You can find the function behind a table of values by testing the input-output pattern: constant first differences mean a linear function, constant second differences mean a quadratic, and a constant ratio between consecutive outputs means an exponential. Every rule below starts from the same raw material β two rows of paired numbers β and ends with an equation you can check against the original cells.
What trips most people up isn't the arithmetic. It's choosing the wrong family before examining the pattern, then forcing a straight line onto data that bends. The method here works in a fixed order, so the table itself tells you which family to use instead of you guessing first.
Read the Table's Input-Output Pattern Before Choosing a Model
Label your columns before touching a calculator. Put the input values (usually x) in the left column in increasing order and the outputs (usually f(x) or y) in the right, spaced evenly. Even spacing is not optional: if the x-values jump by 2, then 5, then 1, difference methods still work but every division must account for the actual gap, which is where most errors start.
Calculate first differences next β the change in output each time the input advances one step. Write them in a third column beside the gaps. For the pairs (1, 3), (2, 5), (3, 7), (4, 9), the differences read 2, 2, 2. That single column already narrows the family: one layer of constant differences is the signature of a linear rule, and nothing more complicated is needed.
If the first differences themselves change, add another column: second differences, which are the differences of the differences. The Khan Academy algebra curriculum teaches this ladder as the standard test for polynomial degree β constant at the first layer means degree 1, constant at the second means degree 2, and so on. Two layers of your own arithmetic beat any guess.
One habit pays off here: keep the original table intact and do all scratch work to the right. Filling in a difference where an output belongs is the single most common mechanical mistake students make on exams, and it silently corrupts every later step.
The Four-Step Procedure That Works for Any Table
Follow the same sequence every time, regardless of family. The steps below are in the order that catches errors fastest, because the family check comes before any equation building.
1: Verify the inputs are equally spaced. Subtract consecutive x-values and confirm a single common gap. If not, note each gap; it changes the divisor in step 3. 2: Build the difference ladder or the ratio column. First differences, then second differences if needed; separately, divide each output by the one before it. 3: Identify the family from whichever column goes constant. Constant first difference = linear; constant second difference = quadratic; constant ratio = exponential. 4: Solve for the parameters, then test a cell you didn't use. Substitute one original pair into the general form, find the remaining constant, and verify with the last row of the table.
Step 4 is where the method earns its keep. Suppose the differences point to a linear rule with slope 2 and you've written f(x) = 2x + b. Substituting the pair (3, 7) gives 7 = 6 + b, so b = 1. Testing the unused pair (4, 9): 2(4) + 1 = 9. The check costs ten seconds and catches sign errors that survived every earlier stage.
When the final column refuses to go constant at any layer you're willing to compute, the data isn't polynomial. That's not a failure β it's a signal to test for an exponential ratio or to accept that the table comes from something richer, like a trigonometric or piecewise rule, which needs different tools.
Constant First Differences Mean a Linear Function
A linear function has the form f(x) = mx + b, where m is the slope and b the y-intercept. When first differences are constant, that constant is the slope. For the table (0, 5), (1, 8), (2, 11), (3, 14), every difference equals 3, so m = 3, and the output at x = 0 gives b = 5 directly. The function is f(x) = 3x + 5.
The intercept shortcut only works when x = 0 appears in the table. If it doesn't, substitute any pair and solve. Take (2, 11) with slope 3: 11 = 3(2) + b, so b = 5 again. Same answer, one extra line. When the inputs are spaced by something other than 1 β say x jumps by 3 each row β divide the output difference by the input gap to get the true slope, because a difference of 9 across a gap of 3 is a slope of 3, not 9.
Watch the direction of travel. A table whose outputs fall by 4 per step has slope β4, and students who record differences as positive magnitudes lose the sign. Write each difference as later value minus earlier value, in that order, and the sign takes care of itself.
Linear tables show up constantly in real data: a phone plan charging a flat fee plus a per-minute rate, a car burning fuel at a steady highway rate, a savings account with fixed weekly deposits and no interest. For the fuel case, if the table lists 12 gallons at 300 miles and 8 gallons at 200 miles, the differences give a slope of β25 miles per gallon of fuel burned β negative because miles remaining drop as gallons used rise.
Constant Second Differences Mean a Quadratic Function
Quadratics have the form f(x) = axΒ² + bx + c, and their second differences are constant at exactly 2a. That factor of 2 is the detail most tables of values pages skip, and it saves a division: if the second difference is 8, then a = 4. For the classic squares table β (1, 1), (2, 4), (3, 9), (4, 16) β first differences run 3, 5, 7 and second differences run 2, 2, so a = 1, matching f(x) = xΒ².
Finding b and c takes two substitutions from the original pairs. With a known, plug in two rows to build two equations, then subtract them to eliminate c. Using (1, 1) and (2, 4) with a = 1: the first gives 1 = 1 + b + c, the second gives 4 = 4 + 2b + c. Subtracting yields 3 = 3 + b, so b = 0, and back-substitution gives c = 0. The check against (4, 16) confirms it.
