Why Your Partial Fraction Decomposition Keeps Failing in Integral Calculus

Partial fraction decomposition is a template. Given $\frac{P(x)}{Q(x)}$, you factor $Q$, write one fraction for every power of every factor — a constant over a linear power, $\frac{A}{x-a}$, a linear numerator over an irreducible quadratic power, $\frac{Bx+C}{x^2+bx+c}$ — then clear denominators and solve. Every step is mechanical, which is why it fails quietly: set it up where the form does not fit and the first sign of trouble is an impossible equation ten lines down.

A review of research on mathematics learning found that knowing how to carry out a method and understanding why it works each help build the other, in both directions^[1]. That review did not study partial fractions, and it reports which to teach first is largely untested^[1]. The diagnosis here is a reading of the mathematics, not a research finding: memorizing skips the entry conditions, which never appear in the worked steps.

The first condition is on degrees: the method applies directly only when $\deg P < \deg Q$. Consider

$$\int \frac{x^3 + 2x^2 + 3}{x^2 + 1} \, dx.$$

The numerator has degree 3, so divide first:

$$\frac{x^3 + 2x^2 + 3}{x^2 + 1} = x + 2 + \frac{1 - x}{x^2 + 1},$$

checking that $(x+2)(x^2+1) + (1-x) = x^3 + 2x^2 + 3$. That last term needs no decomposition, since $x^2 + 1$ is irreducible — only splitting into $\frac{1}{x^2+1} - \frac{x}{x^2+1}$. So the integral is

$$\frac{x^2}{2} + 2x + \arctan x - \frac{1}{2}\ln\!\left(x^2+1\right) + C.$$

A review of the worked-examples research concluded that good teaching pairs several worked solutions per problem type with a matching practice problem^[2]. It set out design principles, not measured results, so treat it as a reason to work the next one with a pencil^[2].

Take $\int \frac{5x+7}{x^2-x-2} \, dx$. Factor the denominator completely, $x^2 - x - 2 = (x-2)(x+1)$, and write

$$\frac{5x+7}{(x-2)(x+1)} = \frac{A}{x-2} + \frac{B}{x+1}, \qquad 5x + 7 = A(x+1) + B(x-2).$$

Setting $x = 2$ gives $17 = 3A$, so $A = \frac{17}{3}$. Setting $x = -1$ gives $2 = -3B$, so $B = -\frac{2}{3}$. The coefficients check: $A + B = 5$ and $A - 2B = 7$. The integral is

$$\frac{17}{3}\ln|x-2| - \frac{2}{3}\ln|x+1| + C.$$

In two classroom experiments with algebra students, the first found that those who explained worked examples while practising understood the algebra better than those who only practised^[3]. The second compared example types, and wrong ones looked especially useful — a result the team called suggestive, not settled^[3].

There is a limit: in two experiments where learners studied probability problems, wrong solutions helped only those who already knew the topic well; those still shaky did better with correct solutions alone^[4]. If the example above felt unfamiliar, skip what follows.

Here is a decomposition with a real error:

$$\frac{2x+1}{(x-1)^2(x+2)} = \frac{A}{(x-1)^2} + \frac{B}{x+2}.$$

Clearing denominators gives $2x + 1 = A(x+2) + B(x-1)^2$, with right side $Bx^2 + (A-2B)x + (2A+B)$. The $x^2$ term forces $B = 0$ and the $x$ term forces $A = 2$, but the constant term demands $2A + B = 1$, that is $4 = 1$. No such constants exist.

A squared factor needs two terms, not one:

$$\frac{2x+1}{(x-1)^2(x+2)} = \frac{A}{x-1} + \frac{B}{(x-1)^2} + \frac{C}{x+2}$$

Now $2x+1 = A(x-1)(x+2) + B(x+2) + C(x-1)^2$. At $x = 1$, $3 = 3B$ gives $B = 1$; at $x = -2$, $-3 = 9C$ gives $C = -\frac{1}{3}$. Matching $x^2$ coefficients gives $A + C = 0$, so $A = \frac{1}{3}$. Both sides equal $\frac{1}{2}$ at $x = 0$. Integrating gives $\frac{1}{3}\ln|x-1| - \frac{1}{x-1} - \frac{1}{3}\ln|x+2|$, plus a constant.

Drop the $\frac{A}{x-1}$ term and the coefficient equations have no solution — unless that coefficient was zero all along, which you cannot know before solving. So write both, every time.

You might think the method runs out on a denominator like $x^3 + x + 1$. It does not: over the real numbers every polynomial factors into linear and irreducible quadratic factors, so a proper rational function always has a decomposition.

The rational root theorem offers only $1$ and $-1$, which give $3$ and $-1$. The cubic does cross zero near $-0.6823$; dividing that root out leaves roughly $x^2 - 0.6823x + 1.4656$, with no real roots.

What fails is the bookkeeping, not the theory. The root is irrational, and Cardano's formula does write it down exactly, as $\sqrt[3]{-\frac{1}{2} + \sqrt{\frac{31}{108}}} + \sqrt[3]{-\frac{1}{2} - \sqrt{\frac{31}{108}}}$ — but carrying that expression through a decomposition by hand is hopeless. If your denominator resists both inspection and the rational root theorem, re-read the question before you doubt your algebra.

The judgment "does partial fractions apply here?" only gets exercised when it is in doubt, and a heading naming the method removes the doubt.

In one experiment, college students who practised several kinds of problem mixed together did far better on a test a week later than those who practised one kind at a time^[5]. In a larger classroom study, fifty-four seventh-grade mathematics classes periodically completed assignments of one kind or the other across four months, the mixed ones arranged so no two problems in a row needed the same method^[6]. Those classes averaged 61% on a surprise test a month later, against 38% for the blocked classes — seventh-graders on seventh-grade material, not calculus students^[6].

Assemble your own set: a partial-fractions problem, a substitution, an integration by parts, and a rational function with too big a numerator. Shuffle them and cover the headings.

Yes-or-no checklists are easy to answer without thinking. A review that pooled 64 studies found that prompting people to explain things to themselves, as they study or solve problems, is a promising way to help learning in many settings^[7]. Turning each check into a why-question follows that idea^[7]:

• Why does this fraction need dividing before anything else?
• Why does this factor take a constant on top and that one a linear numerator?
• Why does a squared factor need two terms rather than one?
• Why do those values of $x$ isolate the constants I want?

Answer them out loud. The first one you cannot answer is where your decomposition breaks.

None of the studies behind this post examined partial fractions, rational functions, or calculus of any kind^{[1][5][6][2][3][4][7]}; the closest subject matches are equation-solving with algebra students^[3] and probability problems^[4]. The practice advice carries over as a general claim about learning mathematics, not as anything measured here.

The diagnosis is the clearest gap: that memorizing a template hides its entry conditions is reasoning about the mathematics, not a tested result, and the review behind it found only that procedures and understanding build each other^[1]. The worked-examples review recommends a structure rather than reporting a gain^[2]. The case for wrong examples is unsettled: one team called its result suggestive rather than conclusive^[3], another found the benefit reversed for learners with weak background knowledge^[4].