Partial Derivatives: Measuring Change One Input at a Time

How to calculate partial derivatives, interpret their values, and avoid confusing local rates with finite changes.

A partial derivative measures how a function changes with respect to one input while its other inputs are held constant. This turns a multivariable problem into ordinary single-variable differentiation. The essential skill is not learning a new set of differentiation rules, but identifying which variable changes and which quantities stay fixed. [1]

In this article

1. The central idea: choose what changes
2. Read the notation and apply the rules
3. An original worked example
4. A local rate is not a finite-change guarantee
5. What partial derivatives tell you in practice

1. The central idea: choose what changes

Isolate one input: a partial derivative answers a controlled question—how does the output change when this input varies and the others remain fixed? [1]
Use familiar rules: differentiate with the usual single-variable methods, treating every other input as a constant. [1]
Interpret the result locally: the value at a particular point describes a rate there, not necessarily an exact output change over a whole unit of input. [1]

For a single-variable function, f(x), the derivative describes its rate of change as x changes. For a function such as f(x,y), that question needs another detail: are we changing x, changing y, or changing both? Partial differentiation selects one of these possibilities by freezing all but one input. For example, fixing y = b produces the single-variable function g(x) = f(x,b). Differentiating g tells us how f responds to x under that fixed-y condition. This is why a multivariable function has separate partial derivatives rather than one unspecified derivative. [1]

2. Read the notation and apply the rules

The subscript in fₓ(x,y) identifies x as the variable being differentiated; fᵧ(x,y) identifies y. The distinction matters because these generally give different answers. Also distinguish a derivative function, such as fₓ(x,y), from its value at a point, such as fₓ(1,2). The first gives a formula that can be evaluated at different inputs; the second gives the rate at one specified input pair. Keeping the variable in the notation avoids the ambiguity of writing only a prime. [1]

Lamar University's calculus notes illustrate the calculation with f(x,y) = 2x²y³. When differentiating with respect to x, the factor 2y³ is constant, so fₓ(x,y) = 4xy³. When differentiating with respect to y, the factor 2x² is constant, so fᵧ(x,y) = 6x²y². Notice that the variable held fixed can still appear in the answer. Holding y constant during differentiation does not mean deleting y; it means treating its value as a fixed coefficient while x varies. [1]

3. An original worked example

Consider the hypothetical mathematical model F(x,y) = 3x²y + y². To find Fₓ, hold y constant. The derivative of 3x²y with respect to x is 6xy, while y² contributes zero because it does not vary with x. Thus Fₓ(x,y) = 6xy. To find Fᵧ, hold x constant instead: 3x²y contributes 3x² and y² contributes 2y. Therefore Fᵧ(x,y) = 3x² + 2y. These results follow directly from applying ordinary differentiation separately to each input. [1]

At (1,2), the output is F(1,2) = 10, while the two partial derivatives are Fₓ(1,2) = 12 and Fᵧ(1,2) = 7. The output value and the derivative values answer different questions: 10 is the function's value, whereas 12 and 7 describe its local responses to separate input changes. We can check the meaning by freezing one input explicitly. Setting y = 2 gives F(x,2) = 6x² + 4, whose derivative at x = 1 is 12. Setting x = 1 gives F(1,y) = 3y + y², whose derivative at y = 2 is 7. [1]

The following comparison collects the two calculations. Each row represents a different fixed-input question about the same hypothetical function. [1]

Question	Held constant	Derivative formula	Value at (1,2)
How does F change with x?	y	Fₓ = 6xy	12 [1]
How does F change with y?	x	Fᵧ = 3x² + 2y	7 [1]

4. A local rate is not a finite-change guarantee

Does Fₓ(1,2) = 12 mean increasing x by one increases F by exactly 12? No. In this example, with y fixed at 2, increasing x from 1 to 2 changes the output from 10 to 28—an increase of 18. The partial derivative describes the rate at the starting point, and that rate itself changes with x: Fₓ(x,2) = 12x. A derivative is therefore not automatically an exact rule for a large finite step. [1]

The same example makes the local interpretation precise. For an increment h in x, with y = 2, direct algebra gives F(1+h,2) − F(1,2) = 12h + 6h². For nonzero h, the change per unit of input is therefore 12 + 6h, which approaches 12 as h approaches zero. For h = 0.01, the exact increase is 0.1206, close to the first-order estimate 12 × 0.01 = 0.12. This calculation explains both the usefulness of the derivative and why a finite-step prediction can have a remainder. [1]

5. What partial derivatives tell you in practice

A practical application is separating influences in a model. Wikipedia gives a consumption function depending on income and wealth: its partial derivative with respect to income represents the marginal propensity to consume. The fixed-input interpretation is crucial here: the calculation describes the model's response to income while wealth is held constant. It does not describe a change in which income and wealth both vary. Before interpreting any partial derivative, identify the output, the changing input, and the quantities held fixed. [3], [1]

Likewise, the two derivative values in our example do not by themselves answer what happens when x and y change together. They answer two separate, controlled questions. For a simultaneous change, you must specify how both inputs change; otherwise the question remains incomplete. This distinction is the most useful habit to carry forward: always attach a partial derivative to its fixed-variable condition, rather than treating it as an unrestricted description of every possible change in the function. [1]

Frequently asked questions

Why can y remain in the answer when I differentiate with respect to x?

Because constant does not mean zero. In 2x²y³, the factor y³ acts as a fixed multiplier during differentiation with respect to x. The result is 4xy³, so the rate with respect to x can differ depending on which fixed value of y you choose. [1]

When does a term disappear during partial differentiation?

A term disappears when it is entirely constant with respect to the chosen input. In the hypothetical function F(x,y) = 3x²y + y², differentiating with respect to x makes the derivative of y² zero. But 3x²y does not disappear, because it still varies with x. [1]

Should I substitute the point before or after differentiating?

A reliable approach is to differentiate first and then substitute both coordinates. Alternatively, fix only the other input first, then differentiate the resulting single-variable function. Do not substitute both coordinates before differentiating: that leaves only a number and removes the variable whose rate you wanted. [1]

Does a positive partial derivative describe every possible change?

No. It describes a positive local rate with respect to the selected input under the stated fixed-input condition. If other inputs change too, that is a different question. The way those inputs vary must be specified before interpreting the resulting output change. [1]

Can partial derivatives be differentiated again?

Yes. Just as single-variable functions can have derivatives of higher orders, multivariable functions can have higher-order derivatives. The important notation habit remains the same: keep track of which variable is being differentiated at each stage. [1]

Conclusion

Partial derivatives make multivariable change manageable by asking one precise question at a time. Choose the changing input, hold the others fixed, apply ordinary differentiation rules, and then evaluate at the point of interest. In the worked example, Fₓ(1,2) = 12 and Fᵧ(1,2) = 7 describe different local responses—not output values or guaranteed finite changes. Understanding that distinction is more important than memorizing a formula: it connects the calculation to what the derivative actually measures. [1]

Explore next

Sources

Open a source or expand its cited passages.

Relevance scores are model estimates of how well each source addresses this question, not probabilities or fact-check ratings. Reasoned guidance is identified on hover.