You decide by testing how the output changes: compute the differences between consecutive y-values and see if those differences are constant (that's linear), or compute the ratios of consecutive y-values and see if those ratios are constant (that's exponential). A table with x = 1, 2, 3, 4 and y = 5, 8, 11, 14 is linear because every step adds 3; y = 5, 10, 20, 40 is exponential because every step multiplies by 2. Neither pattern holds when the data is quadratic, or when the x-values aren't evenly spaced.

The Difference Test: Constant First Differences Mean Linear

A linear function has the form y = mx + b, where m is the slope and b is the y-intercept. The defining property is that equal steps in x produce equal steps in y, no matter which two consecutive points you pick. If your table goes from x = 2 to x = 3 and y rises by 7, then x = 3 to x = 4 should also raise y by 7.

For the table x = 0, 1, 2, 3 and y = 4, 9, 14, 19, the first differences are 5, 5, and 5. Constant first differences confirm linearity. The slope is 5 in this case, and the y-intercept is 4, so the function is y = 5x + 4. You can verify by checking that each output equals 5 times its input plus 4.

Watch how the starting value behaves. A linear function with slope 5 that starts at x = 0 with y = 4 hits y = 0 at x = −0.8, so the line crosses the x-axis between table entries. That's normal for linear functions and doesn't disqualify the data. Linear functions pass through zero exactly once (unless the slope is zero, when they never do).

The most common failure is subtracting the wrong way. If you calculate differences as y₁ − y₂ in one pair and y₂ − y₁ in the next, the signs flip and you'll misread constant numbers as non-constant. Always subtract in the same direction: later y minus earlier y. Sign matters when y decreases, but the magnitude stays constant for linear functions even when the inputs are increasing.

The Ratio Test: Constant Ratios Mean Exponential

An exponential function has the form y = a · bˣ, where a is the initial value at x = 0 and b is the growth (or decay) factor. The defining property is that equal steps in x multiply y by equal factors. If your table goes from x = 1 to x = 2 and y doubles, then x = 2 to x = 3 must also double y.

Take x = 0, 1, 2, 3 and y = 6, 18, 54, 162. Divide 18 by 6 to get 3, 54 by 18 to get 3, and 162 by 54 to get 3. Constant ratio of 3 means exponential. Because the y-value at x = 0 is 6, the function is y = 6 · 3ˣ. You can check any row: 6 · 3¹ = 18, 6 · 3² = 54, 6 · 3³ = 162.

Division by zero destroys the ratio test. If any y-value in the table is 0, the next ratio is either 0 (if the numerator is 0) or undefined (if the denominator is 0). A true exponential function never outputs 0 for real x, so a table with a 0 in the y-column is not exponential. It could be linear, or it could be something else entirely.

Ratios near 1 signal slow growth; ratios below 1 signal decay. A half-life table with y = 80, 40, 20, 10 has a constant ratio of 0.5, which is exponential decay with base 0.5. A ratio of exactly 1 means the function is constant, and constant functions count as linear with slope 0 but not as exponential because the base must be positive and not equal to 1 in standard definitions. Some textbooks exclude b = 1 from exponential functions; check your course's convention if the ratio comes out to exactly 1.

Worked Side-by-Side Comparison: Two Tables, Same x-Values

Comparing two tables on the same x-values is the fastest way to see the tests side by side. Below are two tables with identical x-values but different behavior, tested with both methods.

x y (Table A) First difference Ratio to previous y y (Table B) First difference Ratio to previous y
0 3 — — 2 — —
1 7 4 7/3 ≈ 2.33 6 4 3
2 11 4 11/7 ≈ 1.57 18 12 3
3 15 4 15/11 ≈ 1.36 54 36 3
4 19 4 19/15 ≈ 1.27 162 108 3

Table A is linear because column three is constant at 4, and column four drifts downward with no pattern. Table B is exponential because column six is constant at 3, and column five accelerates (4, 12, 36, 108). Notice the trap: Table B also has one difference of 4, matching the first row of Table A. A single matching difference proves nothing. You need the full column.

That trap trips students on exams. A table where the first difference happens to equal the first ratio of a neighboring column invites you to look at one cell and decide. Build the complete column of at least three differences or ratios before you commit.

Why the x-Values Must Be Evenly Spaced Before Either Test Works

Both tests assume equal steps in x. If x jumps 1, then 2, then 3, the y-differences can't be compared directly because they cover different distances. A table of x = 1, 2, 4, 8 with y = 3, 6, 12, 24 looks linear by naive subtraction (differences of 3, 6, 12 aren't constant, so a hasty reader rejects linear and moves on) and looks exponential by naive division (ratios of 2, 2, 2 look constant, so they call it exponential). The actual pattern is y = 3x: doubling x doubles y, which is direct proportionality, a linear form with b = 0.

