Linear Approximation: Estimating Function Values Near a Point

On this page: The Principle | Conditions | Misconceptions | EE Questions | Retrieval Practice | Worked Example | Solve a Problem | FAQ

---

The Principle

Statement

The linear approximation of $f$ at $a$ estimates $f(x)$ for inputs $x$ close to $a$ by replacing the curve with its tangent line at that point: $f(x) \approx f(a) + f'(a)(x - a)$. The tangent line agrees with $f$ in both output value and slope at $x = a$, so for small displacements $x - a$ the curve hugs the line closely. Choosing a nearby base point $a$ where $f(a)$ and $f'(a)$ are easy to compute exactly turns a hard evaluation into straightforward arithmetic.

Mathematical Form

$f(x) \approx f(a) + f'(a)(x - a)$

Where:

• $x$ = the input at which the approximation is evaluated
• $a$ = the base point where the derivative is known (or easy to compute)
• $f(a)$ = the exact function value at the base point
• $f'(a)$ = the derivative at the base point, supplying the slope of the tangent line
• $f(x)$ = the estimated function value near $a$

Alternative Form

In some texts the same formula appears as the linearization $L(x)$:

• Linearization notation: $L(x) = f(a) + f'(a)(x - a)$, with $f(x) \approx L(x)$

The symbols differ; the content is identical.

---

Conditions of Applicability

Condition: x near a; f’(a) exists

Practical modeling notes

• “x near a” has no universal threshold. Accuracy degrades as $|x - a|$ grows, especially when $|f''|$ is large near $a$. A rough local-error scale is $\tfrac{1}{2}|f''(a)|(x-a)^2$ (heuristic only; a formal bound requires $\max|f''|$ on an interval, not just at $a$).
• You choose $a$: pick a value where $f(a)$ and $f'(a)$ are easy to compute exactly. Useful choices include perfect squares for $\sqrt{x}$, multiples of $\pi/6$ or $\pi/4$ for trig functions, and $a = 0$ for $e^x$ or $\ln(1+x)$.

When It Doesn’t Apply

• Non-differentiable point: If $f'(a)$ does not exist (corner, cusp, vertical tangent, or discontinuity at $a$), there is no slope to use. The tangent line equation likewise fails at these same points.
• x far from a: For large displacements, the tangent line diverges from the curve. Use a higher-order Taylor polynomial when a closer estimate is required.

Want the complete framework behind this guide? Read Masterful Learning.

---

Common Misconceptions

Misconception 1: The approximation gives the exact value

The truth: $f(x) \approx f(a) + f'(a)(x - a)$ is exact only at $x = a$ itself, where both sides equal $f(a)$. For inputs other than $a$, the expression is generally an estimate rather than the exact value.

Why this matters: Treating the approximation as exact suppresses the error term. In applied contexts, reporting the approximated value without acknowledging that it is an estimate is a precision error.

Misconception 2: Any base point $a$ works equally well

The truth: Accuracy depends jointly on how small $|x - a|$ is and on how curved $f$ is near $a$ (controlled by $|f''|$). A poorly chosen base point that is far from $x$, or one where $f$ bends sharply, produces larger errors even for modestly small displacements.

Why this matters: Part of the skill is selecting a convenient $a$ that balances ease of computation against the required approximation quality.

---

Elaborative Encoding

Use these questions to build deep understanding. (See Elaborative Encoding for the full method.)

Within the Principle

• In the formula $f(x) \approx f(a) + f'(a)(x - a)$, which term anchors the approximation to the curve and which term accounts for the slope-driven change from $a$ to $x$?
• If $f'(a) = 0$, what does the linear approximation reduce to, and what does that reveal about the shape of $f$ immediately near $a$?

For the Principle

• How do you select the base point $a$? What properties make one choice of $a$ more useful than another for a given target input $x$?
• What happens to the accuracy of the approximation as $x$ moves progressively farther from $a$, and what measure captures how fast the error grows?

Between Principles

• The tangent line equation and the linear approximation share the same algebraic formula. In what way do they frame the same mathematical object from different angles?

Generate an Example

• Name a specific function and base point where the linear approximation replaces a calculation that would otherwise require a calculator. Describe the approximation step explicitly.

---

Retrieval Practice

Answer from memory, then click to reveal and check. (See Retrieval Practice for the full method.)

