2-Value vs 3-Value Boundary Value Analysis
The short answer
2-value boundary value analysis takes two coverage items for each boundary value - the boundary value and its closest neighbour in the adjacent partition; 3-value boundary value analysis takes three - the boundary value and both its neighbours.
That is the definition, and it is the easy half. The half that costs marks is counting: a border between two partitions has two boundary values, not one, so applying the 3-value rule at a single border yields four distinct coverage items rather than three. Most material online, and most candidates, answer three.
The syllabus defines a boundary value as the minimum or maximum of a partition, and it defines the coverage items per boundary value. It does not spell out what happens where two partitions meet, which is the gap this page exists to close - and the gap ISTQB itself published a white paper about in October 2025.
Side by side
| 2-value BVA | 3-value BVA | |
|---|---|---|
| Coverage items per boundary value | 2 - the boundary value and its closest neighbour in the adjacent partition | 3 - the boundary value and both its neighbours |
| Are all coverage items boundary values | Yes | No - the syllabus notes some coverage items may not be boundary values |
| At one border, for partitions ...9 and 10... | Boundary values 9 and 10. Items: 9, 10 - two in total | Boundary values 9 and 10. Items: 8, 9, 10 and 9, 10, 11 - four in total |
| Integer field accepting 10 to 99 | 9, 10, 99, 100 - four values | 8, 9, 10, 11, 98, 99, 100, 101 - eight values |
| Rigour | Weaker - can miss a boundary implemented one position inside the partition | Stronger - the syllabus states it may detect defects 2-value BVA overlooks |
| Applies to | Ordered partitions only | Ordered partitions only |
How to count without losing the mark
Nearly every lost mark here is an arithmetic slip rather than a misunderstanding. Work in this order.
- 1Read the stem for 'two-value', '2-value', 'three-value' or '3-value'. A v4.0 question will say which form it wants, because both are examinable.
- 2Check the scope. 'On the LOWER boundary only' is not the same as the whole range, and it is the most common way to lose a mark you actually understood.
- 3Decide whether the question means one boundary value or a border between two partitions. For a single boundary value, 3-value BVA gives exactly three coverage items. For a border, each side contributes a boundary value, so you get four.
- 4Confirm the partition is ordered. If the values are unordered categories, BVA does not apply at all and the question is really about equivalence partitioning.
- 5Count coverage items, not test cases, unless the question asks for test cases. One test case can carry several coverage items when the input has more than one field.
Why the count is four and not three
Take a field where 6 to 10 is one partition and 11 to 16 is the next. The border sits between 10 and 11, and both of those are boundary values - 10 is the maximum of the first partition, 11 is the minimum of the second. The syllabus says three coverage items for each boundary value: 9, 10, 11 for the first, and 10, 11, 12 for the second. Taking the union gives 9, 10, 11, 12. Four items, not three.
This matters because the two you would drop are the two that catch real defects. Test only 9, 10, 11 and you never check that 12 is still handled by the second partition; test only 10, 11, 12 and you never check that 9 is still handled by the first. Each of the four items detects a misplacement the other three do not.
The syllabus's own example makes the same point from the other direction. A decision that should fire for values up to and including 10 is instead implemented to fire only when the value is exactly 10. The partitions here are unbounded on their far sides, so the border has just two boundary values, 10 and 11, and the 2-value coverage items are 10 and 11 - neither of which detects the defect, because 10 fires either way and 11 fires under neither. The value 9, which only 3-value BVA adds, is what exposes it. Note that 3-value BVA gives four items at this border too: 9, 10, 11 and 12.
None of this changes how you find the partitions. Boundary value analysis is always applied on top of equivalence partitioning, never instead of it - the boundaries you test are the edges of partitions you have already identified. A question that hands you a range and asks how many groups exist wants equivalence partitioning; one that asks which values to test at the edges wants BVA.
Worked exam questions
The answer and the reasoning are shown in full, including why each wrong option is wrong. These are drawn from the questions-and-answers page, so working them here does not spend the unseen pool the practice drills draw from.
1. A numeric field accepts integers from 10 to 99 inclusive. Using three-value boundary value analysis on the LOWER boundary only, which values must be tested?
K3- A.10, 11 and 12
- B.9, 10 and 11
- C.9 and 10
- D.10 and 11
Correct answer: B
Three-value BVA tests the boundary itself and its neighbours on both sides: 9 (just outside), 10 (the boundary), 11 (just inside).
2. An input field accepts integer ages from 18 to 65 inclusive. Using equivalence partitioning, how many partitions exist for this field?
K3- A.Two
- B.Three
- C.Forty-eight
- D.Four
Correct answer: B
One valid partition (18-65) and two invalid partitions (below 18, above 65). Equivalence partitioning covers the whole value domain, invalid values included.
Test your knowledge with real ISTQB-style questions
You’ve learned 2-Value vs 3-Value Boundary Value Analysis. Now sit a full 40-question paper and see if it holds up under the clock.
Or read questions with worked answers, or take a chapter-only practice drill.