diff --git a/doc/manual/c_reference.rst b/doc/manual/c_reference.rst
index ce9f3d15407..3e7e7f0fadf 100644
--- a/doc/manual/c_reference.rst
+++ b/doc/manual/c_reference.rst
@@ -1479,7 +1479,7 @@ anatomy only, not on the MEG/EEG data to be analyzed.
 		R_{13} & R_{13} & R_{13} & z_0 \\
 		0 & 0 & 0 & 1
 		\end{bmatrix}
-	     
+
 defined so that if the augmented location vectors in MRI
 head and MRI coordinate systems are denoted by :math:`r_{head}[x_{head}\ y_{head}\ z_{head}\ 1]` and :math:`r_{MRI}[x_{MRI}\ y_{MRI}\ z_{MRI}\ 1]`,
 respectively,
diff --git a/mne/io/base.py b/mne/io/base.py
index 60921292d50..79d4e74cf87 100644
--- a/mne/io/base.py
+++ b/mne/io/base.py
@@ -14,6 +14,7 @@
 import numpy as np
 
 from .constants import FIFF
+from .utils import _construct_bids_filename
 from .pick import pick_types, channel_type, pick_channels, pick_info
 from .pick import _pick_data_channels, _pick_data_or_ica
 from .meas_info import write_meas_info
@@ -2258,9 +2259,8 @@ def _write_raw(fname, raw, info, picks, fmt, data_type, reset_range, start,
         if split_naming == 'neuromag':
             # insert index in filename
             use_fname = '%s-%d%s' % (base, part_idx, ext)
-        else:
-            # insert index in filename
-            use_fname = '%s_part-%02d_meg%s' % (base, part_idx, ext)
+        elif split_naming == 'bids':
+            use_fname = _construct_bids_filename(base, ext, part_idx)
             # check for file existence
             _check_fname(use_fname, overwrite)
 
diff --git a/mne/io/fiff/tests/test_raw_fiff.py b/mne/io/fiff/tests/test_raw_fiff.py
index e3c1fc6d70c..6422983496c 100644
--- a/mne/io/fiff/tests/test_raw_fiff.py
+++ b/mne/io/fiff/tests/test_raw_fiff.py
@@ -398,24 +398,26 @@ def test_split_files():
     raw_crop = raw_1.copy().crop(0, 1.)
 
     assert_allclose(raw_1.buffer_size_sec, 10., atol=1e-2)  # samp rate
-    split_fname = op.join(tempdir, 'split_raw.fif')
+    split_fname = op.join(tempdir, 'split_raw_meg.fif')
     # intended filenames
-    split_fname_elekta_part2 = op.join(tempdir, 'split_raw-1.fif')
+    split_fname_elekta_part2 = op.join(tempdir, 'split_raw_meg-1.fif')
     split_fname_bids_part1 = op.join(tempdir, 'split_raw_part-01_meg.fif')
     split_fname_bids_part2 = op.join(tempdir, 'split_raw_part-02_meg.fif')
     raw_1.set_annotations(Annotations([2.], [5.5], 'test'))
-    raw_1.save(split_fname, buffer_size_sec=1.0, split_size='10MB')
+    with pytest.warns(RuntimeWarning, match='does not conform to MNE'):
+        raw_1.save(split_fname, buffer_size_sec=1.0, split_size='10MB')
 
     # check that the filenames match the intended pattern
     assert op.exists(split_fname_elekta_part2)
     # check that filenames are being formatted correctly for BIDS
-    raw_1.save(split_fname,
-               buffer_size_sec=1.0, split_size='10MB',
-               split_naming='bids',
-               overwrite=True)
+    with pytest.warns(RuntimeWarning, match='does not conform to MNE'):
+        raw_1.save(split_fname, buffer_size_sec=1.0, split_size='10MB',
+                   split_naming='bids', overwrite=True)
     assert op.exists(split_fname_bids_part1)
     assert op.exists(split_fname_bids_part2)
 
+    split_fname = op.join(tempdir, 'split_raw.fif')
+    raw_1.save(split_fname, buffer_size_sec=1.0, split_size='10MB')
     raw_2 = read_raw_fif(split_fname)
     assert_allclose(raw_2.buffer_size_sec, 1., atol=1e-2)  # samp rate
     assert_array_almost_equal(raw_1.annotations.onset, raw_2.annotations.onset)
diff --git a/mne/io/utils.py b/mne/io/utils.py
index 2cbd878589b..81617eedf84 100644
--- a/mne/io/utils.py
+++ b/mne/io/utils.py
@@ -249,3 +249,18 @@ def _synthesize_stim_channel(events, n_samples):
     for onset, duration, trigger in events:
         stim_channel[onset:onset + duration] = trigger
     return stim_channel
+
+
+def _construct_bids_filename(base, ext, part_idx):
+    """Construct a BIDS compatible filename for split files."""
+    # insert index in filename
+    deconstructed_base = base.split('_')
+    bids_supported = ['meg', 'eeg', 'ieeg']
+    for mod in bids_supported:
+        if mod in deconstructed_base:
+            idx = deconstructed_base.index(mod)
+            modality = deconstructed_base.pop(idx)
+    base = '_'.join(deconstructed_base)
+    use_fname = '%s_part-%02d_%s%s' % (base, part_idx, modality, ext)
+
+    return use_fname