diff --git a/lib/iris/fileformats/cf.py b/lib/iris/fileformats/cf.py
index f2fc895d1c..ddf7f2d64d 100644
--- a/lib/iris/fileformats/cf.py
+++ b/lib/iris/fileformats/cf.py
@@ -936,6 +936,11 @@ def __init__(self, filename, warn=False, monotonic=False):
 
         self._dataset = netCDF4.Dataset(self._filename, mode='r')
 
+        # Return variable data as np.ma.MaskedArray *only* when masked points
+        # are found in the slice requested, otherwise as 'normal' np.ndarray.
+        # Note: this suits us better, and was standard up to netcdf 1.3.
+        self._dataset.set_always_mask(False)
+
         # Issue load optimisation warning.
         if warn and self._dataset.file_format in ['NETCDF3_CLASSIC', 'NETCDF3_64BIT']:
             warnings.warn('Optimise CF-netCDF loading by converting data from NetCDF3 ' \