From Average Change to a Local Estimate
A secant slope describes the average rate of change of a function between two inputs. To investigate how a function is changing at one particular input, we can compare secant slopes over intervals that get progressively shorter and have that input as one endpoint.
If the slopes from the left and right settle near the same number as the intervals shrink, that number is a reasonable estimate of the function’s instantaneous rate of change at the specified input. A table can provide strong numerical evidence for an estimate, though a finite list of slopes alone does not prove an exact value.
Setting Up the Secant-Slope Table
Let \(f\) be a function and let \(a\) be the input of interest. For another input \(x\), the secant slope between \(a\) and \(x\) is the change in output divided by the change in input:
The condition \(x\ne a\) matters: when \(x=a\), the denominator is zero, so this calculation is not defined. Instead, use inputs close to \(a\), on both sides when possible. For a left-hand input \(x<a\), both the output difference and input difference may be negative; their quotient is still the slope of the secant line.
A useful table includes the nearby input, its function value, and the calculated slope. The intervals should shrink toward \(a\), for example from \(0.1\) away to \(0.01\) away. Compare the left-side and right-side slopes separately before deciding on an estimate.
Worked Example: A Quadratic Near an Input
Worked Example: A Quadratic Near an Input
Estimate the instantaneous rate of change of \(f(x)=x^2+1\) at \(x=2\) using secants from both sides.
First, evaluate the function at the specified input:
For a nearby input \(x\), calculate the secant slope \(\frac{f(x)-5}{x-2}\). For example, at \(x=1.9\), \(f(1.9)=1.9^2+1=4.61\), so the slope is \(\frac{4.61-5}{1.9-2}=\frac{-0.39}{-0.1}=3.9\). At \(x=2.1\), \(f(2.1)=2.1^2+1=5.41\), so the slope is \(\frac{5.41-5}{2.1-2}=\frac{0.41}{0.1}=4.1\).
| Nearby input \(x\) | \(f(x)=x^2+1\) | Secant slope \(\frac{f(x)-f(2)}{x-2}\) |
|---|---|---|
| 1.9 | 4.61 | \((4.61-5)/(1.9-2)=3.9\) |
| 1.99 | 4.9601 | \((4.9601-5)/(1.99-2)=3.99\) |
| 1.999 | 4.996001 | \((4.996001-5)/(1.999-2)=3.999\) |
| 2.001 | 5.004001 | \((5.004001-5)/(2.001-2)=4.001\) |
| 2.01 | 5.0401 | \((5.0401-5)/(2.01-2)=4.01\) |
| 2.1 | 5.41 | \((5.41-5)/(2.1-2)=4.1\) |
On the left, the slopes \(3.9,\ 3.99,\ 3.999\) approach \(4\) from below. On the right, \(4.001,\ 4.01,\ 4.1\) are close to \(4\), with the closest listed slope only \(0.001\) above it. The estimate is therefore \(4\).
Check the arithmetic in the closest rows: \((4.996001-5)/(1.999-2)=(-0.003999)/(-0.001)=3.999\), and \((5.004001-5)/(2.001-2)=0.004001/0.001=4.001\). The calculated slopes support the estimate from both sides.
Why Shrinking Secants Can Reveal a Stable Rate
A short algebraic calculation explains the pattern for this quadratic. For \(f(x)=x^2+1\) and a fixed input \(a\), the secant slope from \(a\) to \(x\), where \(x\ne a\), is
As \(x\) moves closer to \(a\), the expression \(x+a\) moves closer to \(2a\). This accounts for the observed pattern: the nearby secant slopes become close to one stable value. At \(a=2\), that value is \(2a=4\), matching the table.
This calculation gives a complete justification for the pattern in this particular example. More generally, when estimating from a numerical table, the slopes themselves may suggest a value without establishing it exactly. Whether the actual instantaneous rate exists and what its exact value is can require further mathematical justification.
Call it \(a\), and calculate \(f(a)\).
Use some inputs below \(a\) and some above \(a\), with distances that shrink toward zero.
Subtract the function values and divide by the corresponding difference in inputs.
Look for left- and right-side slopes that settle near the same number, and report that number as an estimate.
Worked Example: Estimating a Rate from Measured Data
Worked Example: Estimating a Rate from Measured Data
A position sensor records a moving object’s position \(s\), in meters, at times \(t\), in seconds. Use the data to estimate its instantaneous velocity at \(t=3\) seconds.
| Time \(t\) (s) | Position \(s\) (m) |
|---|---|
| 2.8 | 10.35 |
| 2.9 | 10.87 |
| 3.0 | 11.40 |
| 3.1 | 11.94 |
| 3.2 | 12.49 |
Use \(s(3)=11.40\) meters as the fixed position value. On the left, the slope from \(2.9\) to \(3.0\) is \(\frac{11.40-10.87}{3.0-2.9}=\frac{0.53}{0.1}=5.3\) meters per second. Over the wider left interval, the slope is \(\frac{11.40-10.35}{3.0-2.8}=\frac{1.05}{0.2}=5.25\) meters per second.
On the right, the slope from \(3.0\) to \(3.1\) is \(\frac{11.94-11.40}{3.1-3.0}=\frac{0.54}{0.1}=5.4\) meters per second. Over the wider right interval, it is \(\frac{12.49-11.40}{3.2-3.0}=\frac{1.09}{0.2}=5.45\) meters per second.
