case $k=0$; a coefficient like $2x$ is handled separately below). By the addition law, $a^{x+k}=a^k \cdot a^x$. Set $u=a^x$: every such term becomes $a^k u$, a constant multiple of $u$ — never a higher power, because the exponent's VARIABLE part is always exactly $x$, contributing exactly one factor of $u$. A term with exponent $2x$ is different in exactly one respect: by the power law, $a^{2x}=(a^x)^2=u^2$, contributing two factors of $u$ instead of one. Collect every term of the equation as a multiple of $u^2$, $u$, or a constant (a term with exponent $0\cdot x + k$, i.e. no $x$ at all, is already a plain number), and what remains is, by construction, a polynomial in $u$ of degree at most 2 — a quadratic, exactly because the highest power of $x$ appearing in any exponent was $2x$ and nothing higher. Change that highest exponent to $3x$ and the same argument produces a CUBIC in $u$, not a quadratic; WMA11's hidden-quadratic questions are hidden quadratics specifically because every exponent stays at or below $2x$, not because substitution itself is limited to quadratics.

Traps — 6

stops-at-the-substituted-variable

The named trap this lesson exists to fix, confirmed independently across two series. A real WMA11 examiner report states it directly: "a significant number of responses did not go any further than finding solutions to the quadratic in [the substituted variable]" (Jan 2025, Q4(b)) — the same failure the Jan 2023 Q5 equation this lesson's marked-solution is built on documents in its own report. The substituted letter is never the thing the question asked for; it is scaffolding on the way to it.

exponent-shift-mapped-as-a-power-not-a-multiple

Confirmed verbatim on the real question this lesson's marked-solution reproduces: "Common misconceptions included, $3^{x+2}=p^2$ and $3^{x-1}=p^{-1}$" (Jan 2023, Q5). Both errors treat a SHIFTED exponent as though it behaved like a raised power of the substituted letter — $3^{x+2}$ becomes $p^2$ (which is what $3^{2x}$ would give, not $3^{x+2}$), and $3^{x-1}$ becomes $p^{-1}$ (which is what $3^{-x}$ would give). The correct laws are additive on the base, not on the letter after substitution: $a^{x+k}=a^k\cdot u$, never $u^k$.

negative-value-of-u-not-rejected

Not itself a phrase the research bank's examiner-report extracts quote by name — this is mathematical necessity, not a documented misconception, and is named here because leaving it out would understate what a full solution requires. $u=a^x$ for $a>0$ can never be negative, so a value of $u$ that comes out negative when solving the quadratic is not a second solution waiting to be converted — it is not a solution at all, and the correct response states the rejection with a reason rather than omitting it silently. The real Jan 2023 Q5 equation has exactly this shape (p=-3 is one of its two roots).

answer-from-the-calculator-with-no-method-shown

The real Jan 2023 Q5 question this lesson's marked-solution is built on carries the instruction "In this question you must show all stages of your working. Solutions relying on calculator technology are not acceptable" (verbatim, from the question paper itself), and the official mark scheme for this exact question backs the instruction with a direct, unambiguous rule of its own in part (b)'s general guidance: "Answers just written down scores M0" (WMA11_01_MS_2301, Q5). Re-checking a related Jan 2025 quote against its own report during this pass found it does NOT transfer cleanly to this trap: that series' Q4(b) comment on calculator-only responses ends "...this approach could have been penalised on a different occasion" — the report's own general comments confirm the explicit show-your-working warning that series was printed only on Q2(a) and 5(b), not Q4, so a prior pass's use of that quote here overstated what actually happened on that specific question. The Jan 2023 Q5 mark scheme's own M0 rule is the correct, unhedged source for this trap. Treat the printed warning as a cue that the algebra itself, not just the final value, is what earns the marks.

quadratic-not-yet-formed-before-a-quadratic-technique-is-applied

A related but distinct discipline failure, confirmed on a different real question (not a substitution question, but the same underlying skill): "This question proved challenging for some candidates, who failed to realise that they needed to form a quadratic in x in order to use the discriminant" (Jan 2019, Q9). The substitution trap and this one are mirror images — one is about recognising you HAVE built a quadratic and still needing to finish the job; this one is about failing to recognise a quadratic needs building in the first place, before any quadratic technique (discriminant, factorising, the formula) has anything to act on.

hence-ignored-original-equation-attacked-directly

A genuinely distinct failure from every trap above, confirmed on the real Jan 2023 Q5 question this lesson's marked-solution is built on, and found only by re-reading the full examiner report during this pass rather than the excerpt an earlier pass had already pulled: "In part (b), some candidates ignored the “Hence” and attempted to solve the original equation in terms of x and could score no marks in this part" (WMA11_01_ER_2301). Part (b) says "Hence solve" specifically because it is built to be answered by using part (a)'s own quadratic, 9p²+26p−3=0 — not by any other route that happens to reach the right x. A response that abandons the substitution and attacks $3\times9^x+3^{x+2}=1+3^{x-1}$ by some other method, even one that is mathematically valid in principle, scores zero on part (b): the mark scheme credits the substitution-and-solve method specifically, not "any way of getting the right x." "Hence" is an instruction about which method to use, not a stylistic connective — treat it as one every time it appears attached to a part-mark question.

The card

xⁿ → nxⁿ⁻¹ works for every rational n — negative, fractional, or whole. Not in the formula booklet; memorise it.
Rewrite before you differentiate: √x is x to the power ½, 1/x is x⁻¹, 1/x² is x⁻². The power rule needs a power, not a root or a fraction.
A constant's derivative is 0, not +c. '+c' belongs to integration (the reverse process) — never to differentiation.
'Exact value' means a fraction or a surd, not a rounded decimal — even when the decimal looks tidy at a glance.
The derivative evaluated at a point IS the gradient of the tangent there (spec 4.1) — not the curve's height at that point.

