diff --git a/python/pyspark/sql/dataframe.py b/python/pyspark/sql/dataframe.py
index 09647ff6d074..da742d7ce7d1 100644
--- a/python/pyspark/sql/dataframe.py
+++ b/python/pyspark/sql/dataframe.py
@@ -722,8 +722,6 @@ def __getitem__(self, item):
         [Row(age=5, name=u'Bob')]
         """
         if isinstance(item, basestring):
-            if item not in self.columns:
-                raise IndexError("no such column: %s" % item)
             jc = self._jdf.apply(item)
             return Column(jc)
         elif isinstance(item, Column):
diff --git a/python/pyspark/sql/tests.py b/python/pyspark/sql/tests.py
index 38c83c427a74..6a09b8e29748 100644
--- a/python/pyspark/sql/tests.py
+++ b/python/pyspark/sql/tests.py
@@ -770,7 +770,7 @@ def test_access_column(self):
         self.assertTrue(isinstance(df['key'], Column))
         self.assertTrue(isinstance(df[0], Column))
         self.assertRaises(IndexError, lambda: df[2])
-        self.assertRaises(IndexError, lambda: df["bad_key"])
+        self.assertRaises(AnalysisException, lambda: df["bad_key"])
         self.assertRaises(TypeError, lambda: df[{}])
 
     def test_column_name_with_non_ascii(self):
@@ -794,7 +794,9 @@ def test_field_accessor(self):
         df = self.sc.parallelize([Row(l=[1], r=Row(a=1, b="b"), d={"k": "v"})]).toDF()
         self.assertEqual(1, df.select(df.l[0]).first()[0])
         self.assertEqual(1, df.select(df.r["a"]).first()[0])
+        self.assertEqual(1, df.select(df["r.a"]).first()[0])
         self.assertEqual("b", df.select(df.r["b"]).first()[0])
+        self.assertEqual("b", df.select(df["r.b"]).first()[0])
         self.assertEqual("v", df.select(df.d["k"]).first()[0])
 
     def test_infer_long_type(self):
diff --git a/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala b/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala
index c466d9e6cb34..ec5084af1460 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala
@@ -634,6 +634,7 @@ class DataFrame private[sql](
 
   /**
    * Selects column based on the column name and return it as a [[Column]].
+   * Note that the column name can also reference to a nested column like `a.b`.
    * @group dfops
    * @since 1.3.0
    */
@@ -641,6 +642,7 @@ class DataFrame private[sql](
 
   /**
    * Selects column based on the column name and return it as a [[Column]].
+   * Note that the column name can also reference to a nested column like `a.b`.
    * @group dfops
    * @since 1.3.0
    */