diff --git a/doc/source/whatsnew/v0.23.5.txt b/doc/source/whatsnew/v0.23.5.txt
index 88ea7a6caecfc..39c47303142a6 100644
--- a/doc/source/whatsnew/v0.23.5.txt
+++ b/doc/source/whatsnew/v0.23.5.txt
@@ -20,10 +20,11 @@ and bug fixes. We recommend that all users upgrade to this version.
 Fixed Regressions
 ~~~~~~~~~~~~~~~~~
 
-- Constructing a DataFrame with an index argument that wasn't already an
+- Constructing a :class:`DataFrame` with an index argument that wasn't already an
   instance of :class:`~pandas.core.Index` was broken in `4efb39f
   <https://github.com/pandas-dev/pandas/commit/4efb39f01f5880122fa38d91e12d217ef70fad9e>`_ (:issue:`22227`).
--
+- Passing :meth:`DataFrame.groupby` as grouper a callable or mapping which
+  returns tuples was broken in 0.21.1 (:issue:`22257`).
 -
 
 .. _whatsnew_0235.bug_fixes:
diff --git a/pandas/core/groupby/grouper.py b/pandas/core/groupby/grouper.py
index 35d4a024a4e6c..945d55c63da92 100644
--- a/pandas/core/groupby/grouper.py
+++ b/pandas/core/groupby/grouper.py
@@ -508,6 +508,15 @@ def _get_grouper(obj, key=None, axis=0, level=None, sort=True,
             warnings.warn(msg, FutureWarning, stacklevel=5)
             key = list(key)
 
+    if callable(key) or isinstance(key, dict):
+        if level is None:
+            key = group_axis.map(key)
+        else:
+            key = group_axis.get_level_values(level=level).map(key)
+            # If the grouper is a mapping, 'level' is _only_ used to determine
+            # the mapping input
+            level = None
+
     if not isinstance(key, list):
         keys = [key]
         match_axis_length = False
diff --git a/pandas/tests/groupby/test_grouping.py b/pandas/tests/groupby/test_grouping.py
index 58e9797dbeea5..5913c2f481e10 100644
--- a/pandas/tests/groupby/test_grouping.py
+++ b/pandas/tests/groupby/test_grouping.py
@@ -202,6 +202,23 @@ def test_grouper_creation_bug(self):
         expected = s.groupby(level='one').sum()
         assert_series_equal(result, expected)
 
+    @pytest.mark.parametrize('func', [False, True])
+    def test_grouper_returning_tuples(self, func):
+        # GH 22257 , both with dict and with callable
+        df = pd.DataFrame({'X': ['A', 'B', 'A', 'B'], 'Y': [1, 4, 3, 2]})
+        mapping = dict(zip(range(4), [('C', 5), ('D', 6)] * 2))
+
+        if func:
+            gb = df.groupby(by=lambda idx: mapping[idx], sort=False)
+        else:
+            gb = df.groupby(by=mapping, sort=False)
+
+        name, expected = list(gb)[0]
+        assert name == ('C', 5)
+        result = gb.get_group(name)
+
+        assert_frame_equal(result, expected)
+
     def test_grouper_column_and_index(self):
         # GH 14327
 
@@ -346,7 +363,7 @@ def test_groupby_grouper_f_sanity_checked(self):
         # when the elements are Timestamp.
         # the result is Index[0:6], very confusing.
 
-        pytest.raises(AssertionError, ts.groupby, lambda key: key[0:6])
+        pytest.raises(ValueError, ts.groupby, lambda key: key[0:6])
 
     def test_grouping_error_on_multidim_input(self, df):
         pytest.raises(ValueError,