Why it works — Why the power rule doesn't care whether n is negative, a fraction, or a whole number

The rule $x^n \to nx^{n-1}$ is usually proved first for positive whole numbers — expanding $(x+h)^n$ and taking the limit as $h \to 0$ leaves exactly $nx^{n-1}$, once every other term is carrying a leftover $h$ that vanishes. What lets it apply just as completely to $n = -2$ or $n = \frac{1}{2}$ is that the laws of indices (spec 1.1) already guarantee $x^n$ means the same kind of object whatever rational number $n$ is — $x^{-2}$ is $\frac{1}{x^2}$, $x^{1/2}$ is $\sqrt{x}$, and both still obey $x^a \times x^b = x^{a+b}$ exactly as $x^2 \times x^3 = x^5$ does. Differentiation only ever needs that one algebraic property to reach the power rule for every rational exponent — never that $n$ specifically be a positive integer. This is exactly why the exam formula list does not give three separate rules for 'positive powers', 'negative powers' and 'fractional powers' — there is only ever one rule, and the arithmetic that follows it (decreasing the power by 1, multiplying by the original power) is identical in every case. What genuinely differs case to case is bookkeeping, not method: a negative $n$ means the new power $n-1$ is even more negative, and $n = \frac{1}{2}$ means the new power is $-\frac{1}{2}$ — both ordinary outputs of 'decrease by 1', not signs that something unusual has happened.

Traps — 4

plus-c-added-after-correct-differentiation

Confirmed directly on a real differentiate-then-evaluate question: "a significant number of candidates who differentiated but then added + c to their expression" (Jan 2023, Q1) — and independently repeated on the OTHER real anchor this lesson is built from: "There were a few instances of inclusion of '+ c' or + 9 as a third term" (Jan 2025, Q5(a)). The differentiation itself was done correctly — this is not a calculus error, it is a habit carried over from integration, where a genuine "+c" is required every time (spec 5.1). Differentiating a constant gives 0, not an unknown: spec 4.1 defines the derivative as the gradient of the tangent, and a constant function is a horizontal line with zero gradient everywhere, not a line whose gradient still needs to be found. Real WMA11 mark schemes never give this step an independent mark of its own — on both anchors it is folded into the SAME accuracy mark that requires the finished derivative to have no extra terms, not a separate B-mark. There is a sharper, real second-order version of this trap too: a real Jan 2023 examiner report records "a significant minority of candidates who had introduced '+ c' in part (a) went on to try and establish its value by for example using x = 2 and y = 0" — and the real Jan 2023 Q1(b) mark scheme explicitly withholds the SUBSTITUTION method mark itself for exactly this, not just an accuracy mark: "Do not allow this mark if they have (dy/dx =) ... + c in part (a) AND subsequently go on to try and establish a value for 'c' using x = 2." Carrying an unresolved '+c' through a substitution without trying to solve for it (as in this lesson's Trap 1 above) still earns that method mark; actually trying to pin down a numeric value for it does not.

exact-value-given-as-decimal

Confirmed on the same question's second part: "A significant number of candidates ignored the request to give an exact value for the gradient, resorting immediately to a decimal answer" (Jan 2023, Q1(b)). This is an instruction-following trap, not an arithmetic one — the general principles for pure marking state it as a standing convention across the whole paper, not just this question: "where, for example, an exact answer is asked for... marks will normally be lost if the candidate resorts to using rounded decimals." A fraction whose denominator has a prime factor other than 2 or 5 (81 = 3⁴, in the marked solution above) is the clearest sign a decimal was never going to be exact in the first place. The same exact-value convention applies whether the exact answer is a clean fraction or a surd — the general principles quote above covers both explicitly, naming "an exact answer" and "working with surds" in the same breath, not as two different rules.

negative-power-term-harder-than-its-positive-power-sibling

A real Jan 2025 examiner report finds a striking split within a single question: "Almost all candidates obtained 12x², but it was more common to fail to obtain 2x⁻²" (Jan 2025, Q5(a)) — the exact same rule, xⁿ → nxⁿ⁻¹, applied to two different terms in the same derivative, with a much higher failure rate on the negative-index one. The rule does not get harder; tracking two negative signs at once (the one already on the term, and the one the negative power itself introduces) does.

power-rule-botched-on-a-negative-index-term

The same real report gives the concrete version of the error above: "there were a few responses where 2/x² was differentiated to 2x or just 2" (Jan 2025, Q5(a)). Neither wrong answer follows any version of the power rule — d/dx(2x⁻²) genuinely gives −4x⁻³; "2x" looks as though the power went up, and "2" looks as though the power vanished entirely. Both are the signature of applying the rule to the ORIGINAL fraction, 2/x², without first rewriting it as 2x⁻² — exactly the step this lesson's opening teach block insists on doing first.

The card

y=af(x): stretch, y-direction (from x-axis), scale factor a. a<0 also reflects in the x-axis.
y=f(x)+a: translation, vector (0,a) — a units in +y direction if a>0.
y=f(x+a): translation, vector (−a,0) — a units in −x direction (LEFT) if a>0. Sign flips.
y=f(ax): stretch, x-direction (from y-axis), scale factor 1/a. a<0 also reflects in the y-axis.
Required word: translation (never shift/move). Required word: stretch (never compression), whatever the scale factor.

Why it works — Why $y = f(x+a)$ moves the 'wrong' way — traced from a single point, not stated as a rule

Take a point on $y=f(x)$ where the curve reaches some height $q$ at input $p$: so $f(p) = q$. Ask where that SAME height $q$ appears on the new curve $y = f(x+a)$. It appears wherever $f(x+a) = q$ — and since $f(p) = q$ already, that happens exactly when $x + a = p$, i.e. $x = p - a$. So the point that was at $x = p$ is now found at $x = p - a$: every point has moved by $-a$ in the x-direction. If $a$ is positive, $-a$ is negative, so the whole curve moves in the NEGATIVE x-direction — left — even though $a$ was added, not subtracted. This is the entire mechanism behind the 'sign flips' rule, derived rather than memorised: the transformation is a translation by the vector $\binom{-a}{0}$. Run the same kind of argument on $y=af(x)$, but track the point forward instead of backward, because this transformation acts on the OUTPUT rather than the input: take the same point, height $q$ at input $p$, so $f(p)=q$. On the new curve, the value at that SAME input $p$ is $af(p) = aq$ — nothing about which input reaches $f$ has changed, since $a$ multiplies whatever $f$ already produced. So the point moves from $(p,q)$ to $(p,aq)$: the x-coordinate is untouched, and only the height scales, directly by $a$ (not by $1/a$), because the OUTPUT of $f$ is what got multiplied. And for $y=f(ax)$: the same height $q$ now appears where $f(ax) = q = f(p)$, i.e. $ax = p$, i.e. $x = p/a$ — every point moves to $\frac{1}{a}$ of its original distance from the y-axis, a stretch of scale factor $\frac{1}{a}$, the reciprocal of $a$, precisely because it is the INPUT that was multiplied by $a$ before $f$ ever saw it, and undoing that multiplication is what finds the new position. Two of the four transformations act directly with the number $a$ ($af(x)$: scale by $a$; $f(x)+a$: shift by $a$); the other two act with its negative or its reciprocal ($f(x+a)$: shift by $-a$; $f(ax)$: scale by $\frac{1}{a}$) — and the reason is always the same one-line piece of algebra: solve for $x$ (or for the output) in terms of the OLD variable, and whatever operation undoes what was done to $x$ (or to the output) is the transformation that actually happens.

