Skipping the Condition Behind a Step Is What Breaks Your Solution

One question is left on a Sunday night: solve $\sqrt{x-3} = x-5$. Square both sides:

so $x = 4$ or $x = 7$. Every line is correct.

Substitute back. For $x = 7$: $\sqrt{7-3} = 2$ and $7 - 5 = 2$, true. For $x = 4$: $\sqrt{4-3} = 1$ but $4 - 5 = -1$, so $x = 4$ is not a solution.

Squaring runs one way only: $a = b$ gives $a^2 = b^2$, but $a^2 = b^2$ gives $a = b$ or $a = -b$. The squared equation carries both, and $x = 4$ belongs to the other branch, since $-(4-5) = 1$.

The condition is short: a square root never returns a negative number, so $x - 5 \geq 0$, that is $x \geq 5$. Write it beside the squaring step and $x = 4$ is ruled out the moment it appears.

You can square both sides perfectly. Knowing when it is allowed is a separate piece of knowledge, rarely taught as one. In a study comparing university students with doctoral students writing proofs in abstract algebra, the undergraduates often knew the facts the proof needed and could apply them, and still could not assemble the proof^[1]. What the doctoral students had was a sense of which fact to reach for^[1].

In one small study, thirteen university students were shown an argument whose last line was true but did not follow from the lines above it^[2]. Only six noticed it was not a valid proof, and only two for sound mathematical reasons^[2]. Asked directly whether the last line followed, ten rejected it^[2]. Thirteen students and one argument, in a study its authors called exploratory, measure nothing about how often anybody misses anything^[2]. Read it as a demonstration that what separated missing the broken step from catching it was being asked the question^[2].

Researchers tracked the eyes of beginning university students and of working mathematicians reading proofs^[3]. The beginners spent proportionally more time on how the mathematics looked on the page^[3]. The mathematicians shifted their eyes back and forth between consecutive lines far more often — a sign, the researchers said, of working out the unwritten reason for each step^[3].

That unwritten reason is what you are asked to write: the restriction $x \geq 5$ appears in neither line above, only in the arrow between them.

Solve $\frac{x^2-1}{x-1} = 0$. Multiplying both sides by $x-1$ gives $x^2 - 1 = 0$, so $x = 1$ or $x = -1$. Substituting back, $x = -1$ gives $\frac{0}{-2} = 0$, true, while $x = 1$ gives $\frac{0}{0}$, which is not a number. The multiplication was legal only for $x \neq 1$, and assumed it silently.

Cancelling instead gets the right answer here, but only because the forbidden value never came up. Since $x^2 - 1 = (x-1)(x+1)$, you cancel to $x + 1 = 0$ and get $x = -1$, which the original expression allows, so nothing is lost. Change the right-hand side and the same move bites: $\frac{x^2-1}{x-1} = 2$ cancels to $x + 1 = 2$, so $x = 1$ — the one value the original forbids, so that equation has no solution at all. Cancelling is legal only for $x \neq 1$, and you learn whether that mattered only by writing it down.

Division fails the other way: it deletes. Dividing $x^2 = 5x$ by $x$ gives $x = 5$ and throws away $x = 0$, which also satisfies it; factoring to $x(x-5) = 0$ keeps both.

• Dividing or multiplying both sides by an expression in $x$: can it be zero?
• Squaring, or any even power: does a sign-flipped answer sneak in?
• Cancelling a factor: what did the original expression exclude?
• A logarithm or an even root: is the input inside the allowed domain? Integrating $\frac{1}{x}$ is the standard case: $\ln x$ assumes $x > 0$, $\ln|x|$ covers every nonzero $x$, and the symbols never say which you meant.
• A diagram you drew yourself: is it to scale, or did you assume a right angle or an equal side?
• A named theorem: are its conditions actually met?

A review that pooled many studies of mathematics learning — university students again, not school-age ones — found that prompting them to explain their own reasoning gave a small to moderate boost, measured straight afterwards^[4]. The boost was larger when they were given a structure for what a good explanation contains, not just told to explain^[4]. The same review found much thinner evidence that the gain survives weeks, or holds up in an ordinary classroom^[4].

So use a frame. This step needs ___, because ___. For the radical equation: $x \geq 5$, because squaring lets $a = -b$ in alongside $a = b$. For the rational one: $x \neq 1$, because the expression has no value there. A question that never says what kind of number $x$ is needs a different note: assume all variables are real, written once at the top.

This can be trained. In three experiments, undergraduates given a short booklet on explaining how each line of a proof follows from the one before came away understanding the proof better^[5]. What was measured was their grasp of somebody else's proof, not their own solutions^[5].

You may already ask whether $x$ can be zero, and wonder why it must go on paper. A review that pooled 56 classroom experiments with students from the first year of school to the end of high school found that writing about what you are studying helped you learn it^[6]. It helped in mathematics as much as in science, and high-school students as much as younger ones^[6]. Those studies asked students to write about content, not about the conditions behind a step^[6].

That narrower point is settled by the mathematics alone. The line $x - 3 = (x-5)^2$ does not remember that it came from a square root. A number you carry forward stays visible; a restriction you merely thought about is gone once you turn the page.

Only one source above studied people of school age: the review of writing^[6]. The rest looked at university students, doctoral students or working mathematicians^{[1][2][3][5][4]}, and most concerns reading proofs, a task many high-school students never meet^[1][2][3][5]. That work describes how the error happens; it is not a measurement that carries to your homework.

Nothing here measures what this post is about: every outcome above is understanding, validating or learning material somebody else produced^{[1][2][3][5][4][6]}. Nobody tested whether writing your own conditions makes your solutions come out right more often; that is an argument from the mathematics, not a finding anyone reports.

One more limit: keeping the frame to one line is my judgement, not a result; the review supports having a structure at all^[4].

Take one problem you solved this week. Find every place you divided, squared, cancelled, or took a root or logarithm, and write the frame above it. Then substitute each answer back into the original equation, not into your third line, and see which survive.

If you work with a Learn4Less tutor, the question worth handing them is not "is this right" but "which step needed a condition I did not write".