diff --git a/databricks/koalas/frame.py b/databricks/koalas/frame.py
index a911ecf714..3d7f807d7a 100644
--- a/databricks/koalas/frame.py
+++ b/databricks/koalas/frame.py
@@ -5261,37 +5261,24 @@ def replace(
             if len(value) != len(to_replace):
                 raise ValueError("Length of to_replace and value must be same")
 
-        subset = self._internal.data_spark_column_names
-
-        sdf = self._internal.resolved_copy.spark_frame
-        if (
-            isinstance(to_replace, dict)
-            and value is None
-            and (not any(isinstance(i, dict) for i in to_replace.values()))
+        if isinstance(to_replace, dict) and (
+            value is not None or all(isinstance(i, dict) for i in to_replace.values())
         ):
-            sdf = sdf.replace(to_replace, value, subset)
-        elif isinstance(to_replace, dict):
-            for name, replacement in to_replace.items():
-                if isinstance(name, str):
-                    name = (name,)
-                df_column = self._internal.spark_column_name_for(name)
-                if isinstance(replacement, dict):
-                    sdf = sdf.replace(replacement, subset=df_column)
+
+            def op(kser):
+                if kser.name in to_replace:
+                    return kser.replace(to_replace=to_replace[kser.name], value=value, regex=regex)
                 else:
-                    sdf = sdf.withColumn(
-                        df_column,
-                        F.when(scol_for(sdf, df_column) == replacement, value).otherwise(
-                            scol_for(sdf, df_column)
-                        ),
-                    )
+                    return kser
+
         else:
-            sdf = sdf.replace(to_replace, value, subset)
+            op = lambda kser: kser.replace(to_replace=to_replace, value=value, regex=regex)
 
-        internal = self._internal.with_new_sdf(sdf)
+        kdf = self._apply_series_op(op)
         if inplace:
-            self._internal = internal
+            self._internal = kdf._internal
         else:
-            return DataFrame(internal)
+            return kdf
 
     def clip(self, lower: Union[float, int] = None, upper: Union[float, int] = None) -> "DataFrame":
         """
diff --git a/databricks/koalas/tests/test_dataframe.py b/databricks/koalas/tests/test_dataframe.py
index a12788f872..63b852833b 100644
--- a/databricks/koalas/tests/test_dataframe.py
+++ b/databricks/koalas/tests/test_dataframe.py
@@ -1577,6 +1577,7 @@ def test_replace(self):
         self.assert_eq(kdf.replace({"A": 0, "B": 5}, 100), pdf.replace({"A": 0, "B": 5}, 100))
 
         self.assert_eq(kdf.replace({"A": {0: 100, 4: 400}}), pdf.replace({"A": {0: 100, 4: 400}}))
+        self.assert_eq(kdf.replace({"X": {0: 100, 4: 400}}), pdf.replace({"X": {0: 100, 4: 400}}))
 
         # multi-index columns
         columns = pd.MultiIndex.from_tuples([("X", "A"), ("X", "B"), ("Y", "C")])
@@ -1600,6 +1601,9 @@ def test_replace(self):
         self.assert_eq(
             kdf.replace({("X", "A"): {0: 100, 4: 400}}), pdf.replace({("X", "A"): {0: 100, 4: 400}})
         )
+        self.assert_eq(
+            kdf.replace({("X", "B"): {0: 100, 4: 400}}), pdf.replace({("X", "B"): {0: 100, 4: 400}})
+        )
 
     def test_update(self):
         # check base function