ERF in Apple Numbers: Gaussian Error Integrals and Bounds
Learn Apple Numbers ERF syntax, optional upper bound, normal-curve examples, errors and safe numerical checks.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
Quick answer. Apple’s ERF(lower, upper) returns the Gaussian error-function integral between bounds. =ERF(0,1) has Apple’s published result 0.842700792949715. The second argument is optional: =ERF(1) is mathematically equivalent to =ERF(0,1). The optional-argument rule is documented by Apple; the equivalence is the mathematical definition, not a screenshot of execution in Numbers. [A1, C1]
Availability and version history
- Introduced: Documented in Apple’s Formulas and Functions Help 15.4; the original introduction release is not established.
- Availability: Currently documented for Numbers in that help version; it is not a final supported-version endpoint.
- Scope: Shared iWork function help. Availability in individual clients and older releases is not established by this check.
- Checked: October 14, 2026, against the official ERF reference. Documentation check; not a new native test.
What ERF calculates and why it is useful
For real numbers a and b, the Gaussian error-function integral is
ERF(a,b) = (2 / SQRT(PI)) Γ integral from a to b of EXP(-tΒ²) dt
= erf(b) β erf(a)
If only a is supplied, the integral starts at zero and ends at a. Because the integrand is positive, reversing mathematical bounds changes the sign. This last statement is a mathematical property; Apple does not specify in its article how Numbers handles every reversed-bound input, so native verification is still appropriate before documenting an application-dependent edge case. [A1, C1]
The error function arises in normal-distribution mathematics, diffusion approximations and measurement noise. It is not the same thing as the probability of an error in your spreadsheet; the name refers to a particular special function. In particular, erf(x) ranges mathematically from β1 to 1 for finite real x, while a two-bound integral such as ERF(-1,1) can exceed 1 because it spans both sides of zero. Don’t label its raw result a probability without a valid transformation. [C1]
Exact Apple Numbers syntax
=ERF(lower, upper)
=ERF(lower)
| Position | Apple argument | Required? | Apple-documented type | Meaning | Omitted/default behavior |
|---|---|---|---|---|---|
| 1 | lower |
Yes | Number | Lower integration limit in the two-argument form; the end point when one argument is used | None |
| 2 | upper |
No | Number | Upper integration limit | Integration starts at 0 and ends at lower |
Do not enter an empty comma as a substitute for omitted upper; simply use the one-argument form. Both examples use Apple’s English-language argument order, not another spreadsheet application’s default syntax. Apple’s help does not state a separate special default value for lower or prescribe the display’s decimal precision. [A1, A2]
Reproducible Numbers worksheet β compare one and two bounds
Create a Numbers table named ErrorBounds. In row 1, put Lower in B1, Upper in C1, Two-bound integral in D1, and One-bound erf in E1. Type the following numbers:
| Row | B (Lower) | C (Upper) | D formula | Expected D |
|---|---|---|---|---|
| 2 | 0 | 1 | =ERF(B2,C2) |
0.842700792950 |
| 3 | β1 | 1 | =ERF(B3,C3) |
1.685401585899 |
| 4 | 1 | 8 | =ERF(B4,C4) |
0.157299207050 |
| 5 | 0.5 | 1.5 | =ERF(B5,C5) |
0.445605268662 |
In E2 enter =ERF(C2) and copy it to E3:E5. The expected E results are 0.842700792950, 0.842700792950, approximately 1 (the single-argument ERF(8) is extremely close to 1), and 0.966105146475 respectively. E3 does not equal D3: the E3 formula runs from zero to 1, whereas D3 runs from β1 to 1. Rows 2β4 include values Apple itself prints for the two-argument form; row 5 and E-column comparisons are independent calculations. [A1, C1]
For an isolated quick check use:
=ERF(0,1)
=ERF(-1,1)
=ERF(1,8)
=ERF(0.5)
The first three reproduce Apple’s documented examples. The fourth is expected to be 0.520499877813. If your cell displays fewer digits, change the number format/decimal places; visual rounding alone does not prove a different underlying value. [A1, C1]
Advanced use 1 β area under the standard normal curve
For a standard normal value Z, the probability between β1 and +1 can be written mathematically as
=ERF(-1/SQRT(2),1/SQRT(2))/2
Expected answer 0.682689492137 (about 68.27%). The factor 1/2 and scaling by SQRT(2) matter: using =ERF(-1,1) directly would yield ~1.6854 and therefore could not be a valid probability. This is a transformation of the Gaussian integral, not an Apple claim that ERF() automatically knows which distribution or units your data use. [C1, A1]
For any real standardized limits z1 β€ z2, the general probability expression is
=(ERF(z2/SQRT(2))-ERF(z1/SQRT(2)))/2
If z1 is in B2 and z2 in C2, the sheet formula is:
=(ERF(C2/SQRT(2))-ERF(B2/SQRT(2)))/2