State the principle in words: _____Near a, f(x) is approximated by its tangent line: f(x) is approximately f(a) + f'(a)(x-a).

Write the canonical equation: _____$f(x)\approx f(a)+f'(a)(x-a)$

State the canonical condition: _____x near a; f'(a) exists

---

Worked Example

Use this worked example to practice Self-Explanation.

Problem

Approximate $\sqrt{4.1}$ using the linear approximation of $f(x) = \sqrt{x}$ at $a = 4$.

Step 1: Verbal Decoding

Target: $f(4.1) = \sqrt{4.1}$
Given: $f$, $a$, $x$
Constraints: $x = 4.1$ is near $a = 4$; $f'(a)$ exists for all $x > 0$

Step 2: Visual Decoding

Draw $f(x) = \sqrt{x}$ near $x = 4$ and mark the base point $(4, 2)$. Draw the tangent line at $a = 4$; it rises with slope $1/4$. Locate $x = 4.1$ slightly right of $a$. (Because $\sqrt{x}$ is concave down near $a$, the tangent line lies above the curve, so the approximation slightly overestimates.)

Step 3: Mathematical Modeling

1. $f(4.1) \approx f(4) + f'(4)(4.1 - 4)$

Step 4: Mathematical Procedures

1. $f(4) = \sqrt{4}$
2. $f(4) = 2$
3. $f'(x) = \frac{1}{2\sqrt{x}}$
4. $f'(4) = \frac{1}{2\sqrt{4}}$
5. $f'(4) = \frac{1}{4}$
6. $f(4.1) \approx 2 + \frac{1}{4}(0.1)$
7. $\underline{f(4.1) \approx 2.025}$

Step 5: Reflection

• Verification: At $x = a = 4$: $L(4) = 2 + \tfrac{1}{4}(0) = 2 = f(4)$ — the approximation is exact at the base point, as required.
• Magnitude: The actual value $\sqrt{4.1} \approx 2.0248$; the error $|2.025 - 2.0248| \approx 0.0002$, which is small relative to the answer.
• Graphical meaning: $f''(x) = -\tfrac{1}{4}x^{-3/2} < 0$ near $a = 4$, so the curve is concave down and the tangent line overestimates, consistent with $2.025 > 2.0248$.

---

Before moving on: self-explain the model

Try explaining Step 3 out loud (or in writing): why the tangent line at $a = 4$ applies here, what each term of the approximation formula represents, and why $a = 4$ is the natural choice for estimating $\sqrt{4.1}$.

Mathematical model with explanation

Principle: Linear approximation — $f(x) \approx f(a) + f'(a)(x-a)$, valid when $f'(a)$ exists and $x$ is near $a$.

Conditions: $x = 4.1$ is $0.1$ away from $a = 4$, a small displacement. $f'(4) = 1/4$ is a finite real number, so both conditions are satisfied.

Relevance: $\sqrt{4.1}$ has no closed exact form. The nearby perfect square $a = 4$ makes both $f(4) = 2$ and $f'(4) = 1/4$ trivially computable, so the problem reduces to arithmetic.

Description: The base value $f(4) = 2$ anchors the approximation. The slope $f'(4) = 1/4$ multiplied by the displacement $0.1$ adds the first-order correction $0.025$, yielding the estimate $2.025$.

Goal: Estimate $\sqrt{4.1}$ without a calculator by leveraging known values at the base point $a = 4$.

---

Solve a Problem

Apply what you’ve learned with Problem Solving.

Problem

Approximate $e^{0.1}$ using the linear approximation of $f(x) = e^x$ at $a = 0$.

Hint (if needed): Recall that $e^0 = 1$ and that the derivative of $e^x$ is $e^x$ itself.

Show Solution

Step 1: Verbal Decoding

Target: $f(0.1) = e^{0.1}$
Given: $f$, $a$, $x$
Constraints: $x = 0.1$ is near $a = 0$; $f'(0)$ exists for all $x$

Step 2: Visual Decoding

Draw $f(x) = e^x$ near $x = 0$ and mark the base point $(0, 1)$. Draw the tangent line at $a = 0$; it has slope $1$ and passes through $(0, 1)$. Locate $x = 0.1$ slightly right of $a$. (Because $e^x$ is concave up, the tangent line lies below the curve, so the approximation underestimates.)