The closer left and right slopes, \(5.3\) and \(5.4\), bracket an estimate near their midpoint: \(\frac{5.3+5.4}{2}=\frac{10.7}{2}=5.35\) meters per second. The wider-interval slopes, \(5.25\) and \(5.45\), are also centered at \(5.35\). The data therefore support an estimated instantaneous velocity of \(5.35\) meters per second. Because these are measured values, the estimate should not be treated as an exact rate.
Check the calculations: the closer intervals each have length \(0.1\) second, and their position changes are \(0.53\) meter and \(0.54\) meter. Dividing gives \(5.3\) and \(5.4\) meters per second. The wider intervals each have length \(0.2\) second, and changes of \(1.05\) and \(1.09\) meters give \(5.25\) and \(5.45\) meters per second.
Worked Example: A Reciprocal Function
Worked Example: A Reciprocal Function
Estimate the instantaneous rate of change of \(f(x)=\frac{1}{x}\) at \(x=2\) using nearby inputs.
The fixed function value is \(f(2)=\frac{1}{2}=0.5\). For each nearby input, calculate the secant slope \(\frac{f(x)-0.5}{x-2}\).
| Input \(x\) | Function value \(f(x)\) | Secant slope |
|---|---|---|
| 1.9 | \(1/1.9\approx0.526316\) | \((0.526316-0.5)/(1.9-2)\approx-0.26316\) |
| 1.99 | \(1/1.99\approx0.502513\) | \((0.502513-0.5)/(1.99-2)\approx-0.25126\) |
| 2.01 | \(1/2.01\approx0.497512\) | \((0.497512-0.5)/(2.01-2)\approx-0.24876\) |
| 2.1 | \(1/2.1\approx0.476190\) | \((0.476190-0.5)/(2.1-2)\approx-0.23810\) |
The left-side slopes are below \(-0.25\), and the right-side slopes are above \(-0.25\). The closest listed slopes are approximately \(-0.25126\) and \(-0.24876\), so a reasonable estimate is \(-0.25\).
An exact simplification checks the table’s pattern for any \(x\ne2\):
At \(x=1.99\), this gives \(-1/3.98\approx-0.25126\); at \(x=2.01\), it gives \(-1/4.02\approx-0.24876\), agreeing with the table. As \(x\) gets closer to \(2\), \(-1/(2x)\) gets closer to \(-1/4=-0.25\). This confirms the estimate for this function.
Interpreting Estimates Carefully
Secant slopes have units of output per unit of input. If \(f(x)\) measures distance in meters and \(x\) measures time in seconds, slopes are measured in meters per second. If \(f(x)\) measures temperature in degrees and \(x\) measures position in centimeters, the slopes have units of degrees per centimeter. The estimated instantaneous rate uses the same units.
Several habits make these estimates more reliable:
- Use the same specified input as one endpoint in every secant.
- Compare slopes from the left and right, rather than relying on just one side.
- Use increasingly close inputs, but retain enough digits in intermediate calculations to see the pattern.
- Do not claim more precision than the table supports, especially when values are measured or rounded.
A Common Mistake: Mixing Up the Endpoints
Suppose \(f(x)=x^2+1\), and the specified input is \(2\). A common error is to use the slope between \(1.9\) and \(2.1\) as though it were one of the shrinking secants from \(2\). That slope is \(\frac{f(2.1)-f(1.9)}{2.1-1.9}=\frac{5.41-4.61}{0.2}=4\). Although the arithmetic is correct, neither endpoint is fixed at \(2\) alone in the intended comparison: this is a secant across the target, not a secant from the target to one nearby input. It cannot show how the left and right slopes behave separately.
Instead, calculate the two slopes separately: from \(1.9\) to \(2\), \(\frac{5-4.61}{2-1.9}=\frac{0.39}{0.1}=3.9\); from \(2\) to \(2.1\), \(\frac{5.41-5}{2.1-2}=\frac{0.41}{0.1}=4.1\). These values reveal the two-sided pattern around the specified input.
What a Table Can Establish
A table is useful for estimating, but its conclusion has limits. A handful of slopes near \(4\), for example, does not by itself prove that the instantaneous rate is exactly \(4\); values farther from the input could behave differently, and rounded data can conceal variation. When a formula is available, algebra may justify the pattern, as in the quadratic and reciprocal examples. When only measurements are available, report a reasonable estimate and its units, while recognizing the uncertainty.
Check Your Understanding
Use the secant-slope idea to answer each question. Show the calculation when a numerical slope is requested.
- For \(f(x)=x^2\) at \(x=3\), calculate the secant slopes from \(3\) to \(2.9\) and from \(3\) to \(3.1\). What value do these slopes suggest?
- Why should a table include nearby inputs on both sides of the specified input when possible?
- If position is measured in kilometers and time in hours, what are the units of the secant slopes?
- Explain why a finite table of nearby slopes can support an estimate without proving that the exact instantaneous rate equals that estimate.
- For \(f(x)=1/x\) at \(x=2\), use the identity \(-1/(2x)\) for the secant slope to find the exact slope when \(x=2.1\).
