diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index af5f1d8c66cf9bb018b8d81cfe867c04ea64a711..0347a5df351c952d247feb8efaa3416d70d62c36 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -2,6 +2,27 @@ This document contains the ``fslpy`` release history in reverse chronological
 order.
 
 
+2.1.0 (Under development)
+-------------------------
+
+
+Added
+^^^^^
+
+
+* New tensor conversion routines in the :mod:`.dtifit` module (Michiel
+  Cottaar).
+
+
+Fixed
+^^^^^
+
+
+* The :class:`.FeatDesign` class now handles "compressed" voxelwise EV files,
+  such as those generated by `PNM
+  <https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PNM>`_.
+
+
 2.0.1 (Monday April 1st 2019)
 -----------------------------
 
diff --git a/fsl/data/dtifit.py b/fsl/data/dtifit.py
index a58dfe0df6179d20ea8ede8f63acffa5fba00b3c..360eeb9003d6e7d5eb799fa4b7d278e38371399b 100644
--- a/fsl/data/dtifit.py
+++ b/fsl/data/dtifit.py
@@ -3,6 +3,7 @@
 # dtifit.py - The DTIFitTensor class, and some related utility functions.
 #
 # Author: Paul McCarthy <pauldmccarthy@gmail.com>
+# Author: Michiel Cottaar <michiel.cottaar@ndcn.ox.ac.uk>
 #
 """This module provides the :class:`.DTIFitTensor` class, which encapsulates
 the diffusion tensor data generated by the FSL ``dtifit`` tool.
diff --git a/fsl/data/featdesign.py b/fsl/data/featdesign.py
index 633b712b4b04f9de986ae096612987e6814a00a8..ce5b41f592609b5dbd6bc8b987e04016567a761c 100644
--- a/fsl/data/featdesign.py
+++ b/fsl/data/featdesign.py
@@ -214,7 +214,7 @@ class FEATFSFDesign(object):
                             'for ev {}'.format(ev.index))
                 continue
 
-            design[:, ev.index] = ev.image[x, y, z, :]
+            design[:, ev.index] = ev.getData(x, y, z)
 
         return design
 
@@ -228,6 +228,17 @@ class VoxelwiseEVMixin(object):
     ``filename`` Path to the image file containing the data for this EV
     ``image``    Reference to the :class:`.Image` object
     ============ ======================================================
+
+    Some FSL tools (e.g. `PNM
+    <https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PNM>`_) generate *compressed*
+    voxelwise EVs, where there is only one voxel per slice. For example,
+    if the input data has shape ``(x, y, z, t)``, and slices are acquired
+    along the Z plane, voxelwise EV files generated by PNM will have shape
+    ``(1, 1, z, t)``.
+
+    Therefore, using the :meth:`getData` method is preferable to accessing
+    the :meth:`image` directly, as ``getData`` will check for compressed
+    images, and adjust the voxel coordinates accordingly.
     """
 
     def __init__(self, filename):
@@ -272,6 +283,29 @@ class VoxelwiseEVMixin(object):
         return self.__image
 
 
+    def getData(self, x, y, z):
+        """Returns the data at the specified voxel, taking into account
+        compressed voxelwise EVs.
+        """
+        image = self.image
+
+        if image is None:
+            return None
+
+        dx, dy, dz = image.shape[:3]
+
+        # "Compressed" images have one voxel
+        # per slice, i.e. they have shape
+        # [1, 1, nslices, ntimepoints] (if
+        # Z is the slice plane).
+        if sum((dx == 1, dy == 1, dz == 1)) == 2:
+            if dx == 1: x = 0
+            if dy == 1: y = 0
+            if dz == 1: z = 0
+
+        return image[x, y, z, :]
+
+
 class EV(object):
     """Class representing an explanatory variable in a FEAT design matrix.
 
diff --git a/tests/test_featdesign.py b/tests/test_featdesign.py
index dd821127e925bf37a15f17fab5f40d40b5268a26..5c5d1a7e3433cb2c62bcb1c13b0e4bf061aa6028 100644
--- a/tests/test_featdesign.py
+++ b/tests/test_featdesign.py
@@ -412,3 +412,19 @@ def test_VoxelwiseEVs():
             exp = img.dataobj[x, y, z, :]
             assert np.all(ev1.image[x, y, z, :] == exp)
             assert np.all(ev2.image[x, y, z, :] == exp)
+
+
+def test_compressed_voxelwise_ev():
+
+    testcases = [((1, 1, 10, 10), (0, 0, 5)),
+                 ((1, 10, 1, 10), (0, 5, 0)),
+                 ((10, 1, 1, 10), (5, 0, 0))]
+
+    with tempdir():
+
+        for shape, vox in testcases:
+            img = tests.make_random_image('vev.nii.gz',  shape)
+            vev = featdesign.VoxelwiseEV(0, 0, 'ev1', 'vev.nii.gz')
+            x, y, z = vox
+
+            assert np.all(vev.getData(5, 5, 5) == img.dataobj[x, y, z, :])