Run the tableau: bring down, multiply, add, repeat
Match it to long division through the coefficient recurrence
Apply the Remainder Theorem, since the remainder is P(c)
What long division keeps rewriting
Run the long division from the opening example again, 2x3−5x2+3x−7 divided by x−3, but
keep an eye on which parts of the writing are decisions and which parts are ritual. Each stage
picks the next quotient term, multiplies it by x−3, and subtracts:
Stacking the three stages gives 2x3−5x2+3x−7=(x−3)(2x2+x+6)+11, so the quotient
is 2x2+x+6 and the remainder is 11.
Now look at what each subtraction really did. Multiplying the first quotient term 2x2 by x−3
gives 2x3−6x2, and subtracting it is exactly why the leading terms cancel: that half of every
step carries no information at all. The other half changes the next coefficient from −5 to
−5−(−6)=−5+3⋅2=1, since subtracting −6x2 is the same as adding6x2, which is
3 times the new quotient coefficient 2. Strip away the guaranteed cancellation and here is the
entire computation, written in bare numbers. The leading coefficient 2 came straight down, and
after that each new number was the next coefficient of the dividend plus 3 times the previous new
number:
−5+3⋅2=1,3+3⋅1=6,−7+3⋅6=11.
The numbers 2,1,6 are precisely the quotient coefficients, and the final number 11 is the
remainder. Long division computed nothing else. Synthetic division is a layout that computes exactly
this chain and refuses to write anything more.
The tableau: bring down, multiply, add
To divide a polynomial P(x) by x−c, build a three-row tableau:
In the corner, write c, the number that makes the divisor zero. For x−3 that is 3; for
x+2, which is x−(−2), it is −2. The corner always holds the divisor’s sign already
flipped, and that is why the tableau will add where long division subtracted.
Across the top row, write the coefficients of P(x) in descending order of degree, inserting a
0 for every missing power. Skipping a placeholder shifts every later column into the wrong
place, the same way dropping a zero digit corrupts a base-ten numeral.
Then follow the rhythm: bring down the leading coefficient; multiply it by c and write
the product under the next coefficient; add down that column. Next, multiply the sum by c
again, and repeat that pattern until the last column is added.
For the division above, the finished tableau is
322−561336−71811
Read the bottom row from the left. The last number is the remainder. Everything before it lists
the quotient coefficients, starting one degree below the dividend: here 2,1,6 means
Q(x)=2x2+x+6, with remainder 11. The degree drops by one because the divisor spent one
power of x: a cubic divided by a linear polynomial has a quadratic quotient.
The rhythm of the tableau for dividing 2x cubed minus 5x squared plus 3x minus 7 by x minus 3. Bring the leading 2 straight down, multiply by the corner number 3, write the product under the next coefficient, add down the column, and repeat. The bottom row holds the quotient coefficients 2, 1, 6 and, after the tick mark, the remainder 11.
Worked example 1Divide 3x3+4x2−11x+2 by x−2
The divisor is x−2, so the corner number is 2, and the top row is 3,4,−11,2. Bring
down the 3; then 3×2=6 and 4+6=10; then 10×2=20 and −11+20=9;
then 9×2=18 and 2+18=20:
2334610−1120921820
The quotient is Q(x)=3x2+10x+9 and the remainder is 20. To confirm it, expand the
division identity:
The check is a short expansion and catches most slips, so make it a habit.
Worked example 2Divide x4−3x2+5x−6 by x+2
Two traps live in this one. First, the divisor is x+2=x−(−2), so the corner number is −2,
not 2. Second, the dividend has no x3 term, so the top row needs a placeholder zero:
1,0,−3,5,−6. Now run the rhythm, keeping careful track of signs. Bring down the 1;
then 1×(−2)=−2 and 0+(−2)=−2; then (−2)×(−2)=4 and −3+4=1; then
1×(−2)=−2 and 5+(−2)=3; finally 3×(−2)=−6 and −6+(−6)=−12:
−2110−2−2−3415−23−6−6−12
The quotient is Q(x)=x3−2x2+x+3 and the remainder is −12. As a spot check, the
dividend at x=−2 is 16−12−10−6=−12, which matches the remainder. Why that check is
guaranteed to work is exactly where this lesson is headed.