Traps — 5

translation-called-a-shift-or-move

Confirmed directly on the real transformation from $y=6/x$ to $y=6/(x-2)$: "it was extremely rare to see a fully correct description of the transformation. Although most knew that the graph would move 2 units right, very few appeared to know that it was a translation, with most using the words 'shift' or 'move' instead" (Jan 2023, Q7(b)). The mathematics — direction, distance — was not the problem for most candidates. The mark scheme checks a specific noun, and "shift"/"move" are not it, however accurately they describe the picture.

af-and-fax-stretch-direction-confused

Confirmed on the real pair $y=\cos x \to y=\cos 2x$: "Candidates were split between the correct solution of (90, −1) and (180, −2) which presumably came from mistakenly thinking that the transformation mapping cos(x) to cos(2x) is a stretch in the y-axis direction" (Oct 2021, Q4(a)(i)) — the actual transformation is a stretch in the x-direction, scale factor 1/2. The two stretches change different visible properties of a curve (period vs amplitude, for a trig curve), which is the fastest way to tell them apart under time pressure without re-deriving the rule from scratch.

sign-error-on-a-transformed-key-value

Fully re-verified against the real question this pass, not just the examiner-report prose an earlier pass could only quote in isolation: Jan 2019 Q8 gives a curve $C$ with a single maximum at $(4,9)$, crossing the axes at $(-3,0)$ and $(0,6)$, and a single asymptote $y=4$. Part (a) asks for the asymptote of $y=f(-x)$ — a reflection in the y-axis, which changes no y-value on the curve, so the asymptote stays $y=4$ (real mark scheme, verbatim: "B1: States y = 4"). The examiner report confirms the trap directly: "The vast majority of candidates correctly stated y = 4... There was a significant number of candidates who stated y = −4. It is unclear whether this was a misunderstanding of which axis the graph was being reflected in" — i.e. treating $f(-x)$ (reflection in the y-axis, which this lesson's own MCQs test) as though it were $-f(x)$ (reflection in the x-axis, which WOULD flip the asymptote to $y=-4$). A second, independent real confirmation of the "reflected-in-wrong-axis" misconception this lesson already tests directly.

fax-scale-factor-not-inverted

The same Jan 2019 Q8, part (b): given the maximum of $C$ at $(4,9)$, state the turning point of $y=f\left(\frac{x}{4}\right)$ — the $f(ax)$ form with $a=\frac14$, so the real scale factor is $1/a=4$, giving $(16,9)$ (real mark scheme, verbatim: "B1: States (16, 9) only"). The examiner report names the exact reciprocal-confusion trap directly: "A small but significant number of candidates misunderstood the transformation and divided the x-coordinate of the turning point on y = f(x), giving (1,9) as their answer" — i.e. using $a=\frac14$ directly (dividing by 4) instead of its reciprocal $1/a=4$ (multiplying by 4). This is the same underlying slip as the cos(2x) case above, confirmed in the opposite direction: there, a candidate used $a=2$ instead of $1/a=\frac12$ (should have divided, multiplied instead); here, a candidate used $a=\frac14$ instead of $1/a=4$ (should have multiplied, divided instead). Between the two real questions, both directions of "forgetting to invert the scale factor" are independently confirmed on real scripts.

transformed-graph-not-used-to-count-intersections

The same Jan 2019 question's part (c) — using the transformed picture to state, for the ORIGINAL curve $C$, the values of $k$ for which the line $y=k$ meets $C$ at exactly one point — is described by the examiner report as "the least successful part of this question which was often omitted" (Jan 2019, Q8(c)), on a question the report elsewhere calls "a good discriminating question". The real answer needs both the touching case at the maximum ($k=9$) AND the whole ray at or below the asymptote ($k \le 4$, since the curve approaches but never reaches $y=4$, so every horizontal line at or below it still meets the curve exactly once) — verbatim from the real mark scheme: "B1: Sight of either one of k ≤ 4, k = 9 ... B1: Both of k ≤ 4, k = 9 ONLY" (with a documented special case: giving the answer in terms of $y$ instead of $k$ scores only the first of the two marks). Treating "read off one key feature" as the end of the question, rather than as information to be combined with a second, structurally different feature (a single point vs. an entire half-line), is exactly the gap this finding is describing.

The card

f″(x) → integrate → f′(x) + c₁. Fix c₁ NOW using a GRADIENT condition, f′(a) = m — before integrating again.
f′(x) → integrate → f(x) + c₂. Fix c₂ using a POINT on the curve, f(p) = q — never the gradient condition.
Two integrations, two constants, two different functions. A point belongs to f(x); a gradient value belongs to f′(x).
Method mark for integration: power of at least one term increased by 1 (xⁿ → xⁿ⁺¹). A constant is required every time — spec 5.1.
∫xⁿ dx = xⁿ⁺¹/(n+1) + c, n ≠ −1 — not in the formula booklet. Memorise it.
Finding each constant is really TWO marks, not one: a dependent method mark for substituting the condition and solving, then a separate accuracy mark for the value it produces. The second integration's own method mark survives even a wrong or missing first constant — but the SECOND constant cannot be found at all without the first substitution having been attempted.

Why it works — A gradient condition and a point are statements about two different functions

Everything in this lesson follows from one distinction, and the distinction is definitional, not a rule to memorise. Spec 4.1 defines $f'(x)$ AS the gradient of the tangent to $y = f(x)$ — so a condition written as "$f'(a) = m$" is a sentence about the STEEPNESS of the curve at $x = a$, and it can only be checked once $f'(x)$ itself exists as an expression. A condition written as "the curve $y = f(x)$ passes through $(p, q)$" is a sentence about the HEIGHT of the curve at $x = p$ — it says $f(p) = q$, and it can only be checked once $f(x)$ itself exists. These are not two versions of the same fact arriving in different notation. $f'(4) = 3$ and $f(4) = 3$ would be two completely different, unrelated statements about the same curve — one about its slope at $x=4$, one about its height there — and mixing them up is not a small error, because there is no algebra that recovers the right answer once the wrong function has been evaluated. This is exactly the mechanism behind the trap a real Jan 2023 examiner report names: candidates who had correctly integrated once, reaching $f'(x)$, then reached for the point $(p,q)$ — a fact about $f$, not about $f'$ — because it was the only other piece of given information left in the question, not because it belonged there.

