ismember MATLAB: Syntax, Rows and Tolerance

In MATLAB, the ismember function tests set membership across arrays, matrices, cell arrays, and data tables. It identifies whether elements in a test dataset exist within a reference set, returning a logical array indicating membership and an optional index array identifying the location of each match. In research data analysis, mastering ismember and matlab ismember operations is essential for cohort segmentation, multi-condition experimental filtering, coordinate verification, and relational data matching across large scientific datasets.

What it is used for in research

Scientific computing workflows routinely process complex observational, experimental, and simulated datasets that require matching observations against reference lists or inclusion criteria. In scientific computing in MATLAB, the ismember function serves several primary research functions:

• Experimental Cohort Filtering: Isolating participant identifiers, animal tag numbers, or clinical sample IDs that match specific inclusion or exclusion criteria across multi-center trials.
• Tabular Data Subsetting: Selecting rows from a MATLAB data table where categorical groupings, subject markers, or timestamp records match predefined target lists.
• Spatial Coordinate and Grid Verification: Testing whether 2D or 3D coordinate pairs (such as sensor locations or pixel coordinates) exist within a calibration grid using the row-comparison syntax.
• Genomic and Multi-Omics Feature Matching: Screening high-throughput probe IDs, gene symbols, or metabolite accession strings stored in a MATLAB cell array or string array against reference biological pathways.
• Trial State and Event Tracking: Identifying discrete simulation events or hardware state transitions that belong to predefined behavioral phases.

Install or access (official source only)

The ismember function is a built-in core function of standard MATLAB and GNU Octave. It does not require any add-on toolboxes, specialized licenses, or external package installations.

• MATLAB Desktop: Accessible via local installation on Windows, macOS, or Linux, acquired through the official MathWorks portal (mathworks.com/downloads) under standard academic campus-wide, institutional, or individual licenses.
• MATLAB Online: Accessible through any modern web browser at matlab.mathworks.com using authenticated MathWorks institutional or student account credentials.
• Command-Line Documentation: To inspect the integrated reference documentation and function signatures directly in the MATLAB Command Window, enter:

  help ismember
  doc ismember

• Environment Verification: Running which ismember confirms that the function resolves to the built-in elementary math and data structures library of your active MATLAB runtime.

Step-by-step task

Follow these numbered steps to execute set membership checks, index lookups, row comparisons, and table filtering in MATLAB:

Step 1: Test 1D Vector Membership

To check which elements of vector A exist anywhere in vector B, call ismember(A, B). The function returns a logical vector lia possessing the exact dimensions of A:

% Define sample experimental measurements and target threshold values
A = [12, 24, 35, 48, 60, 72];
B = [24, 60, 95];

% Test membership
lia = ismember(A, B);
% lia returns: [0, 1, 0, 0, 1, 0]

Step 2: Retrieve Matching Index Positions

Requesting two output arguments returns both the logical indicator and an array locb containing the first matching index position in B for each corresponding element in A (or 0 if no match exists):

[lia, locb] = ismember(A, B);
% lia  returns: [0, 1, 0, 0, 1, 0]
% locb returns: [0, 1, 0, 0, 2, 0]

Here, 24 is located at index 1 of B, and 60 is located at index 2 of B. Elements without matches return 0.

Step 3: Compare Rows in Multi-Column Matrices

When working with spatial coordinates, feature vectors, or multi-channel measurements, use the 'rows' flag. Both matrices must have an identical number of columns:

% Sample coordinate pairs [X, Y]
sampleCoords = [1.0, 2.0; 3.0, 4.0; 5.0, 6.0; 7.0, 8.0];
calibrationGrid = [0.0, 0.0; 3.0, 4.0; 7.0, 8.0];

% Match entire rows
[liaRows, locbRows] = ismember(sampleCoords, calibrationGrid, 'rows');
% liaRows  returns: [0; 1; 0; 1]
% locbRows returns: [0; 2; 0; 3]