A worked case with all three terms: the table (1, 6), (2, 11), (3, 20), (4, 33) has first differences 5, 9, 13 and second differences 4, 4. So a = 2. Substituting (1, 6) and (2, 11) gives 6 = 2 + b + c and 11 = 8 + 2b + c; subtracting yields 5 = 6 + b, so b = β1, and c = 5. The function is f(x) = 2xΒ² β x + 5, and plugging x = 4 returns 33 as the table requires.
Quadratics model falling objects and area problems, which is why they appear in physics and geometry courses together. A ball's height table typically shows symmetric differences around its peak, and the constant second difference encodes the acceleration β on Earth, roughly β16 feet per second squared in the imperial units most US textbooks use, per the standard gravity figure of 32 feet per second squared halved by the one-half in the position formula. Your table's second difference divided by 2 recovers that coefficient.
A Constant Ratio Between Outputs Means an Exponential Function
Exponentials take the form f(x) = a Β· bΛ£. The test is division, not subtraction: divide each output by the previous one and look for a constant quotient, which is the base b. For (0, 3), (1, 6), (2, 12), (3, 24), every ratio is 2, so b = 2, and the output at x = 0 gives a = 3. The function is f(x) = 3 Β· 2Λ£.
Ratios demand more care than differences because they're sensitive to zeros and signs. A table with an output of 0 anywhere breaks the division, and a table mixing positive and negative outputs produces ratios that flip sign without a stable base. In those cases the exponential test simply doesn't apply, and that negative result is itself useful information.
Growth and decay both appear. A ratio of 0.5 means the quantity halves each step β radioactive decay, cooling coffee, a depreciating asset under a fixed percentage rate. If the input gap isn't 1, the base you compute is the growth over that gap, and you need a root to convert it to a per-unit rate: a ratio of 4 across a gap of 2 means the base is the square root of 4, which is 2 per unit.
Compound interest is the everyday example. At 5% annual growth, the ratio between consecutive years' balances is 1.05, so a table of yearly balances divides out to a constant 1.05 and the function is the starting balance times 1.05 raised to the year. If a table showed a constant ratio of 1.05 but the entries were spaced every two years, the annual rate would be the square root of 1.05 minus 1, about 2.47% β a distinction that changes the answer materially.
Difference Tables Compared: Linear, Quadratic, and Exponential Side by Side
The three families separate cleanly once you build the right column, and comparing them on identical criteria locks in the recognition. The table below uses one example of each, chosen so the deciding column is visible at a glance.
| Family | Example pairs | First differences | Second differences | Ratio | Rule |
|---|---|---|---|---|---|
| Linear | (0, 5), (1, 8), (2, 11), (3, 14) | 3, 3, 3 | 0, 0 | not constant | f(x) = 3x + 5 |
| Quadratic | (1, 1), (2, 4), (3, 9), (4, 16) | 3, 5, 7 | 2, 2 | not constant | f(x) = xΒ² |
| Exponential | (0, 3), (1, 6), (2, 12), (3, 24) | 3, 6, 12 | 3, 6 | 2, 2, 2 | f(x) = 3 Β· 2Λ£ |
Read the columns in order of cheapness: differences first, because subtraction is easier than division and handles negatives without special cases, then ratios only when differences fail. The ratio column for the linear example isn't constant β 8/5, 11/8, 14/11 all differ β which is exactly why the exponential test fails there and the linear one succeeds.
The second-difference column for an exponential is never constant at the first layer but does settle at the second layer in a deceptive way: for 3 Β· 2Λ£ it reads 3, 6, 6, which looks almost constant. Don't be fooled. A genuine quadratic's second differences are identical from the first entry, while an exponential's settle only after the first gap or not at all across longer tables β compute four or five rows before deciding.
Two of these families also diverge in long-run behavior, which matters when you're choosing a model for projections. A quadratic with positive leading coefficient eventually outpaces any exponential, but an exponential with base above 1 beats any quadratic early on. If your table spans only a few rows, either can fit; if it spans ten, the fit tells you which process is actually at work.
Work Backward From the Answer Choices Without Full Solving
Multiple-choice versions of this question rarely need the full parameter hunt. Test the answer choices against two or three table cells and eliminate β the wrong functions usually fail on the very first pair. If the choices are f(x) = 2x + 1, f(x) = xΒ² + 1, and f(x) = 2Λ£, the pair (2, 5) eliminates the third immediately since 2Β² = 4, and the pair (3, 7) eliminates the second since 9 + 1 = 10.
This shortcut is what the College Board's SAT math sections reward, because the digital format gives you a built-in calculator and limited time β checking three options against two cells takes less than a minute, while deriving the parameters from scratch can take three. The catch: it only works when choices are given. Free-response tables still need the full procedure.
Speed up elimination by sorting the choices by family first. If your difference ladder showed constant first differences, cross out every quadratic and exponential option before plugging anything in, then test the remaining linear candidates. Two candidates, one pair of cells, done.