Traps — 4

gradient-condition-applied-at-the-wrong-stage

Confirmed directly on a real two-stage integration question: candidates who had correctly reached f'(x) "were often unsuccessful because they substituted the coordinates for P... rather than realising f'(x) is the gradient function so that f'(4) = 3 was needed" (Jan 2023, Q11(b)). The point belongs to f, once f(x) exists — using it while only f'(x) exists attaches a fact about the curve's height to a function that is describing its slope.

both-constants-deferred-to-the-end

Confirmed on the same question: "It was very common to see candidates integrate twice and apply a constant only after their second integration, thus missing earlier marks" (Jan 2023, Q11(b)). Integrating f'(x) turns its constant into a genuine term of f(x) (c₁ multiplying x); a single constant tacked on only after both integrations has nowhere to put that term, and the marks for finding c₁ correctly, at its own line, are gone before the deferred constant is ever reached.

point-coordinates-transposed-or-sign-flipped

Confirmed on a separate, independent series: an examiner report on a two-stage integration question records "slips such as using the point (1, 8) or even (−8, 1)" in place of the correct point (Oct 2021, Q10). Both wrong versions are the same real point with a coordinate's sign or position moved — a copying error, not a conceptual one, but one the paper still records candidates making under time pressure. Write the point down exactly as given before substituting, not from memory a line later.

constant-of-integration-omitted-entirely

The single most common 'silly' final-mark loss on integration questions across the whole WMA11 corpus reviewed for this course, confirmed as a repeated finding across four separate series: Oct 2021 Q1, Oct 2020 Q9, Jan 2023 Q3, and Jan 2025 Q6(b). In a two-stage problem there are two separate places to make this same omission, not one — dropping either constant, not just both, loses the marks attached to that specific integration.

The card

Fraction under an integral sign? SPLIT first: (A+B)/D = A/D + B/D, then xᵐ/xⁿ = xᵐ⁻ⁿ on each piece. There is no rule for integrating a fraction as one object.
Never integrate numerator and denominator separately then divide — that computes a different quantity, not a slower route to the right one, and real mark schemes give it zero.
A wrong split can still earn the integration method mark, if what follows is integrated correctly and consistently — don't abandon a question over an uncertain split.
Every indefinite integration needs its OWN +c, written the moment you integrate — not deferred until a point is used. Confirmed the single most repeated silly mark loss across FOUR series (Jan 2023 Q3, Oct 2021 Q1, Oct 2020 Q9, Jan 2025 Q6(b)).
∫xⁿ dx = xⁿ⁺¹/(n+1) + c, n ≠ −1 — check no split term is x⁻¹; that case needs ln, outside P1, and real questions avoid it.

Why it works — "Integrate the top, integrate the bottom, then divide" is not risky — it is provably a different calculation

This is not a case of one method being safer than another; the two processes genuinely compute different things, and a small example proves it without needing any exam-specific evidence at all. Take the simplest possible case: $\int \frac{4x}{x}\,dx$. Split first, the correct way: $\frac{4x}{x} = 4$ (a constant, since x cancels), so $\int 4\,dx = 4x + c$ — a function that genuinely depends on x, which it must, since the thing being integrated does not vanish. Now the wrong way: integrate the numerator and denominator as if they were separate, unrelated expressions, then divide the two results. $\int 4x\,dx = 2x^2$; $\int x\,dx = \frac{x^2}{2}$; dividing gives $\frac{2x^2}{x^2/2} = 4$ — a bare constant, with the $x^2$ cancelling out of the ratio entirely. That result cannot be right on its face: it claims $\frac{dy}{dx}$ was a flat, unchanging number when the original expression, $\frac{4x}{x}$, so obviously simplifies to the constant 4 that even a correct calculation should have produced exactly that — and by coincidence of this simple example it does, which is exactly what makes it a fair test: even when the two methods happen to agree on a trivial case, the wrong method reaches its answer by an argument that would give a nonsense, x-dependent-looking ratio on almost any other fraction. Try it on $\int \frac{x^2}{x}\,dx$ instead — correctly, $\frac{x^2}{x}=x$, so $\int x\,dx = \frac{x^2}{2}+c$; the wrong way, $\int x^2\,dx \div \int x\,dx = \frac{x^3}{3} \div \frac{x^2}{2} = \frac{2x}{3}$ — a completely different function of x, not even close to $\frac{x^2}{2}$. This is the real content behind a real Jan 2023 examiner report's finding that this approach "resulted in no marks": it is not a partial-credit method that happens to be marked harshly, it is a calculation that does not compute the integral of the original expression at all.

Traps — 5

no-idea-how-to-split-the-fraction

Confirmed directly on the cleanest fully-worked example this research pass found: "some candidates had no idea how to write the expression as the sum of two terms" (Jan 2023, Q3). The fraction is left as one object because nothing in spec 5.2's rule looks like it applies to a fraction — the fix is mechanical, not conceptual: $\frac{A+B}{D} = \frac{A}{D} + \frac{B}{D}$ is fraction addition run backwards, and each resulting piece reduces to a pure power of x via $x^m/x^n = x^{m-n}$ (spec 1.1). Practise the split as its own separate step, before integration ever enters the working.

numerator-and-denominator-integrated-before-dividing

Confirmed on the same question: "A small minority attempted to integrate all 3 terms separately before division which resulted in no marks" (Jan 2023, Q3). This is not a wrong split that still earns partial credit — it is not a split at all, and it computes something genuinely different from the fraction's actual integral (the mechanism block above proves this on a simple case, not just describes it). Unlike an arithmetic slip inside a correct method, there is no method mark here to fall back on.

constant-of-integration-omitted

The single most common "silly" final-mark loss on integration questions across the whole WMA11 corpus reviewed for this course — independently confirmed as a repeated finding across FOUR separate series: Jan 2023 Q3, Oct 2021 Q1, Oct 2020 Q9, and Jan 2025 Q6(b). Spec 5.1 requires a constant of integration every time indefinite integration is performed, whether or not the question ever gives enough information to find its numeric value. Write $+c$ the instant the integration is finished, not after the point has been substituted — an expression with no free constant left in it has nowhere for a given point to attach.