Check your understanding
Use synthetic division to divide x3−7x+6 by x−2.
The x2 term is missing, so the top row needs a placeholder zero: 1,0,−7,6, with 2 in the corner.
211022−74−36−60
Bring down the 1; then 1×2=2 and 0+2=2; then 2×2=4 and −7+4=−3; then −3×2=−6 and 6+(−6)=0. The quotient is x2+2x−3 and the remainder is 0, so this division comes out exact. Forgetting the placeholder produces the wrong answer x−5 with remainder −4.
Why the shortcut is exactly long division
A pattern spotted in one example is not a proof, but the reason it always works is short. At every
step, the new bottom number is the next coefficient of P(x) plus c times the previous bottom
number, exactly the rule that turned −5,3,−7 into 1,6,11 above. That rule matches long
division step for step: long division’s leading term always cancels for free, and subtracting the
divisor’s −c term is the same as adding c times the last quotient coefficient. Chain these small
steps together, the same way the three lines of long division were stacked earlier, and they rebuild
the full identity P(x)=(x−c)Q(x)+r.
Long division’s leading-term cancellations are exactly the terms this shortcut never bothers to
write, and long division’s subtraction of a −c multiple is the same as the tableau’s addition of a
+c multiple. The tableau is not a different method that happens to agree with long division. It
is long division, holding only the numbers that matter.
Check your understanding
Divide x2−5x+6 by x−2 with long division. The first subtraction removes x2−2x, leaving −3x+6. In the synthetic-division tableau for the same division (corner 2, top row 1,−5,6), which single number equals that −3?
Long division's first subtraction takes the coefficient −5 and removes the −2 that comes from expanding x(x−2), which is the same as adding2: −5−(−2)=−5+2=−3. That is exactly the synthetic-division rule: multiply the previous bottom number by the corner 2 and add it to the next coefficient, −5+2⋅1=−3. The corner already holds the divisor's sign flipped, so the tableau's addition and long division's subtraction land on the identical number. The 1 is only the leading coefficient brought down before this step happens, and −6 is the next column's product, −3×2, not this one's.
The Remainder Theorem
The spot check in Worked Example 2 was no coincidence, and the reason is short enough to fit in
three lines.
The Remainder Theorem: dividing P(x) by x−c leaves remainder P(c)#
Divide any polynomial P(x) by x−c. The division algorithm from the last lesson produces a
quotient Q(x) and a remainder that is zero or of degree less than the divisor’s degree 1. In
either case the remainder is a constant polynomial. Call that constant r, so that
P(x)=(x−c)Q(x)+r
holds as an identity: the two sides are the same polynomial, so they agree at every single value
of x. In particular they agree at the one value chosen to make the first term vanish. Substitute
x=c:
P(c)=(c−c)Q(c)+r=0⋅Q(c)+r=r.
Whatever number Q(c) happens to be, multiplying it by zero erases it. The remainder therefore
equals P(c).
∎
Two remarks keep this proof honest. First, nothing was divided by zero: no fraction ever appears in
the argument. We evaluated a polynomial identity, valid for all x, at one convenient input. That
is also why the theorem needed no tableau: it follows just from the fact that dividing P(x) by
x−c always works and gives one right answer. Synthetic division is simply a fast way to find
that quotient Q and that remainder r.
Second, the theorem is an equality of numbers, not an approximation, and it runs in both
directions. Want the remainder without dividing? Evaluate P(c). Want P(c) without plugging in
powers? Divide and read the remainder. Go Deeper below follows that second direction, where it
turns into a fast way to evaluate any polynomial.
