Why Wrong Answers Feel Right
Most students know the feeling: the working looks tidy, the answer looks plausible, and the page comes back marked wrong. There is usually no answer key at home, no tutor nearby, and no time in an exam to redo everything. What readers want is a fast way to know, not a promise to “be more careful”.
The slip itself is rarely a gap in knowledge. It is a dropped minus sign, a misread exponent or a carried digit that went missing. Re-reading the same working tends to repeat the same slip, because the eye sees what it expects to see. A useful check therefore has to approach the answer from a different direction.
George Pólya built this into his four-step method in How to Solve It. The last phase, “looking back”, opens with two plain questions: “Can you check the result? Can you check the argument?” The seven methods below are that phase made concrete, ordered from quickest to most specialised.
Method 1: Substitute the Answer Back
Definition. Substitution means putting the answer into the original statement and checking that both sides agree. The word “original” matters. If the answer goes into a line already rewritten halfway through the solution, a mistake in that line will pass the check and survive.
Example. Solve 3x + 2 = x + 14. Subtract x and 2 from both sides to get 2x = 12, so x = 6. In the original, the left side is 3(6) + 2 = 20 and the right side is 6 + 14 = 20. They match.
The limit is that substitution proves one number works, not that every solution was found. A quadratic has two roots, and testing one says nothing about the other. The solve for x calculator is handy for comparing against your own result.
Method 2: Estimate Before You Calculate
Estimation catches large errors such as a misplaced decimal point, a doubled value or a forgotten zero. Round each number to something easy, compute roughly, then see whether the exact answer lands nearby.
Example. For 48 × 52, round to 50 × 50 = 2,500. The exact product is 2,496, so an answer such as 24,960 or 249.6 is rejected on sight. For 0.37 × 812, the estimate 0.4 × 800 = 320 sits near the exact 300.44. The estimate runs high because both roundings moved up, but the size is right.
The habit worth building is to estimate before the calculator comes out. An estimate made after seeing the answer tends to bend toward it.
Method 3: Run the Inverse Operation
Multiplication undoes division, a square undoes a square root, and differentiation undoes integration. Apply the inverse to the answer and see whether the starting value returns.
Example. 1,247 ÷ 29 = 43, and 29 × 43 = 1,247 exactly. For calculus, if ∫ 6x² dx = 2x³ + C, then differentiating 2x³ gives 6x², which matches the integrand. The integral calculator and the derivative calculator make a good pair for this loop.
Method 4: Hunt for Extraneous Roots
Squaring both sides, multiplying by an expression containing x, or taking logarithms can create solutions the original equation never had. These are called extraneous solutions.
Example. Solve √(x + 6) = x. Squaring gives x + 6 = x², so x² − x − 6 = 0 and x = 3 or x = −2. Test both in the original: √9 = 3 works, but √4 = 2 does not equal −2. Only x = 3 survives. Whenever a step is not reversible, substitution is the only dependable filter.
Method 5: Solve It a Second Way
A second method rarely repeats the first method’s slip. A quadratic solved by the formula can be rechecked by factoring, or by testing that the roots add to −b/a and multiply to c/a. A system of equations solved by substitution can be rechecked by elimination. A definite area can be rechecked by a rough sketch.
This is the most reliable check and also the slowest, so it earns its place on questions worth many marks. The quadratic equation calculator shows the discriminant and both roots for the comparison.
Method 6: Test Calculus Results Numerically
A derivative can be checked without any rules. The central difference, (f(x + h) − f(x − h)) ÷ 2h with a small h, approximates the slope at a point. This is a standard idea from numerical differentiation.
Example. A student claims the derivative of f(x) = x² sin 3x is 2x sin 3x + x² cos 3x, forgetting the factor 3 from the chain rule. At x = 1 with h = 0.0001, the central difference gives −2.68774. The correct formula, 2x sin 3x + 3x² cos 3x, gives −2.68774. The student’s version gives −0.70775, so the error is exposed at once.
Method 7: Apply Sanity Bounds
Some answers can be rejected by a property alone. A probability must lie between 0 and 1. A variance cannot be negative. A square root of a real number cannot be negative. A triangle’s side cannot exceed the sum of the other two. Units help as well: an area measured in metres cannot be right.
These checks cost a few seconds and catch answers that are impossible, even when the arithmetic behind them looks clean.
One Problem, Four Checks in Action
Take 2x² − 7x + 3 = 0. The quadratic formula gives x = (7 ± √(49 − 24)) ÷ 4 = (7 ± 5) ÷ 4, so x = 3 or x = ½.
Substitution. For x = 3: 2(9) − 21 + 3 = 0. For x = ½: 2(¼) − 3.5 + 3 = 0.5 − 3.5 + 3 = 0. Both roots pass in the original equation.