wrong-split-can-still-earn-the-integration-method-mark

A real, quoted fragment from the Jan 2023 Q3 examiner report records three different wrong ways students split the actual question's fraction: "8x³+6x⁻² or 4x⁵+6x⁻² or 2x⁵+(3/2)x⁻². These candidates could still gain the second method mark for appropriate integration." The real mark scheme explains exactly why: the integration mark is coded dM1 — "Depends on the first M mark" (the split's own M1, which only needs "one processed index correct" to be earned), not on the split's separate A1 (full accuracy). A wrong-but-genuine split attempt still earns the split's M1, and that alone is enough to unlock the dependent integration mark — a wrong division, honestly committed to and then integrated correctly, is not the same failure as trap two above. The practical lesson: if you are unsure your split is right, do not panic, guess randomly, or leave the question blank — commit to a split and integrate it correctly, because the dependent mark only checks that a genuine method attempt came first, not that it was error-free.

correct-integration-left-unsimplified-or-missing-the-x

A second, separate way to lose the final accuracy mark even after splitting AND integrating correctly — newly confirmed by this lesson's own primary-source re-verification, not previously recorded in this course's facts bank: the real Jan 2023 Q3 examiner report continues past the +c trap to name it directly: "marks were lost unnecessarily for example by... not simplifying fully... [or by] other careless mistakes such as missing out 'x' but writing the index... which also cost the final accuracy mark." Concretely, writing the correct power as a bare exponent with no base — e.g. "⁻¹" floating on its own instead of $x^{-1}$ attached to its coefficient — states a number, not the term the mark scheme actually wants. The fix is procedural, not conceptual: once a term is integrated, check that the base variable and its exponent are both actually written down together, and that the whole expression is collected into its simplest single form, before moving on to substitute a point.

The card

f″(x) → integrate → f′(x) + c₁. Fix c₁ NOW using a GRADIENT condition, f′(a) = m — before integrating again.
f′(x) → integrate → f(x) + c₂. Fix c₂ using a POINT on the curve, f(p) = q — never the gradient condition.
Two integrations, two constants, two different functions. A point belongs to f(x); a gradient value belongs to f′(x).
Method mark for integration: power of at least one term increased by 1 (xⁿ → xⁿ⁺¹). A constant is required every time — spec 5.1.
∫xⁿ dx = xⁿ⁺¹/(n+1) + c, n ≠ −1 — not in the formula booklet. Memorise it.
Finding each constant is really TWO marks, not one: a dependent method mark for substituting the condition and solving, then a separate accuracy mark for the value it produces. The second integration's own method mark survives even a wrong or missing first constant — but the SECOND constant cannot be found at all without the first substitution having been attempted.

Why it works — A gradient condition and a point are statements about two different functions

Everything in this lesson follows from one distinction, and the distinction is definitional, not a rule to memorise. Spec 4.1 defines $f'(x)$ AS the gradient of the tangent to $y = f(x)$ — so a condition written as "$f'(a) = m$" is a sentence about the STEEPNESS of the curve at $x = a$, and it can only be checked once $f'(x)$ itself exists as an expression. A condition written as "the curve $y = f(x)$ passes through $(p, q)$" is a sentence about the HEIGHT of the curve at $x = p$ — it says $f(p) = q$, and it can only be checked once $f(x)$ itself exists. These are not two versions of the same fact arriving in different notation. $f'(4) = 3$ and $f(4) = 3$ would be two completely different, unrelated statements about the same curve — one about its slope at $x=4$, one about its height there — and mixing them up is not a small error, because there is no algebra that recovers the right answer once the wrong function has been evaluated. This is exactly the mechanism behind the trap a real Jan 2023 examiner report names: candidates who had correctly integrated once, reaching $f'(x)$, then reached for the point $(p,q)$ — a fact about $f$, not about $f'$ — because it was the only other piece of given information left in the question, not because it belonged there.

Traps — 4

gradient-condition-applied-at-the-wrong-stage

Confirmed directly on a real two-stage integration question: candidates who had correctly reached f'(x) "were often unsuccessful because they substituted the coordinates for P... rather than realising f'(x) is the gradient function so that f'(4) = 3 was needed" (Jan 2023, Q11(b)). The point belongs to f, once f(x) exists — using it while only f'(x) exists attaches a fact about the curve's height to a function that is describing its slope.

both-constants-deferred-to-the-end

Confirmed on the same question: "It was very common to see candidates integrate twice and apply a constant only after their second integration, thus missing earlier marks" (Jan 2023, Q11(b)). Integrating f'(x) turns its constant into a genuine term of f(x) (c₁ multiplying x); a single constant tacked on only after both integrations has nowhere to put that term, and the marks for finding c₁ correctly, at its own line, are gone before the deferred constant is ever reached.

point-coordinates-transposed-or-sign-flipped

Confirmed on a separate, independent series: an examiner report on a two-stage integration question records "slips such as using the point (1, 8) or even (−8, 1)" in place of the correct point (Oct 2021, Q10). Both wrong versions are the same real point with a coordinate's sign or position moved — a copying error, not a conceptual one, but one the paper still records candidates making under time pressure. Write the point down exactly as given before substituting, not from memory a line later.

constant-of-integration-omitted-entirely

The single most common 'silly' final-mark loss on integration questions across the whole WMA11 corpus reviewed for this course, confirmed as a repeated finding across four separate series: Oct 2021 Q1, Oct 2020 Q9, Jan 2023 Q3, and Jan 2025 Q6(b). In a two-stage problem there are two separate places to make this same omission, not one — dropping either constant, not just both, loses the marks attached to that specific integration.

The card

Fraction under an integral sign? SPLIT first: (A+B)/D = A/D + B/D, then xᵐ/xⁿ = xᵐ⁻ⁿ on each piece. There is no rule for integrating a fraction as one object.
Never integrate numerator and denominator separately then divide — that computes a different quantity, not a slower route to the right one, and real mark schemes give it zero.
A wrong split can still earn the integration method mark, if what follows is integrated correctly and consistently — don't abandon a question over an uncertain split.
Every indefinite integration needs its OWN +c, written the moment you integrate — not deferred until a point is used. Confirmed the single most repeated silly mark loss across FOUR series (Jan 2023 Q3, Oct 2021 Q1, Oct 2020 Q9, Jan 2025 Q6(b)).
∫xⁿ dx = xⁿ⁺¹/(n+1) + c, n ≠ −1 — check no split term is x⁻¹; that case needs ln, outside P1, and real questions avoid it.

