diff --git a/dataflows/processors/concatenate.py b/dataflows/processors/concatenate.py
index c6cd8cf..31d4c40 100644
--- a/dataflows/processors/concatenate.py
+++ b/dataflows/processors/concatenate.py
@@ -6,10 +6,10 @@
 def concatenator(resources, all_target_fields, field_mapping):
     for resource_ in resources:
         for row in resource_:
-            processed = dict((k, '') for k in all_target_fields)
+            processed = dict((k, None) for k in all_target_fields)
             values = [(field_mapping[k], v) for (k, v)
                       in row.items()
-                      if k in field_mapping]
+                      if k in field_mapping and v is not None]
             if len(values) == 0:
                 message = 'Got an empty row after concatenation' +\
                     '(resource=%s, source=%r)' % (resource_.res.name, row)
diff --git a/tests/test_lib.py b/tests/test_lib.py
index 22e51c4..1212b5d 100644
--- a/tests/test_lib.py
+++ b/tests/test_lib.py
@@ -308,6 +308,31 @@ def test_concatenate():
     ]
 
 
+def test_concatenate_multifield():
+    from dataflows import concatenate
+
+    f = Flow(
+        [
+            {'a': 1, 'b': 2, 'c': None},
+            {'a': 2, 'b': None, 'c': 3},
+            {'a': 3, 'c': 4},
+            {'a': 3, 'b': 6, 'c': 4},
+        ],
+        concatenate({
+            'f1': ['a'],
+            'f2': ['b', 'c'],
+        })
+    )
+    results, _, _ = f.results()
+    assert results[0] == [
+        {'f1': 1, 'f2': 2},
+        {'f1': 2, 'f2': 3},
+        {'f1': 3, 'f2': 4},
+        {'f1': 3, 'f2': 4},
+    ]
+
+
+
 def test_filter_rows():
     from dataflows import filter_rows