The risk with shortcuts is anchoring β falling in love with an answer that fits the first pair and skipping verification on the last. Always test the final row. Options are frequently designed so two candidates match the early cells and only one survives the last, which is precisely the discrimination the question is testing.
Verify Your Function Against Every Cell, Not Just the First
A function that matches two cells proves little; three parameters can always accommodate two points. The real test uses every pair in the table, and the fastest way to run it is to evaluate your candidate at each input and compare to the stored output. Any mismatch means either the family is wrong or a parameter slipped.
Sign errors are the top offender. A quadratic derived as f(x) = 2xΒ² + x + 5 instead of 2xΒ² β x + 5 fits the first two rows and fails the third, which is why the b = β1 term above matters. Second is the intercept when x = 0 is absent β solving for b from one equation is fine, but only the multi-cell check proves the solve was right.
A useful discipline: keep a small verification table with three columns β input, table output, your function's output β and a fourth column of differences. All zeros in the fourth column and you're done. A single nonzero entry points you to the row that broke and, usually, to which of a, b, or c is off.
Graphing the points alongside your candidate curve is the visual version of the same check and catches family errors that arithmetic misses. If the table's points curve upward but your linear fit passes through three of four, the fourth point will sit visibly off the line β a mismatch that's easy to overlook in columns of numbers but impossible to miss on a plot.
Common Mistakes That Produce the Wrong Equation
Mixing subtraction and division tests is the error that wastes the most time. Students compute differences, see them change, and immediately assume quadratic β but an exponential also has changing differences. Build both the difference ladder and the ratio column before committing to a family, and let the constant one decide.
Forgetting the factor of 2 in the quadratic coefficient produces answers that are off by a clean factor and look plausible. If second differences are 12 and you write a = 12, every cell will miss by 6xΒ². Divide by 2 first: a = 6. That division is not optional bookkeeping; it falls directly out of the algebra of (x+1)Β² β xΒ².
Assuming x = 0 sits in the table is a third recurring slip. Exponential tables without an x = 0 row require a substitution to find the starting value, and linear tables without one require solving for the intercept. The y-intercept is only free when the input 0 is present.
Treating uneven inputs as if they were evenly spaced silently inflates or deflates slopes. A table jumping by 3 each row with output differences of 9 has slope 3, not 9, and quadratics need the same correction before the 2a division. Check the input gaps in step 1 and adjust every divisor downstream.
FAQ
What if the differences never become constant?
Then the table isn't describing a polynomial of low degree, and you should test the ratio column for exponential growth or decay. If ratios also fail, the underlying rule may be trigonometric, piecewise, or rational, which needs different identification techniques. The constant-difference ladder is a first filter, not a universal key.
Can a table fit more than one function?
Yes β any finite set of points fits infinitely many curves, which is why no exam table should be trusted beyond checking a candidate you derive from the pattern. The family test identifies the simplest rule consistent with the differences, and the word "simplest" is the assumption doing the work. With enough rows, the pattern becomes hard to fake.
How do I find the equation when the x-values skip by 2 instead of 1?
Divide the output difference by the input gap to get the per-unit slope or base adjustment. A linear table with outputs rising 8 across a gap of 2 has slope 4, and an exponential with a ratio of 9 across a gap of 2 has a per-unit base of 3, since the square root of 9 is 3. Skipping this division is the most common uneven-spacing error.
What's the quickest way to check my answer on a test?
Evaluate your candidate at the last input in the table and compare to the stored output. If they match, you've tested the cell the question designer most likely varied. A single pair of cells takes seconds and catches the majority of parameter errors, especially sign mistakes in the middle term of a quadratic.
Does a constant second difference always mean a quadratic?
It means the data comes from a quadratic in the chosen input variable, provided the inputs are evenly spaced. If the inputs aren't evenly spaced, second differences can look constant by coincidence. Confirm even spacing first, then trust the column. Uneven tables need the divided-difference version of the same technique.
Why does my exponential function fit the table but look wrong on a graph?
You may have computed the base over a multi-unit gap without taking the root, which stretches the growth rate. If consecutive table entries are two years apart and the ratio is 1.21, the annual base is 1.1, not 1.21. Graphing the candidate against all points exposes the inflation immediately as a curve that overshoots the data.
How do I tell a quadratic from an exponential when the table is short?
A four-row table is genuinely ambiguous, and you should extend it or compute a third row of differences. Quadratics have exact constancy at the second layer from the start; exponentials drift. If the source data can't be extended, state both candidate functions and note the ambiguity rather than asserting one.
What does a constant ratio below 1 mean?
It means decay: the quantity shrinks by a fixed percentage each step. A ratio of 0.8 is a 20% drop per unit, and the function takes the form a Β· 0.8Λ£. Half-life tables are the canonical example, and their ratios sit exactly at 0.5 per half-life interval regardless of the substance.