Why it works — "Integrate the top, integrate the bottom, then divide" is not risky — it is provably a different calculation

This is not a case of one method being safer than another; the two processes genuinely compute different things, and a small example proves it without needing any exam-specific evidence at all. Take the simplest possible case: $\int \frac{4x}{x}\,dx$. Split first, the correct way: $\frac{4x}{x} = 4$ (a constant, since x cancels), so $\int 4\,dx = 4x + c$ — a function that genuinely depends on x, which it must, since the thing being integrated does not vanish. Now the wrong way: integrate the numerator and denominator as if they were separate, unrelated expressions, then divide the two results. $\int 4x\,dx = 2x^2$; $\int x\,dx = \frac{x^2}{2}$; dividing gives $\frac{2x^2}{x^2/2} = 4$ — a bare constant, with the $x^2$ cancelling out of the ratio entirely. That result cannot be right on its face: it claims $\frac{dy}{dx}$ was a flat, unchanging number when the original expression, $\frac{4x}{x}$, so obviously simplifies to the constant 4 that even a correct calculation should have produced exactly that — and by coincidence of this simple example it does, which is exactly what makes it a fair test: even when the two methods happen to agree on a trivial case, the wrong method reaches its answer by an argument that would give a nonsense, x-dependent-looking ratio on almost any other fraction. Try it on $\int \frac{x^2}{x}\,dx$ instead — correctly, $\frac{x^2}{x}=x$, so $\int x\,dx = \frac{x^2}{2}+c$; the wrong way, $\int x^2\,dx \div \int x\,dx = \frac{x^3}{3} \div \frac{x^2}{2} = \frac{2x}{3}$ — a completely different function of x, not even close to $\frac{x^2}{2}$. This is the real content behind a real Jan 2023 examiner report's finding that this approach "resulted in no marks": it is not a partial-credit method that happens to be marked harshly, it is a calculation that does not compute the integral of the original expression at all.

Traps — 5

no-idea-how-to-split-the-fraction

Confirmed directly on the cleanest fully-worked example this research pass found: "some candidates had no idea how to write the expression as the sum of two terms" (Jan 2023, Q3). The fraction is left as one object because nothing in spec 5.2's rule looks like it applies to a fraction — the fix is mechanical, not conceptual: $\frac{A+B}{D} = \frac{A}{D} + \frac{B}{D}$ is fraction addition run backwards, and each resulting piece reduces to a pure power of x via $x^m/x^n = x^{m-n}$ (spec 1.1). Practise the split as its own separate step, before integration ever enters the working.

numerator-and-denominator-integrated-before-dividing

Confirmed on the same question: "A small minority attempted to integrate all 3 terms separately before division which resulted in no marks" (Jan 2023, Q3). This is not a wrong split that still earns partial credit — it is not a split at all, and it computes something genuinely different from the fraction's actual integral (the mechanism block above proves this on a simple case, not just describes it). Unlike an arithmetic slip inside a correct method, there is no method mark here to fall back on.

constant-of-integration-omitted

The single most common "silly" final-mark loss on integration questions across the whole WMA11 corpus reviewed for this course — independently confirmed as a repeated finding across FOUR separate series: Jan 2023 Q3, Oct 2021 Q1, Oct 2020 Q9, and Jan 2025 Q6(b). Spec 5.1 requires a constant of integration every time indefinite integration is performed, whether or not the question ever gives enough information to find its numeric value. Write $+c$ the instant the integration is finished, not after the point has been substituted — an expression with no free constant left in it has nowhere for a given point to attach.

wrong-split-can-still-earn-the-integration-method-mark

A real, quoted fragment from the Jan 2023 Q3 examiner report records three different wrong ways students split the actual question's fraction: "8x³+6x⁻² or 4x⁵+6x⁻² or 2x⁵+(3/2)x⁻². These candidates could still gain the second method mark for appropriate integration." The real mark scheme explains exactly why: the integration mark is coded dM1 — "Depends on the first M mark" (the split's own M1, which only needs "one processed index correct" to be earned), not on the split's separate A1 (full accuracy). A wrong-but-genuine split attempt still earns the split's M1, and that alone is enough to unlock the dependent integration mark — a wrong division, honestly committed to and then integrated correctly, is not the same failure as trap two above. The practical lesson: if you are unsure your split is right, do not panic, guess randomly, or leave the question blank — commit to a split and integrate it correctly, because the dependent mark only checks that a genuine method attempt came first, not that it was error-free.

correct-integration-left-unsimplified-or-missing-the-x

A second, separate way to lose the final accuracy mark even after splitting AND integrating correctly — newly confirmed by this lesson's own primary-source re-verification, not previously recorded in this course's facts bank: the real Jan 2023 Q3 examiner report continues past the +c trap to name it directly: "marks were lost unnecessarily for example by... not simplifying fully... [or by] other careless mistakes such as missing out 'x' but writing the index... which also cost the final accuracy mark." Concretely, writing the correct power as a bare exponent with no base — e.g. "⁻¹" floating on its own instead of $x^{-1}$ attached to its coefficient — states a number, not the term the mark scheme actually wants. The fix is procedural, not conceptual: once a term is integrated, check that the base variable and its exponent are both actually written down together, and that the whole expression is collected into its simplest single form, before moving on to substitute a point.

The card

Period: smallest T with f(x+T)=f(x), a NUMBER. Domain: the set of x where f is defined. Never answer one with the other.
sin x, cos x: period 2π, amplitude 1, range [−1,1]. tan x: period π (not 2π), undefined where cos x = 0, unbounded.
sin x, tan x odd (point symmetry, origin). cos x even (line symmetry, y-axis).
af(x): vertical stretch ×a. f(x+a): translate LEFT by a. f(ax): horizontal stretch ×1/a — period ÷ a.
Always mark a scale on both axes before sketching — an unscaled sketch is a documented, confirmed way to lose marks.

