MatNWB Logo

codecov Azure DevOps tests

MatNWB is a Matlab interface for reading and writing Neurodata Without Borders (NWB) 2.x files.


Step 1: Download MatNWB

Download the current release of MatNWB from the MatNWB releases page or from the View NeurodataWithoutBorders/matnwb on File Exchange. You can also check out the latest development version via

git clone

Step 2a: Reading from a NWB File

If you wish to read from a NWB file, you can do so using the nwbRead command:

File = nwbRead('/path/to/file.nwb');

The returned NwbFile object provides an in-memory view of the underlying NWB data. For more information, see the NWB Overview Documentation

Step 2b: Writing a NWB File

From the MATLAB command line, add MatNWB to the path. The generated classes are normally placed in the +types subdirectory in the MatNWB installation directory. As MATLAB packages, these generated classes comprise the building blocks you will need to write your NWB file.


Once you have configured your NWB File, you may write the NwbFile object to disk using the nwbExport function.

nwbExport(NwbFile, 'path/to/file.nwb');

Extensions: Generate MatNWB Classes for Extensions

The generateExtension command generates extension classes given a file path to the extension’s namespace. This can be useful if you need to work with NWB Extension Schemas outside of Core.

generateExtension('schema/core/nwb.namespace.yaml', '.../my_extensions1.namespace.yaml',...);

Advanced: Generating Legacy MatNWB Classes

The generateCore command can generate older versions of the nwb schema.

generateCore('2.1.0'); % generates classes for NWB schema version 2.1.0

Supported schema versions are provided in the MatNWB installation directory under nwb-schema.


Intro to MatNWB

Basic File Reading a demo showcase for basic visualization from a DANDI dataset.
Extracellular Electrophysiology YouTube walkthrough
Calcium Imaging YouTube walkthrough

Intracellular Electrophysiology



Dynamic tables


Advanced data write YouTube walkthrough

Using Dynamically Loaded Filters

Remote read

Scratch Space

API Documentation

For more information regarding the MatNWB API or any of the NWB Core types in MatNWB, visit the MatNWB API Documentation pages.

Under the Hood

NWB files are HDF5 files with data stored according to the Neurodata Without Borders (NWB) schema. The schema is described in a set of YAML documents which defines the various types and their attributes.

Certain functions, like generateCore and nwbRead, automatically read these specifications and converts them to a MATLAB class file. These classes generally map directly to attributes and constraints of the types defined in the schema.


MatNWB is available online at

Data Dimensions

NWB files use the HDF5 format to store data. There are two main differences between the way MATLAB and HDF5 represents dimensions. The first is that HDF5 is C-ordered, which means it stores data is a rows-first pattern, and the MATLAB is F-ordered, storing data in the reverse pattern, with the last dimension of the array stored consecutively. The result is that the data in HDF5 is effectively the transpose of the array in MATLAB. The second difference is that HDF5 can store 1-D arrays, but in MATLAB the lowest dimensionality of an array is 2-D. Due to differences in how MATLAB and HDF5 represent data, the dimensions of datasets are flipped when writing to/from file in MatNWB. This behavior differs depending on whether VectorData use DataPipe objects to contain the data. It’s important to keep in mind the mappings below to make sure is written to and read from file as expected.

without DataPipes

Writing to File

in MatNWB
in HDF5
(M, 1) (M,)
(1, M) (M,)
(P, O, N, M) (M, N, O, P)

Reading from File

in HDF5
in MatNWB
(M,) (M,1)
(M, N, O, P) (P, O, N, M)

NOTE: MATLAB does not support 1D datasets. HDF5 datasets of size (M,) are loaded into MATLAB as datasets of size (M,1). To avoid changes in dimensions when writing to/from file use column vectors for 1D datasets.

with DataPipes

Writing to File

in MatNWB
in HDF5
(M, 1) (1, M)
(1, M) (M, 1)/(M,)**
(P, O, N, M) (M, N, O, P)

** Use scalar as input to ‘maxSize’ argument to write dataset of shape (N,)

Reading from File

in HDF5
in MatNWB
(M, 1) (1, M)
(1, M) (M, 1)
(M,) (M, 1)
(M, N, O, P) (P, O, N, M)


The NWB schema has regular updates and is open to addition of new types along with modification of previously defined types. As such, certain type presumptions made by MatNWB may be invalidated in the future from a NWB schema. Furthermore, new types may require implementations that will be missing in MatNWB until patched in.

For those planning on using matnwb alongside pynwb, please keep the following in mind:

The master branch in this repository is considered perpetually unstable. If you desire Matnwb’s full functionality (full round-trip with nwb data), please consider downloading the more stable releases in the Releases tab. Most releases will coincide with nwb-schema releases and guarantee compatibility of new features introduced with the schema release along with backwards compatibility with all previous nwb-schema releases.

This package reads and writes NWB 2.0 files and does not support older formats.


Basic Data Retrieval | showcases how one would read and process converted NWB file data to display a raster diagram.

Conversion of Real Electrophysiology/Optophysiology Data | converts Electrophysiology/Optophysiology Data recorded from:

Li, Daie, Svoboda, Druckman (2016); Data and simulations related to: Robust neuronal dynamics in premotor cortex during motor planning. Li, Daie, Svoboda, Druckman, Nature.

Analysis examples will be added in the dandi-example-live-scripts repo

Third-party Support

The +contrib folder contains tools for converting from other common data formats/specifications to NWB. Currently supported data types are TDT, MWorks, and Blackrock. We are interested in expanding this section to other data specifications and would greatly value your contribution!


Run the test suite with nwbtest.


  1. “A class definition must be in an “@” directory.”

Make sure that there are no “@” signs anywhere in your full file path. This includes even directories that are not part of the matnwb root path and any “@” signs that are not at the beginning of the directory path.

Alternatively, this issue disappears after MATLAB version 2017b. Installing this version may also resolve these issues. Note that the updates provided with 2017b should also be installed.

  1. I Have Issues Reading From a NWB File!

Some simple methods to troubleshoot failed NWB file reads can be found in the NWB Overview Documentation.