Two of Unit 8’s applications look unrelated and are the same calculation. The average value of a function is the constant that would enclose the same area; the area between two curves is the integral of their difference. In both cases the integrand is a difference, and in both cases the mistakes come from getting that difference wrong.

Both also rest on the same enduring understanding, that definite integrals let us solve problems involving the accumulation of change over an interval.

Average value is a height, not a rate

The framework’s definition is a formula:

\[f_{\text{avg}} = \frac{1}{b-a}\int_a^b f(x)\,dx.\]

Read it as area divided by width, which makes it the height of the rectangle on \([a,b]\) with the same area as the region under \(f\). It is measured in the units of \(f\), not in units of \(f\) per unit of \(x\).

The framework’s exam guidance names the confusion directly: some students confuse the average value and the average rate of change of a function on an interval. Those are different quantities with different units, and the relationship between them is exact and worth knowing.

\[\text{average rate of change of } f = \frac{f(b)-f(a)}{b-a} = \frac{1}{b-a}\int_a^b f'(x)\,dx = \big(f'\big)_{\text{avg}}.\]

So the average rate of change of \(f\) is the average value of \(f'\) — one level down, never the average value of \(f\) itself. On \(f(x) = x^2\) over \([1,3]\) the average rate of change is 4 and the average value is \(\tfrac{13}{3}\), and no rearrangement turns one into the other.

That middle expression should also look familiar: \(\tfrac{f(b)-f(a)}{b-a}\) is the quantity the Mean Value Theorem promises the derivative attains somewhere. Average rate of change is the derivative’s business; average value is the function’s own.

Why it is not the midpoint of the range

The exam guidance goes further and asks that students understand why an average value may be less than, equal to, or greater than the midpoint of the range. All three happen on \([0,1]\) with functions running from 0 to 1.

For \(f(x) = x\) the average value is \(\tfrac12\), and the midpoint of the range is also \(\tfrac12\). For \(f(x) = x^2\) the average value is \(\tfrac13\), below the midpoint, because the function spends most of the interval near its low values. For \(f(x) = \sqrt{x}\) it is \(\tfrac23\), above, for the mirror-image reason.

The midpoint of the range asks where the outputs sit. The average value asks how long the function spends near each of them, and those are not the same question.

Top minus bottom

Five regions. Dark shading is where the first named boundary is ahead of the second and the difference counts positively; pale shading is where they have swapped and it counts negatively. The panel gives both the signed integral and the total area, which are the same number only when the boundaries never cross in between. The first two presets are average value in disguise: the constant is chosen so the signed integral is exactly zero, which is what makes it the average, and the two shaded pieces then have equal area.

The integrand for an area is always the difference of the two boundaries, taken in the order that keeps it positive. The framework attaches its notation skill to this topic — use appropriate mathematical symbols and notation — and the notation is where the point is earned: writing \(\textstyle\int_0^1 (x - x^2)\,dx\) says which curve is on top, and writing \(\textstyle\int_0^1 x\,dx - \textstyle\int_0^1 x^2\,dx\) says it too, but only the first survives contact with a region where they trade places.

For \(y = x\) and \(y = x^2\) on \([0,1]\) the area is \(\tfrac12 - \tfrac13 = \tfrac16\), and both forms give it, because \(x \geq x^2\) throughout.

Functions of y, and curves that cross three times

Two later topics complicate the setup without changing the idea.

The first is orientation. The framework says areas can be calculated using functions of either \(x\) or \(y\), and the region bounded by \(x = y^2\) and \(x = y+2\) is far easier in \(y\): the boundaries meet where \(y^2 = y+2\), so \(y = -1\) and \(y = 2\), and

\[\int_{-1}^{2}\big((y+2) - y^2\big)\,dy = \frac92.\]

In \(x\) the same region would need splitting, because its left boundary changes character partway up. Choosing the variable is part of the setup, not a preference.

The second is curves that cross more than twice, which is a topic of its own. Take \(y = x\) and \(y = x^3\) on \([-1,1]\). They meet at \(-1\), \(0\), and \(1\), and \(x^3\) is above on the left half and below on the right. The signed integral is

\[\int_{-1}^{1}(x - x^3)\,dx = 0,\]

which is correct and is not the area. The area is \(\tfrac12\), found by splitting at the interior crossing and adding the two pieces as positive quantities. A total area that comes out zero for a region you can see is the same warning as a negative answer for a positive integrand: the integral did what it was told, and it was told the wrong thing.

Both halves of this article fail in the same place, which is why they belong together. An average value is the signed integral divided by the width, so a function that spends half its time negative can have an average value near zero without being near zero anywhere. An area is the unsigned accumulation, so it needs the crossings found first. Before integrating a difference, ask one question: am I allowed to let this cancel? For average value the answer is yes and that is the definition. For area it is no, every time.