Step 4: Filter Research Data Tables Using Logical Indexing

In structured scientific analysis, use ismember to construct a logical mask for MATLAB indexing across tabular datasets:

% Create an experimental cohort table
SubjectID = [101; 102; 103; 104; 105];
Score = [78.2; 88.5; 64.1; 92.0; 81.4];
Cohort = categorical({'Control'; 'Active'; 'Active'; 'Control'; 'Active'});
trialData = table(SubjectID, Score, Cohort);

% Define target subjects to extract
selectedSubjects = [102; 105];

% Extract subset using logical masking
mask = ismember(trialData.SubjectID, selectedSubjects);
activeCohort = trialData(mask, :);

Step 5: Test Membership Across Cell Arrays and String Arrays

For qualitative descriptors, sample annotations, or biological identifiers, ismember operates natively on string arrays and cell arrays of character vectors:

geneList = ["BRCA1", "TP53", "EGFR", "MYC", "PTEN"];
targetPanel = ["TP53", "PTEN", "KRAS"];

geneMask = ismember(geneList, targetPanel);
% geneMask returns: [0, 1, 0, 0, 1]

Output and how to read it

Understanding the structure and data types returned by ismember prevents indexing errors and unexpected logical operations:

• First Output (Lia): A logical array matching the exact dimensions and size of the input array A. Each element contains true (logical 1) if that specific element of A is present anywhere within B, and false (logical 0) otherwise. When using 'rows' on an m × n matrix, Lia is an m × 1 logical column vector.
• Second Output (Locb): An array of type double matching the size of A. If an element in A matches one or more elements in B, Locb contains the lowest linear index in B where that value occurs. If an element has no match in B, Locb evaluates to 0.
• Zero-Index Precaution: MATLAB uses 1-based indexing. Because unmatched elements evaluate to 0 in Locb, attempting to index directly into B(locb) will trigger an out-of-bounds error (Subscript indices must either be real positive integers or logicals). Always gate indexing with lia:

  matchedB = B(locb(lia));

• Algorithmic Complexity: ismember sorts the reference set and performs binary search lookups, providing O(n log2(m)) scaling efficiency rather than the O(n × m) penalty incurred by brute-force nested comparison loops.

Troubleshooting

Address common operational issues and edge cases when working with ismember in MATLAB:

1. Floating-Point Rounding Failures (The Precision Trap)

Because ismember demands exact binary bit-level equality, floating-point calculations frequently fail to match due to IEEE 754 precision artifacts. For example, ismember(0.1 + 0.2, 0.3) evaluates to false.

Fix: Use the dedicated tolerance-aware companion function ismembertol for floating-point data:

A_float = [0.1 + 0.2, 1.0000001];
B_float = [0.3, 1.0];

% Fails with exact ismember
exactMatch = ismember(A_float, B_float); % returns [0, 0]

% Succeeds with ismembertol using relative tolerance
tolMatch = ismembertol(A_float, B_float, 1e-5); % returns [1, 1]

2. Matrix Dimension Mismatch Under ‘rows’

If you execute ismember(A, B, 'rows') when matrix A and matrix B have different numbers of columns, MATLAB throws an error: Inputs must have the same number of columns to use the 'rows' flag.

Fix: Inspect column counts with size(A, 2) and size(B, 2). Ensure both matrices represent identical variable dimensions before comparing row entries.

3. Character Vectors vs String Arrays

Passing a single character vector such as 'Control' treats the input as a 1 × 7 character array, testing each individual character against B rather than evaluating the entire word.

Fix: Enclose text elements in double quotes (string array: "Control") or curly braces (cell array of character vectors: {'Control'}) to treat text labels as distinct individual tokens.

4. Working with NaN Values

By standard IEEE 754 definitions, NaN == NaN evaluates to false. In ismember, however, NaN values are considered equal to each other when comparing numeric arrays. If A contains NaN and B contains NaN, ismember(NaN, [1, NaN, 3]) returns true. If you need to disregard missing values, filter them beforehand using isnan.

