
A matrix item is a table where each row is its own small question, and the only thing that varies is how many selections a row allows. Read the instruction line ('select one option in each row' versus 'select all that apply in each row'), work row by row on its own merits, and never leave a row blank. NCSBN publishes the format as 'Matrix (Grid)'; the grid-versus-multiple-choice split describes two behaviors of that one format, not two official item types.
A clarification before the content, because this is one of those places where the internet is more confident than the source material.
NCSBN’s published name for this format is “Matrix (Grid).” That is what appears in NCSBN’s own presentation materials, and “Matrix/Grid” is the term used in NCSBN’s item-writing research.
The distinction between a “matrix grid” and a “matrix multiple choice” — one selection per row versus multiple selections per row — is real, widely taught, and behaviorally accurate. It is not a distinction NCSBN publishes as two separately named item types. Treat it as a useful way of describing two behaviors of one format rather than as official taxonomy.
With that said, the distinction matters enormously in practice, because the two behave differently and mixing them up costs points.
A table. Rows down the left listing findings, interventions, or client data. Columns across the top listing categories. You place selections in the cells.
That is the whole format. Everything else is variation in how many selections a row permits.
The row asks a question with one correct answer among the columns.
Typical setups:
| Finding | Expected | Unexpected |
|---|---|---|
| Heart rate 118 | ||
| Temperature 37.0°C | ||
| Urine output 15 mL/hr |
Each finding is either expected or unexpected. Not both. One click per row.
Or a three-column version: Effective / Ineffective / Unrelated. Improving / Unchanged / Declining. Indicated / Contraindicated / Non-essential.
The tell: the columns are mutually exclusive. Nothing can be both expected and unexpected.
How to work it: go row by row, ignoring the rest of the table. Each row is an independent true/false or three-way judgment. Do not try to balance the answers or assume the columns should come out evenly distributed.
The row can legitimately belong to more than one column.
Typical setup:
| Intervention | Pneumonia | Heart failure | COPD exacerbation |
|---|---|---|---|
| Administer oxygen | |||
| Elevate head of bed | |||
| Obtain sputum culture |
Oxygen applies to all three. Head of bed elevation applies to all three. Sputum culture applies to pneumonia only.
The tell: the columns are conditions or categories that can overlap, and the row content could reasonably apply to several.
How to work it: for each row, evaluate each cell separately. “Does this intervention apply to this condition?” is a distinct yes/no question per cell. A row may end up with three checks or one.
Read the instruction line. It says what to do. “Select one option in each row” versus “select all that apply in each row.” This is the reliable method and it takes three seconds.
If the instruction is ambiguous, read the column headers. Mutually exclusive headers mean one per row. Overlapping categories mean potentially several.
Check whether the interface lets you. Radio buttons permit one selection per row; checkboxes permit several. The interface enforces the rule.
That last point is worth internalizing. You usually cannot make the wrong kind of error if you are paying attention — the software will not let you select two options in a one-per-row item. The real risk is the opposite: under-selecting in a multi-select row because you assumed one per row.
Assuming a pattern. Candidates notice they have marked four rows “unexpected” in a row and start doubting themselves. There is no requirement that answers be distributed evenly. Judge each row on its merits.
Leaving rows blank. Every row needs an answer. A blank row cannot earn credit under any scoring method. If you are unsure, reason to your best answer and select it.
Reading the table as a whole. A matrix looks like a lot of information at once and that feels overwhelming. It is not one big question. It is six or eight small ones sharing a header.
Rushing because it looks mechanical. Matrix items are often testing precise knowledge — which findings are expected in a specific condition, which interventions are indicated for a specific diagnosis. The grid layout makes them feel like busywork. They are not.
Matrix items have multiple correct selections, so partial credit applies.
NCSBN publishes three partial-credit methods — plus/minus, zero/one, and rationale scoring — but does not publish which method applies to which item type. Anyone telling you “matrix items are scored 0/1” is inferring.
What this means for you is simple and does not depend on the mapping: answer every row. Under zero/one scoring, a wrong answer costs nothing and a right one earns a point, so blanks are strictly worse than guesses. Under plus/minus scoring, a reasoned answer is still better than a certain zero. Either way, complete the grid.
A matrix item is a table where each row is its own small question. Read the instruction to learn whether a row takes one answer or several, work row by row, and never leave a row empty.
References
All sources accessed August 2026. NCSBN names the format “Matrix (Grid)”; the grid-versus-multiple-choice distinction describes two behaviors of that format and is not NCSBN taxonomy. NCSBN does not publish which scoring method applies to which item type.
Common follow-up questions on The 2026 Test Plan & Next Gen NCLEX.
In the one-selection-per-row behavior the columns are mutually exclusive (such as Expected versus Unexpected), so each row gets a single click. In the multiple-selection behavior the columns are conditions or categories that can overlap, so a row may legitimately belong to several and you evaluate each cell as a separate yes-or-no question. NCSBN does not publish these as two separately named item types; treat them as two behaviors of one format.
Read the instruction line, which says 'select one option in each row' or 'select all that apply in each row'; this is the reliable method and takes three seconds. If the instruction is ambiguous, mutually exclusive column headers mean one per row while overlapping categories mean potentially several. The interface also enforces the rule: radio buttons permit one selection per row, checkboxes permit several.
Under-selecting in a multi-select row because you assumed one answer per row. The software will not let you pick two options in a one-per-row item, so the real risk is the opposite, missing selections when a row can take several. Assuming answers should be evenly distributed and doubting yourself after marking several rows the same way also costs points.
Matrix items have multiple correct selections, so partial credit applies. NCSBN publishes three partial-credit methods (plus/minus, zero/one, and rationale scoring) but does not publish which method applies to which item type, so anyone saying 'matrix items are scored 0/1' is inferring. Either way, answer every row, because a blank row cannot earn credit under any scoring method.
Go row by row, treating each as an independent question and ignoring the rest of the table. A matrix looks like a lot of information at once, but it is really six or eight small questions sharing a header, not one big one. Do not rush because it looks mechanical; these items often test precise knowledge of which findings or interventions fit a specific condition.
3,000+ NCLEX questions with think-like-a-nurse rationales, unfolding case studies and adaptive mock exams: free to start, no card required.
Start Free Trial →