Check your understanding
Without dividing, find the remainder when x3−4x2+1 is divided by x+2.
The divisor is x+2=x−(−2), so the Remainder Theorem says the remainder is P(−2), the polynomial evaluated at −2.
P(−2)=(−2)3−4(−2)2+1=−8−16+1=−23
Watch both signs: (−2)3=−8 stays negative, while (−2)2=4 is positive and then picks up the −4 in front. Evaluating at +2 instead gives −7, the classic corner-sign mistake.
Two optional extensions
Two useful extensions of this lesson’s idea live in Go Deeper below, so the core path stays focused
on the tableau and the Remainder Theorem.
When the remainder is zero
In the first checkpoint the remainder came out 0, and the division identity collapsed to a pure
product: x3−7x+6=(x−2)(x2+2x−3), with nothing left over. The division is exact, and
x−2 is a factor of the cubic. Since the remainder isP(c), the remainder is zero exactly
when P(c)=0, that is, exactly when c is a root of the polynomial. A zero remainder is how
synthetic division certifies that x−c is a factor, the link the next lesson calls the Factor
Theorem.
Common mistakes
Practice
Multiple Choice Questions (MCQ)
Progressively harder sets of questions. Each opens on its own page.
Practice problems at the level of the course, to be worked out on paper. Hints one at a
time, then the answer or the full worked solution, with your progress kept in this browser.
You can skip this and keep going. Read it if you want to know more.
The full proof, for any polynomial and any corner number
The earlier section showed, using one cubic’s numbers, that the tableau’s rhythm rebuilds long
division. Here is the same argument written in symbols, so it works for every polynomial and every
corner number c at once.
The tableau computes the true quotient and remainder#
Let P(x)=anxn+an−1xn−1+⋯+a1x+a0 with n≥1, and fix any number c.
Define the bottom-row numbers exactly as the rhythm does: the bring-down step sets
bn−1=an, each multiply-and-add step sets
bk−1=ak+cbkfor k=n−1,n−2,…,1,
and the final addition sets r=a0+cb0. Collect the bottom row into the polynomial
Q(x)=bn−1xn−1+⋯+b1x+b0. The claim is that P(x)=(x−c)Q(x)+r.
Expand the product. Multiplying Q(x) by x shifts every coefficient up one degree, and
multiplying by −c scales it, so
Now add r and compare against P(x) one coefficient at a time. The xn coefficient is
bn−1, which the bring-down step made equal to an. For each k from 1 to n−1, the
coefficient of xk is bk−1−cbk, and the recurrence says bk−1=ak+cbk, so
this coefficient is (ak+cbk)−cbk=ak. The constant term is r−cb0, and the
final addition says r=a0+cb0, so the constant term is a0. Every coefficient matches,
which proves the identity P(x)=(x−c)Q(x)+r.
It remains to say why Q and r are the quotient and remainder rather than merely a way of
rewriting P. The number r is a constant, so it is either zero or of degree less than 1, the
degree of the divisor x−c. The division algorithm from the last lesson has a uniqueness half. By
that half, only one pair consisting of a quotient and a remainder meeting that condition can satisfy
such an identity. The tableau produced that pair.
∎
Horner's method: reading the remainder as an evaluation
Unwind the tableau’s recurrence for a cubic P(x)=a3x3+a2x2+a1x+a0 and something
familiar appears. The bottom row is b2=a3, then b1=a2+cb2, then b0=a1+cb1, then r=a0+cb0. Substitute each line into the next:
r=a0+c(a1+c(a2+ca3))=((a3c+a2)c+a1)c+a0.