Free alternatives (table)

While MATLAB provides a polished environment for matrix operations, several open-source programming languages offer equivalent set membership capabilities for reproducible research. The following table compares primary tools:

Software / Package	Primary Function	Row-Wise Matching	Index Retrieval	Tolerance Comparison	License Model
MATLAB	ismember(A, B)	Yes ('rows' flag)	Yes ([lia, locb])	Yes (via ismembertol)	Commercial proprietary
GNU Octave	ismember(a, s)	Yes ('rows' flag)	Yes ([tf, s_idx])	Yes (via ismembertol)	Free open-source (GPLv3)
Python (NumPy)	numpy.isin(a, b)	Requires structured view	Requires np.nonzero()	Requires np.isclose()	Free open-source (BSD)
R (base)	x %in% table	Requires merge / semi_join	Yes (via match(x, table))	Custom distance threshold	Free open-source (GPLv2/v3)
Julia (Base)	in.(A, Ref(B))	Yes (via eachrow)	Yes (via findfirst)	Yes (via isapprox)	Free open-source (MIT)

For researchers migrating or establishing open-science pipelines without commercial dependencies, GNU Octave provides direct drop-in script compatibility with MATLAB ismember and ismembertol syntax. For multi-language workflows evaluating Python, R, and Julia ecosystems, consult our detailed analysis in MATLAB vs R for research.

FAQ

What does ismember do in MATLAB?

The ismember function checks whether elements of a test array exist in a reference set. It outputs a logical array indicating which elements are present, and can optionally output an index vector showing the first matching position in the reference set.

How do I return the index location of matching elements?

Call ismember with two output arguments: [lia, locb] = ismember(A, B). The second variable, locb, contains the 1-based index of the first occurrence in B for each element in A that matched. Elements in A that do not appear in B receive a value of 0.

Why does ismember fail on decimal values?

Standard ismember performs exact bitwise comparison. Due to floating-point rounding errors inherent to binary arithmetic, computed numbers rarely equal identical decimal representations. Use ismembertol(A, B, tol) to evaluate floating-point membership within a defined numerical tolerance.

How do I match entire rows between two matrices?

Pass the 'rows' argument: lia = ismember(A, B, 'rows'). Both matrices must share the same number of columns. MATLAB evaluates each row as a single unit and returns a logical column vector indicating which rows of A are present in B.

Can ismember filter rows in a MATLAB data table?

Yes. You can test membership on a specific table column, such as mask = ismember(dataTable.SubjectID, targetIDs), and then use that logical vector to slice the table: filteredTable = dataTable(mask, :).

What is the difference between ismember and intersect?

The ismember function evaluates set membership while preserving the exact size, shape, and element order of the input array A. In contrast, intersect extracts only the common values between two sets, returning a sorted, deduplicated vector of unique shared elements.

Sources (official docs)

• MathWorks. MATLAB ismember Documentation: Array Set Membership. Available at: https://www.mathworks.com/help/matlab/ref/ismember.html (read 2026-11-09).
• MathWorks. MATLAB ismembertol Documentation: Set Membership Within Tolerance. Available at: https://www.mathworks.com/help/matlab/ref/double.ismembertol.html (read 2026-11-09).
• GNU Octave Documentation. Set Operations: ismember and ismembertol. Available at: https://docs.octave.org/latest/Set-Operations.html (read 2026-11-09).
• NumPy Documentation. numpy.isin: Element-wise Set Membership. Available at: https://numpy.org/doc/stable/reference/generated/numpy.isin.html (read 2026-11-09).
• R Core Team. Value Matching: match and %in%. Available at: https://stat.ethz.ch/R-manual/R-devel/library/base/html/match.html (read 2026-11-09).
• Bucknell University. Introduction to MATLAB (Handout H01). Available at: https://www.eg.bucknell.edu/~rmn009/H01.IntroMATLAB.pdf (read 2026-11-09).