Step 3: Mathematical Modeling

1. $e^{0.1} \approx f(0) + f'(0)(0.1 - 0)$

Step 4: Mathematical Procedures

1. $f(0) = e^0$
2. $f(0) = 1$
3. $f'(x) = e^x$
4. $f'(0) = e^0$
5. $f'(0) = 1$
6. $e^{0.1} \approx 1 + 1 \cdot (0.1)$
7. $\underline{e^{0.1} \approx 1.1}$

Step 5: Reflection

• Verification: At $x = 0$: $L(0) = 1 + 1 \cdot 0 = 1 = e^0$ — exact agreement at the base point confirmed.
• Magnitude: Actual $e^{0.1} \approx 1.10517$; the error $\approx 0.005$, about $0.5\%$ of the result, consistent with a small displacement.
• Graphical meaning: $f''(x) = e^x > 0$, so $e^x$ is concave up and the tangent line underestimates: $1.1 < e^{0.1}$.

---

Related Principles

Principle	Relationship to Linear Approximation
Tangent line equation	Linear approximation IS the tangent line $y = f(a) + f'(a)(x-a)$ used as an estimate of $f(x)$
Derivative at a point	$f'(a)$ provides the slope; without it there is no approximation
Taylor polynomial definition	Linear approximation is the $n = 1$ Taylor polynomial at $a$; Taylor extends it to add quadratic and higher correction terms

---

FAQ

What is linear approximation in calculus?

Linear approximation (also called linearization) uses the tangent line at a base point $a$ to estimate nearby function values: $f(x) \approx f(a) + f'(a)(x-a)$. It works because the tangent line matches $f$‘s value and slope at $a$, so the curve barely departs from the line for inputs close to $a$.

When should I use linear approximation?

Use it when you need an estimate of $f(x)$ for an input $x$ close to a base point $a$ where $f(a)$ and $f'(a)$ are easy to compute exactly. Common choices: $a$ is a perfect square for $\sqrt{x}$, a multiple of $\pi/6$ or $\pi/4$ for trig functions, or $a = 0$ for $e^x$ and $\ln(1+x)$.

What’s the difference between linear approximation and the tangent line?

They are the same formula. The tangent line $y = f(a) + f'(a)(x-a)$ is a geometric object—a line through the curve at $a$. Linear approximation uses the same expression as an estimation tool: $f(x) \approx L(x)$ where $L(x) = f(a) + f'(a)(x-a)$.

How accurate is a linear approximation?

Accuracy depends on how fast the curve bends. A rough local-error scale is $\tfrac{1}{2}|f''(a)|(x-a)^2$—use this as a heuristic, not a guaranteed bound (a formal bound requires $|f''|$ bounded on an interval). The smaller $|x - a|$ and $|f''|$ near $a$, the better the estimate. For higher precision at larger displacements, use a second-order Taylor polynomial instead.

How does linear approximation relate to the Taylor polynomial?

The linear approximation $f(a) + f'(a)(x-a)$ is identical to the degree-$1$ Taylor polynomial centered at $a$. Taylor polynomials extend the idea by adding terms proportional to $f''(a)(x-a)^2$, $f'''(a)(x-a)^3$, and so on, improving accuracy farther from $a$ at the cost of more computation.

---

Related Guides

• Calculus Subdomain Map — Return to the calculus hub to see where approximation sits after tangent-line work and before higher-order Taylor methods
• Principle Structures — Organize linear approximation in its hierarchical context: derivative at a point → tangent line → linear approximation → Taylor polynomial
• Self-Explanation — Learn to narrate worked solutions step by step
• Retrieval Practice — Keep this formula immediately accessible
• Problem Solving — Apply the Five-Step Strategy to approximation problems

---

How This Fits in Unisium

Within the calculus subdomain, Unisium treats linear approximation as the first local-estimation use of the tangent line equation: the same line that matches value and slope at $a$ becomes a nearby-value model for $f(x)$. The platform links the approximation to the elaborative encoding questions, retrieval cloze prompts, and worked problems in this guide, then tracks which exercises each student has completed. When you attempt a problem tagged linearApproximation, the platform schedules follow-up retrieval at spaced intervals so the formula remains accessible under timed exam conditions.

Ready to master linear approximation? Start practicing with Unisium or explore the complete learning framework in Masterful Learning.

Masterful Learning

The study system for physics, math, & programming that works: retrieval, connection, explanation, problem solving, and more.

Read the book (opens in new tab) ISBN 979-8-2652-9642-9