The fix is to verify spacing before testing. Subtract consecutive x-values and confirm they match: 1, 1, 1 works, but 1, 2, 4 fails. If the x-values are unevenly spaced, you can still classify the function, but you have to normalize. One method: represent each y as a function of its x-value on x = 0, 1, 2, 3 by treating the actual x-column as the input to y = mx + b or y = a · bˣ and using substitution to solve for the parameters.

For the y = 3x example, solving y = a · bˣ with x = 1, y = 3 and x = 2, y = 6 gives b = 2; then x = 2 to x = 4 gives y = 6 to 12, which would require b = √2, not 2. The mismatch proves it isn't exponential.

Half-Lives, Doubling Times, and Why Real Data Rarely Lands Exactly

Textbook tables produce clean ratios because textbook authors pick them that way. Real measurements don't. A population table might show 100, 149, 224, 335, 502 with ratios of 1.49, 1.50, 1.50, 1.50 — close, but the first ratio is 1.49 because of rounding. The function is exponential with base about 1.5; the tiny deviation is noise from rounded inputs.

How close is close enough depends on how the values were rounded. If each y is given to the nearest whole number, expect ratio deviations up to about half a unit divided by the smaller y-value. For y around 100 rounded to integers, that's roughly 0.005, so ratios of 1.495–1.505 all fit a true base of 1.5. For y around 10, the allowance is much wider, about 0.05.

Exponential decay in half-life tables works the same way: y = A · 0.5^(x/h), where h is the half-life in x-units. If the table is x = 0, 5, 10, 15 with y = 320, 160, 80, 40, then every 5 units halves y, so the half-life is 5 x-units and the ratio per 5-unit step is 0.5. If your x-steps are 1 unit apart, the per-step ratio is 0.5^(1/5) ≈ 0.871, not 0.5.

That's a common exam trap. If the table's x-step is 1 but the half-life is 5, you'll see ratios around 0.87, not 0.5. Students who expect a clean 0.5 misclassify the function as "not exponential" when it clearly is.

When Neither Test Fits: Quadratic, Cubic, and None-of-the-Above Tables

Tables that pass neither test exist, and they aren't failures of your method. A quadratic y = x² gives x = 1, 2, 3, 4 with y = 1, 4, 9, 16. First differences are 3, 5, 7 (not constant), and ratios are 4, 2.25, 1.78 (not constant). But the second differences are 2, 2, 2 — constant second differences prove a quadratic. The next level up, third differences, characterizes cubics.

Pattern of y Constant first difference Constant ratio Constant second difference
Linear, slope 3 3 varies 0
Exponential, base 2 doubles 2 doubles
Quadratic, y = x² 1, 3, 5, 7 varies 2
Constant, y = 5 0 1 0

Second differences are computed by taking the first differences and then differencing those. For y = x², first differences are 1, 3, 5, 7 and second differences are 2, 2, 2. If you see constant second differences but not constant first differences, it's quadratic. Linear is the special case where second differences are all zero, which is why linear functions are sometimes called "degree 1 polynomials" and quadratics "degree 2."

The practical rule: check linear first (constant first differences), then exponential (constant ratios), then quadratic (constant second differences). Any table that fails all three is some other function — perhaps a higher-degree polynomial, a trig function, a piecewise function, or noise. Don't force the wrong label.

Estimating Missing Entries and Choosing the Right Model

Exam questions often give a partial table and ask you to fill in a gap. Identify the pattern first, then extend it. If the table shows exponential behavior with a ratio of 1.4 and you need y at x = 5 when y(x = 4) = 13.7, multiply: 13.7 · 1.4 = 19.18. If the table shows linear behavior with a slope of 4 and you need y at x = 5 when y(x = 4) = 15, add 4: y = 19.

For exponential gaps, divide instead when you're going backward. If y(3) = 20 with a ratio of 1.4, then y(2) = 20 / 1.4 ≈ 14.29, not 20 · 1.4. Direction matters.

Real-world decisions about which model fits rarely depend on a textbook table. Mold growth modeling in food science, for instance, is often approximated as exponential at early stages and logistic later because resources run out. A table showing 10, 20, 40, 80, 150, 220 for consecutive hours fits exponential for the first four rows but flattens after that; the correct label is "exponential in the early phase, logistic overall." Linear models often win for short intervals even when underlying growth is exponential — near x = 0, the curve y = eˣ and the line y = 1 + x are close enough for many estimates.

Calculator and Spreadsheet Shortcuts That Don't Lie

Graphing calculators and spreadsheets automate the test, but the automation can hide the failure cases. In Excel or Google Sheets, add a helper column for ratios (e.g., =B2/B1) and another for differences (=B2−B1), then look at the columns. A constant column means you've found the pattern. Dragging formulas down is quick, and conditional formatting can highlight which column is flatter.