Multiply the nest out and you get a3c3+a2c2+a1c+a0, which is P(c) on the nose. So
the tableau’s final number computes P(c) by direct multiply-and-add arithmetic, using the same
recurrence already shown to match long division, and the same unwinding works in every degree. This
nested way of evaluating a polynomial is called Horner’s method, and the tableau performs it
with one multiply-and-add for each coefficient after the leading one. So a degree-n polynomial
costs just n multiplications, instead of building each power of c from scratch. As a bonus, the
intermediate numbers bk are not waste products. They are the quotient coefficients, so evaluating
P(c) this way hands you the whole division for free.
Worked example 3Evaluate P(4) for P(x)=x5−4x4+2x3+x−10
Computing 45, 4⋅44, and 2⋅43 separately invites arithmetic errors. Run the
tableau instead. The top row needs a placeholder for the missing x2 term:
1,−4,2,0,1,−10, with 4 in the corner:
411−44020208813233−10132122
Each step is a single multiply and add: 1×4=4 and −4+4=0; then 0×4=0
and 2+0=2; then 2×4=8 and 0+8=8; then 8×4=32 and 1+32=33;
finally 33×4=132 and −10+132=122. By the Remainder Theorem,
P(4)=122,
and the brute-force check agrees: 1024−1024+128+4−10=122. Five multiplications replaced
the whole tower of powers, and the bottom row even reports the quotient
x4+2x2+8x+33 from dividing P(x) by x−4.
Dividing by a divisor that isn't monic
The tableau, as built, divides only by x−c: a linear divisor whose leading coefficient is 1.
A divisor like 2x−1 does not fit the corner directly, but factoring out the leading coefficient
rescues it, because 2x−1=2(x−21). Divide by the monic part
x−21 first, then hand the leftover factor of 2 to the quotient.
Worked example 4Divide 2x3+5x2−7x+6 by 2x−1
Since 2x−1=2(x−21), run the tableau with corner number 21.
Bring down the 2; then 2×21=1 and 5+1=6; then
6×21=3 and −7+3=−4; then −4×21=−2 and
6+(−2)=4:
2122516−73−46−24
The tableau says 2x3+5x2−7x+6=(x−21)(2x2+6x−4)+4. Now
rewrite the monic factor as 21(2x−1) and let the 21 act on the quotient:
2x3+5x2−7x+6=(2x−1)(x2+3x−2)+4.
Dividing by 2x−1 gives quotient x2+3x−2 and remainder 4. Notice exactly what changed in
that rewrite and what did not. The quotient got divided by 2, while the remainder stayed put,
because the factor of 2 was absorbed inside the product term only. Halving the remainder as well
is the standard mistake here. Consistently with the Remainder Theorem, the remainder is
P(21)=41+45−414+424=4.
The same reasoning in general: to divide P(x) by ax−b with a=0, run the tableau with
corner c=ab, divide the resulting quotient by a, and leave the remainder alone. In
particular the Remainder Theorem extends verbatim: the remainder on dividing by ax−b is
P(ab), still a single evaluation.
A bit of history (optional)
A name on a method tends to record who advertised it, not who found it.
The three rows you just ran are called Horner’s method in the English-speaking world. William George
Horner, a schoolmaster in Bath, presented the scheme in 1819. He offered it as the engine inside a
method for hunting the roots of an equation. The name stuck for a reason with nothing to do with the
arithmetic. An influential writer of the day pushed it for years in print, and the English textbooks
copied him.
The same tableau had gone to press fifteen years earlier, in a prize memoir by Paolo Ruffini. That is
the physician who argued the quintic unsolvable, two lessons back. He wrote in Italian, for a
readership that barely overlapped with the one reading Horner. In Italy and Spain the three rows are
still taught as Ruffini’s rule.
Neither man was first, and neither claim is as old as the method. Rows of coefficients were being
multiplied and added in this exact rhythm centuries before either memoir, in traditions that never
entered the argument at all.
The rhythm outlived every name on it because it is cheap. It spends one multiply and one add for
each coefficient after the leading one. That is why the nested form used to find P(c) in Go Deeper
still runs in the software that graphs a curve for you.