Sum and product. The roots should add to −b/a = 7/2 and multiply to c/a = 3/2. Indeed 3 + ½ = 3.5 and 3 × ½ = 1.5.
Second route. Factoring gives (2x − 1)(x − 3), which is zero at x = ½ and x = 3. The graph is an upward parabola crossing the x-axis twice, which agrees with a positive discriminant of 25.
Sanity bound. The discriminant is a perfect square, so rational roots were expected. A messy decimal here would have signalled an arithmetic slip. Four independent checks took under two minutes, and each one could have exposed a different error.
Choosing a Check by Subject
Algebra leans on substitution and second routes. Arithmetic leans on estimation and remainders. Geometry leans on units, sketches and the triangle inequality. Statistics leans on sanity bounds: a standard deviation cannot be negative, and a z-score beyond ±4 deserves a second look. Calculus leans on central differences and inverse operations. Readers who work in several subjects can keep the same three-step habit and swap only the middle step.
What 80,000 Simulated Mistakes Showed
Mathvelo tested how well the cheap checks catch realistic slips. The test used 20,000 random products of two three-digit numbers, then corrupted each answer in four common ways. The estimate rounded each factor to the nearest ten and flagged any answer more than 10% away. The other tests compared remainders after dividing by 10, 9 or 11. The table shows the share of mistakes each check caught.
| Mistake type | Estimate | Last digit | Mod 9 | Mod 11 |
|---|---|---|---|---|
| One digit changed | 24.4% | 17.6% | 98.1% | 100% |
| Two digits swapped | 26.6% | 21.5% | 0% | 100% |
| Decimal shifted one place | 100% | 80.8% | 35.0% | 86.2% |
| Off by 10, 100 or 1,000 | 0% | 0% | 100% | 100% |
Two lessons stand out. First, estimation is excellent against decimal slips and nearly useless against small digit errors, which is exactly why it should be paired with something else. Second, casting out nines (the mod 9 check) catches almost every changed digit yet never catches swapped digits, because swapping digits leaves their sum unchanged. Casting out elevens catches both.
Which Check Catches Which Mistake
| Typical mistake | Best check |
|---|---|
| Wrong decimal place or size | Estimate first |
| Sign error or dropped term | Substitute into the original |
| False root after squaring | Substitute every candidate |
| Missing the second root | Sum and product of roots |
| Forgotten chain-rule factor | Central difference |
| Impossible value | Sanity bounds |
A 60-Second Checking Routine
Estimate before calculating, then spend the minute afterwards in this order: compare the answer with the estimate (10 seconds), substitute into the original statement (20 seconds), run one sanity bound (10 seconds), and finish with a second method only on the questions worth the time. For practice, solve in the simplify calculator, then verify by hand. A related habit is described in the guide on how to study math, where self-testing replaces re-reading.
What Is Overrated, Underrated and Often Wrong
Overrated: asking “does it look right?” An answer that looks right is the one most likely to slip through, since the eye accepts it. Underrated: checking the units and the size. It is the cheapest test, and the simulation above shows it removing every decimal slip. Often wrong: the belief that casting out nines proves an answer correct. It can only prove an answer wrong, and the data shows it is blind to swapped digits. Treat every check as a filter, not a proof.
Quick Answers
How do I check if my math answer is correct without an answer key?
Estimate first, substitute the answer into the original problem, and test it against a sanity bound. If a question is worth many marks, solve it a second way as well.
How do I check a quadratic equation answer?
Substitute each root into the original equation, then test that the roots add to −b/a and multiply to c/a. Both facts hold for correct roots.
How do I check a derivative answer?
Compare it with a central difference at one or two points, using a small h such as 0.001 or 0.0001. A forgotten chain-rule factor shows up at once.
Does casting out nines prove an answer is right?
No. It only rules answers out. It also misses swapped digits completely, so pair it with an estimate or a mod 11 check.
Does estimation work for algebra?
Yes, for size and sign. Plug in a convenient value such as x = 1 or x = 10 and see whether the claimed formula behaves sensibly.
Related Mathvelo Tools
Further Reading and Sources
- Pólya, G. (1945). How to Solve It. Princeton University Press. — Overview of the four-step method
- How to Solve It: the “looking back” questions — Hosted by Harvard Mathematics
- Extraneous and missing solutions — Why squaring or clearing denominators can add false roots
- Casting out nines — What the mod 9 check can and cannot detect
A note on these guides. Worked examples were computed and checked step by step. The simulation used a fixed random seed and can be reproduced. Opinions are labelled as opinions. This page is educational and not professional advice.