In Desmos, type the x-column and y-column as a table, then plot. A linear function traces a straight line; exponential traces a curve that flattens near the x-axis and shoots upward. The visual check confirms your arithmetic but shouldn't replace it. A table with only three points can look straight and curved depending on the viewing window.

For regression work, use the trendline feature. Excel and Google Sheets both provide a linear trendline R² and an exponential trendline R². The higher R² is the better fit; values above 0.99 for one and below 0.95 for the other make the choice easy. When both are close, examine residuals — a linear model residuals showing a curve indicates the fit is secretly wrong. According to the National Council of Teachers of Mathematics, students who check residuals against the original table catch model-choice errors that a single R² misses.

Edge Cases: Vertical Tables, Negative Values, and the Origin Question

A table oriented vertically (x across the top, y down a column) tests the same way; just read the pairs correctly. Columns of x across the top usually pair with columns below, not with entries to the side. Misreading the layout is a frequent source of errors.

Negative y-values complicate the ratio test but not the difference test. A ratio of two negative numbers is positive (e.g., −8 / −4 = 2), so exponential functions can go negative as long as they don't hit zero. However, the standard exponential form y = a · bˣ with a > 0 and b > 0 is always positive for all real x, so a table crossing zero must be linear or piecewise. Linear functions cross zero without issue, as already noted.

The origin question — does the line pass through (0, 0)? — only matters when b = 0. Many applied contexts like direct proportionality (cost per item, distance at constant speed) are linear with b = 0, but linear does not require b = 0. Exponential functions in standard form have y-intercept a, not zero, so a table showing y = 0 at x = 0 cannot be exponential in the standard sense. Direct-variation tables look exponential because doubling x doubles y, as noted earlier; check whether y/x is constant to confirm linear-with-zero-intercept.

FAQ

What if only two rows of data are given?

Two rows prove nothing on their own because infinitely many functions pass through any two points. Check the tripling pattern with a third row: if y goes 4, 12, 36, the ratios are 3 and 3 (exponential), while if it goes 4, 12, 20, the differences are 8 and 8 (linear). A two-row table is genuinely ambiguous between linear and exponential, and the correct answer on a test is often a caveat: "insufficient information."

Do exponential functions ever have negative ratios?

No, the ratio between any two y-values of a standard exponential y = a · bˣ with b > 0 is positive across its domain. If your ratio column contains a negative value, the table is not a standard exponential — it's likely linear with a negative slope, or a piecewise combination. Consider checking whether the y-values pass through zero: any table where one entry is 0 cannot be exponential.

What's the difference between exponential growth and a quadratic table?

A quadratic y = x² has ratios that grow slowly (1, 4, 9, 16 involves ratios 4, 2.25, 1.78) while an exponential y = 2ˣ has ratios that stay fixed (2, 4, 8, 16 involves ratios 2, 2, 2). The decisive test is the ratio column: constant means exponential, slowly drifting means polynomial. Second differences of the quadratic are 2, 2, 2; second differences of the exponential double.

Can an exponential table have a constant first difference somewhere?

Yes, by coincidence. In the table x = 0, 1, 2 with y = 1, 3, 9, the differences are 2 and 6 (not constant), but if you sample only x = 0 and x = 1 and x = 2 of a carefully chosen exponential, one difference can match another difference from a different table. One matching value in a column doesn't establish a pattern; you need at least three consecutive matches to confirm.

How do half-life problems change the ratio when the x-step isn't the half-life?

If the half-life is 5 units and the table is spaced 1 unit apart, divide 0.5 into five equal multiplicative steps: 0.5^(1/5) ≈ 0.871. Your ratio column should be roughly 0.871 per row, not 0.5. Students who expect 0.5 per 1-step row misclassify the function as non-exponential when it's actually exponential with a scaled base.

Does a table have to include x = 0 to identify the function?

No, x = 0 isn't required. You need at least three rows to test the pattern, and the constant ratio or difference determines the base or slope regardless of where x starts. To find the parameter a or b (intercept), you can back-solve from any row: for y = a · bˣ, divide the y-value at x = k by bᵏ.

What if the ratio is very close to 1 but not exactly?

A ratio near 1 indicates very slow exponential growth (or decay) that can be indistinguishable from a constant function over a short table. If ratios are 1.001, 0.999, 1.002, 1.001, the deviation is within rounding and the correct answer is "approximately constant — consistent with either linear slope near zero or exponential base near 1." With longer tables the two models diverge: linear grows by a fixed amount per step, exponential grows by a fixed percentage.

Why do my spreadsheet ratios flip around when I divide backward?

Because division isn't symmetric: y₂/y₁ and y₁/y₂ are reciprocals, not negatives. Ratios computed in opposite directions give values like 2 and 0.5 that average to about 1. Make the direction consistent — always later y divided by earlier y — so the ratio column stays uniform when the function is exponential.