Why it works — Why the period of tan x is π, not 2π — and why that isn't a coincidence to memorise

Tangent is defined as a ratio: for an angle $x$, $\tan x$ is $\sin x$ divided by $\cos x$ (the same ratio a right-angled triangle gives as opposite over adjacent, extended to every real angle). Two facts about half-turns of $\sin$ and $\cos$ are enough to derive the whole period from scratch, without memorising it as a separate rule: rotating a point on the unit circle by $180°$ sends $(\cos x, \sin x)$ to $(-\cos x, -\sin x)$ — which is exactly the point at angle $x + \pi$, so $\sin(x + \pi) = -\sin x$ and $\cos(x + \pi) = -\cos x$. Both individual functions flip sign under a shift of $\pi$ — neither one, on its own, has returned to its original value; a shift of $\pi$ is NOT a period for $\sin x$ or $\cos x$ alone. But form the ratio: $\tan(x + \pi) = \dfrac{\sin(x+\pi)}{\cos(x+\pi)} = \dfrac{-\sin x}{-\cos x} = \dfrac{\sin x}{\cos x} = \tan x$. The two minus signs cancel. A shift that fails for the numerator and fails for the denominator, individually, succeeds for their ratio — because both flip together. That is the entire reason $\tan x$'s period is shorter than $\sin x$'s and $\cos x$'s: not an arbitrary extra fact about tangent, but a direct consequence of it being built from two functions that both change sign at exactly the same rate. And $\pi$ is not just A working shift, it is the SMALLEST one: any shift smaller than $\pi$ would have to be tested against $\sin$ and $\cos$ individually failing to repeat that early (their own period is a full $2\pi$), and no smaller shift produces the same sign-cancellation the half-turn does, so nothing shorter closes the ratio back to itself. One more thing worth noticing, because it is exactly why the documented wrong answer is so tempting: $\tan x$'s domain has a gap between consecutive asymptotes — one continuous branch runs from $-\frac{\pi}{2}$ to $\frac{\pi}{2}$, and that interval happens to be $\pi$ WIDE. The width of one branch's domain and the period are, for tangent specifically, the same number — which is precisely the coincidence that makes writing the interval, instead of its length, feel like a reasonable answer to "state the period".

Traps — 3

period-confused-with-domain-or-interval

Confirmed directly on the real question this lesson is built around: "less than half of all candidates gave a correct answer for the period of $\tan x$... Quite frequently the answer was given as an interval, including $-\frac{\pi}{2} < x < \frac{\pi}{2}$... It is possible that this type of answer arose through a confusion between period and domain/range" (Jan 2023, Q9(a)) — on a question worth exactly one mark (B1), with no method needed, making a majority-wrong result on a one-line "state" question one of the most lopsided findings anywhere in this research pass. The period is always a single number; a set of x-values, however correctly described, cannot be it.

sketch-lacks-a-marked-scale

Confirmed independently on a different question — sketching y = sin x on axes already given for y = cos 2x — where the examiner report records that "the absence of a scale on the y = cos 2x graph lead to some candidates not sketching y = sin x carefully enough" and, separately, that "it was noticeable that many candidates did not mark a scale on the x-axis, despite part (a) encouraging them to do this" (Jan 2019, Q5). A curve with the right shape but no marked scale cannot show that a period is 2π rather than, say, 4 — the examiner has no way to check the claim the sketch is supposed to be making.

stretch-direction-inverted-y-for-x

Confirmed on a real transformation question — mapping cos(x) to cos(2x) — where the report records that the incorrect answer "presumably came from mistakenly thinking that the transformation mapping cos(x) to cos(2x) is a stretch in the y-axis direction" (Oct 2021, Q4(a)(i)); the correct transformation is a stretch in the x-axis direction, scale factor ½. That report is about cos(2x) specifically, not sin(2x) — but the mechanism is identical for any y = sin(kx)/cos(kx): a coefficient multiplying x itself, before the trig function is applied, is always a horizontal effect, never a vertical one, whichever of the two trig functions it is attached to.

The card

f(x) > 0: curve above the x-axis. f(x) < 0: below. The roots are the boundary — find them the usual three ways.
Upward parabola (a>0): positive OUTSIDE the roots, negative BETWEEN them. Downward (a<0): reversed.
Two inequalities combined: always sketch a number line — a whole branch of one can vanish against the other, invisibly in pure algebra.
State the answer in the variable the question named (usually x) — never y, and never a region's label (e.g. R).
Strict (< / >): boundary excluded. Non-strict (≤ / ≥): boundary included. Carry the original's strictness into the final line.

Why it works — Why 'outside' and 'between' aren't two rules to memorise — it's one sign pattern

Write the quadratic in factorised form, $f(x) = a(x-p)(x-q)$, with $p < q$ the two DISTINCT real roots (which exist only when $b^2-4ac > 0$, strictly — the same discriminant condition from the previous lesson, tightened here to its strict form because $p<q$ itself already rules out the repeated-root case $b^2-4ac=0$, where the curve only touches the axis and there is no 'between the roots' region to speak of). Consider the product $(x-p)(x-q)$ on its own, ignoring $a$ for a moment. For $x < p$: both $(x-p)$ and $(x-q)$ are negative, so their product is positive. For $p < x < q$: $(x-p)$ is positive but $(x-q)$ is still negative, so the product is negative. For $x > q$: both factors are positive, so the product is positive again. That sign pattern — positive, negative, positive — is a plain consequence of two linear expressions each changing sign only at their own zero and nowhere else; it has nothing specifically to do with quadratics, and it holds for ANY $p < q$, not particular ones. Now bring $a$ back in: multiplying by a positive $a$ leaves every sign exactly as it was, giving the 'upward: outside positive, between negative' rule stated above. Multiplying by a negative $a$ flips every sign, giving 'downward: outside negative, between positive' — exactly its reverse. A single test point substituted into ANY interval settles that interval's sign directly, and is often faster in the exam than trying to recall which shape goes with which case.

Traps — 5

final-answer-stated-in-the-wrong-variable

Confirmed directly, and by Pearson's own account a genuinely widespread error on the first sitting of this content: "stating y < 16 was a frequent error" (Jan 2019, Q4) — the number 16 there is the real, verified boundary value from that question; candidates had correctly found it and then written the final line of the answer using y, the OUTPUT variable, instead of x, the one actually being solved for. The two letters sit on the same picture, which is exactly why the swap is easy to make and easy to miss when checking your own work.

region-label-used-as-the-answer-variable

Confirmed on the same question: "others stated inequalities involving R, which gained no credit" (Jan 2019, Q4) — a diagram will sometimes name a shaded region with a letter, most often R, purely as a caption for the picture. R is not a coordinate and does not belong in an inequality about x; treating a label as if it were algebra scores nothing, however correct the reasoning behind it was.