Alternatively, because Numbers supports two bound arguments to ERF, =ERF(B2/SQRT(2),C2/SQRT(2))/2 gives the mathematically same expected result. Choosing the two-bound form removes one subtraction from the formula. [A1, C1]
Advanced use 2 β diffusion-style dimensionless comparison
A normalized diffusion model sometimes uses erf(x/(2*SQRT(D*t))), where D > 0 is a model parameter, t > 0 is elapsed time and x is a distance. If B2 holds x, C2 holds D, and D2 holds t, an illustrative mathematical model is:
=ERF(B2/(2*SQRT(C2*D2)))
For x=2, D=1, and t=1, the expected value is ERF(1) = 0.842700792950. This is not a universal physical boundary condition: real diffusion solutions require units, geometry and assumptions matched to the specific problem. Use the formula as a numerical building block, not as a validated engineering prediction. [C1]
Advanced use 3 β cross-check with ERFC
Mathematically ERFC(x)=1βERF(x), so a quick comparison at x=1 is
=1-ERF(1)
=ERFC(1)
Both are expected near 0.157299207050. The direct ERFC() form can be preferable when calculating very small upper-tail values because 1βERF(x) subtracts nearly equal quantities and may lose precision. ERFC is a separate Apple-documented function with a complementary result. [A1, A6, C1]
Errors, limits and troubleshooting
| Symptom | What to check | Evidence boundary |
|---|---|---|
| Formula editor rejects the formula | Start with =; spell ERF correctly; use one or two numeric arguments in documented order. |
Argument count/types are Apple documented. [A1] |
| A decimal-comma locale shows a syntax error | Use Numbers’ regional argument separator (often ; instead of ,) and decimal commas if required. |
Locale syntax is in Apple’s functions overview. [A2] |
| Result is greater than 1 | You may have integrated from a negative to a positive bound. That is not by itself an ERF error; the two-bound integral can be >1. | Mathematical definition, not a Numbers run. [C1] |
| Unexpected sign | Check whether your bounds or the one-bound input are negative. Verify the direction of the intended integral. | Math property; unusual execution cases not native-tested. [C1] |
| Only a few digits appear | Set the cell’s number display to show more decimal places. | Display settings do not change the mathematical definition. |
| A cross-table cell reference fails | Use the Numbers table-qualified reference syntax ErrorBounds::B2 when needed. |
Apple reference guide. [A3] |
| Large input rounds visually to 1 | Many erf(x) values are mathematically extremely close to 1. For a tiny complement use ERFC(x) instead. |
Math plus separately documented ERFC. [A6, C1] |
Apple’s ERF page does not publish a complete coercion/error taxonomy or a maximum documented input magnitude. This manuscript therefore avoids inventing a Numbers error string, claiming text is silently coerced, or asserting reversed bounds are accepted or rejected in all versions. Native verification is required for any such additional claim. [A1]
Related functions and reader questions
ERF versus ERFC? ERF(x) integrates from 0 to x (with one argument), whereas ERFC(x) gives its mathematical complement integrated from x toward infinity. See Apple’s ERFC documentation. [A6]
Is ERF(-1,1) an 168.5% probability? No. An ERF two-bound integral is not directly a probability until the distribution scaling and normalization have been applied. [C1]
Does omitting upper mean upper=0? No. Apple explicitly says one-argument ERF(lower) integrates between 0 and lower. That is not the same as setting its second argument to zero in the two-bound form. [A1]
Can I use an input cell rather than constants? Yes. Put a number in B2 and enter =ERF(B2) in C2; this is a standard Numbers table-cell reference. [A3]
Was ERF introduced in Numbers 15.4? The 15.4 help documents ERF but does not establish its introduction version. Don’t turn the documentation’s version number into a release-history assertion. [A1, A4, A5]
Official source and verification register β checked 13 October 2026
- [A1] Apple, ERF β Formulas and Functions Help 15.4: syntax, argument types, optional upper bound, three published examples.
- [A2] Apple, Functions overview 15.4: formula syntax and regional function conventions.
- [A3] Apple, Refer to cells in formulas 15.4: Numbers table references and anchoring.
- [A4] Apple, Numbers for Mac feature/release information: current Mac release context; not individual first-introduction proof.
- [A5] Apple, Numbers for iPad feature/release information: current iPad release context; not native function execution.
- [A6] Apple, ERFC β Formulas and Functions Help 15.4: direct companion function.
- [C1] Independent arithmetic, Python 3
math.erf,math.erfcandmath.sqrt, calculated 13 October 2026; not Apple Numbers execution. The source of computed examples is distinct from Apple’s published examples.
Native-validation limitation: No Apple Numbers execution or screenshots were performed. Relevant edge cases and locale-specific behavior require testing in the intended application.