sketch-skipped-a-sign-case-goes-unnoticed

Confirmed, though on a different underlying question type from this lesson's own worked examples: "it was rare to see a sketch graph, which would have helped students to see that x > 0 was required" (Oct 2021, Q3(i) — solving 3/x > 4 alone, a single reciprocal inequality, not a quadratic one and not itself two inequalities combined). The specific equation is not this topic's, and the real trap there was a dropped CASE (the necessary x > 0, missed by candidates who rearranged straight to x < 3/4), not a dropped branch — but the report's own explanation generalises: skipping the picture is what let a sign condition slip past a purely algebraic approach. This lesson's worked-chain and chain-drill blocks rehearse the same discipline — checking a combined region on a number line rather than symbolically — on genuinely quadratic material, adapted from that diagnosis rather than reproducing its structure.

between-and-outside-the-roots-reversed

Not a specific quoted WMA11 finding for this topic in the research pass this lesson draws on — flagged honestly as a natural, mathematically-grounded slip rather than a documented script error. Swapping which region belongs to an upward parabola and which to a downward one is exactly the failure the factorised-form derivation above exists to make unnecessary: $(x-p)(x-q)$ is negative between its roots and positive outside them regardless of any picture, and multiplying by $a$ only flips which of those counts as 'positive' — a single test point rebuilds the correct region in one line, faster than trying to recall which shape goes with which case.

boundary-strict-vs-non-strict-mismatched

Not a specific quoted WMA11 finding for this exact topic either — the same discipline IS documented on a closely related question type in this course's discriminant lesson ('some gave their final answer as 0 ≤ k ≤ 5/9, losing the final mark', Jan 2023 Q4), and it transplants directly here: a strict original inequality ($<$ or $>$) demands a strict final answer, and a non-strict one ($\leq$ or $\geq$) demands the boundary be included. This lesson's own marked-solution and worked-chain examples are deliberately built with different strictness at each end for exactly this reason — copying one convention onto both ends of an answer is a fast way to lose a mark that has nothing to do with the algebra.

The card

f(x) > 0: curve above the x-axis. f(x) < 0: below. The roots are the boundary — find them the usual three ways.
Upward parabola (a>0): positive OUTSIDE the roots, negative BETWEEN them. Downward (a<0): reversed.
Two inequalities combined: always sketch a number line — a whole branch of one can vanish against the other, invisibly in pure algebra.
State the answer in the variable the question named (usually x) — never y, and never a region's label (e.g. R).
Strict (< / >): boundary excluded. Non-strict (≤ / ≥): boundary included. Carry the original's strictness into the final line.

Why it works — Why 'outside' and 'between' aren't two rules to memorise — it's one sign pattern

Write the quadratic in factorised form, $f(x) = a(x-p)(x-q)$, with $p < q$ the two DISTINCT real roots (which exist only when $b^2-4ac > 0$, strictly — the same discriminant condition from the previous lesson, tightened here to its strict form because $p<q$ itself already rules out the repeated-root case $b^2-4ac=0$, where the curve only touches the axis and there is no 'between the roots' region to speak of). Consider the product $(x-p)(x-q)$ on its own, ignoring $a$ for a moment. For $x < p$: both $(x-p)$ and $(x-q)$ are negative, so their product is positive. For $p < x < q$: $(x-p)$ is positive but $(x-q)$ is still negative, so the product is negative. For $x > q$: both factors are positive, so the product is positive again. That sign pattern — positive, negative, positive — is a plain consequence of two linear expressions each changing sign only at their own zero and nowhere else; it has nothing specifically to do with quadratics, and it holds for ANY $p < q$, not particular ones. Now bring $a$ back in: multiplying by a positive $a$ leaves every sign exactly as it was, giving the 'upward: outside positive, between negative' rule stated above. Multiplying by a negative $a$ flips every sign, giving 'downward: outside negative, between positive' — exactly its reverse. A single test point substituted into ANY interval settles that interval's sign directly, and is often faster in the exam than trying to recall which shape goes with which case.

Traps — 5

final-answer-stated-in-the-wrong-variable

Confirmed directly, and by Pearson's own account a genuinely widespread error on the first sitting of this content: "stating y < 16 was a frequent error" (Jan 2019, Q4) — the number 16 there is the real, verified boundary value from that question; candidates had correctly found it and then written the final line of the answer using y, the OUTPUT variable, instead of x, the one actually being solved for. The two letters sit on the same picture, which is exactly why the swap is easy to make and easy to miss when checking your own work.

region-label-used-as-the-answer-variable

Confirmed on the same question: "others stated inequalities involving R, which gained no credit" (Jan 2019, Q4) — a diagram will sometimes name a shaded region with a letter, most often R, purely as a caption for the picture. R is not a coordinate and does not belong in an inequality about x; treating a label as if it were algebra scores nothing, however correct the reasoning behind it was.

sketch-skipped-a-sign-case-goes-unnoticed

Confirmed, though on a different underlying question type from this lesson's own worked examples: "it was rare to see a sketch graph, which would have helped students to see that x > 0 was required" (Oct 2021, Q3(i) — solving 3/x > 4 alone, a single reciprocal inequality, not a quadratic one and not itself two inequalities combined). The specific equation is not this topic's, and the real trap there was a dropped CASE (the necessary x > 0, missed by candidates who rearranged straight to x < 3/4), not a dropped branch — but the report's own explanation generalises: skipping the picture is what let a sign condition slip past a purely algebraic approach. This lesson's worked-chain and chain-drill blocks rehearse the same discipline — checking a combined region on a number line rather than symbolically — on genuinely quadratic material, adapted from that diagnosis rather than reproducing its structure.

between-and-outside-the-roots-reversed

Not a specific quoted WMA11 finding for this topic in the research pass this lesson draws on — flagged honestly as a natural, mathematically-grounded slip rather than a documented script error. Swapping which region belongs to an upward parabola and which to a downward one is exactly the failure the factorised-form derivation above exists to make unnecessary: $(x-p)(x-q)$ is negative between its roots and positive outside them regardless of any picture, and multiplying by $a$ only flips which of those counts as 'positive' — a single test point rebuilds the correct region in one line, faster than trying to recall which shape goes with which case.

boundary-strict-vs-non-strict-mismatched

Not a specific quoted WMA11 finding for this exact topic either — the same discipline IS documented on a closely related question type in this course's discriminant lesson ('some gave their final answer as 0 ≤ k ≤ 5/9, losing the final mark', Jan 2023 Q4), and it transplants directly here: a strict original inequality ($<$ or $>$) demands a strict final answer, and a non-strict one ($\leq$ or $\geq$) demands the boundary be included. This lesson's own marked-solution and worked-chain examples are deliberately built with different strictness at each end for exactly this reason — copying one convention onto both ends of an answer is a fast way to lose a mark that has nothing to do with the algebra.