diff --git a/lower-constraints.txt b/lower-constraints.txt
index b883ff0..590c2fa 100644
--- a/lower-constraints.txt
+++ b/lower-constraints.txt
@@ -1,6 +1,5 @@
 docutils==0.11
 dulwich==0.15.0
 PyYAML==3.10.0
-six==1.9.0
 Sphinx==1.6.1
 stestr==2.0.0
diff --git a/reno/_exts/show_reno_config.py b/reno/_exts/show_reno_config.py
index 79dfa5f..c9d79af 100644
--- a/reno/_exts/show_reno_config.py
+++ b/reno/_exts/show_reno_config.py
@@ -13,7 +13,6 @@
 from docutils import nodes
 from docutils.parsers import rst
 from docutils.statemachine import ViewList
-import six
 from sphinx.util import logging
 from sphinx.util.nodes import nested_parse_with_titles
 
@@ -37,7 +36,7 @@ def _format_option_help(options):
         for l in _multi_line_string(opt.help, '  '):
             yield l
         yield ''
-        if isinstance(opt.default, six.string_types) and '\n' in opt.default:
+        if isinstance(opt.default, str) and '\n' in opt.default:
             # Multi-line string
             yield '  Defaults to'
             yield ''
diff --git a/reno/loader.py b/reno/loader.py
index b0b2a03..ad571b2 100644
--- a/reno/loader.py
+++ b/reno/loader.py
@@ -14,7 +14,6 @@ import collections
 import logging
 import os.path
 
-import six
 import yaml
 
 from reno import scanner
@@ -107,7 +106,7 @@ class Loader(object):
 
         for section_name, section_content in content.items():
             if section_name == self._config.prelude_section_name:
-                if not isinstance(section_content, six.string_types):
+                if not isinstance(section_content, str):
                     LOG.warning(
                         ('The %s section of %s '
                          'does not parse as a single string. '
@@ -115,7 +114,7 @@ class Loader(object):
                         (self._config.prelude_section_name, filename),
                     )
             else:
-                if isinstance(section_content, six.string_types):
+                if isinstance(section_content, str):
                     # A single string is OK, but wrap it with a list
                     # so the rest of the code can treat the data model
                     # consistently.
@@ -129,7 +128,7 @@ class Loader(object):
                     )
                 else:
                     for item in section_content:
-                        if not isinstance(item, six.string_types):
+                        if not isinstance(item, str):
                             LOG.warning(
                                 ('The item %r in the %s section of %s '
                                  'parses as a %s instead of a string. '
diff --git a/reno/setup_command.py b/reno/setup_command.py
index 102ab25..42fe909 100644
--- a/reno/setup_command.py
+++ b/reno/setup_command.py
@@ -24,7 +24,6 @@ from distutils import cmd
 from distutils import errors
 from distutils import log
 
-import six
 
 from reno import cache
 from reno import config
@@ -108,7 +107,7 @@ class BuildReno(cmd.Command):
         if val is None:
             setattr(self, option, default)
             return default
-        elif not isinstance(val, six.string_types):
+        elif not isinstance(val, str):
             raise errors.DistutilsOptionError("'%s' must be a %s (got `%s`)"
                                               % (option, what, val))
         return val
diff --git a/reno/tests/test_loader.py b/reno/tests/test_loader.py
index 33b6d28..af78909 100644
--- a/reno/tests/test_loader.py
+++ b/reno/tests/test_loader.py
@@ -17,7 +17,6 @@ import textwrap
 
 import fixtures
 import mock
-import six
 import yaml
 
 from reno import config
@@ -73,7 +72,7 @@ class TestValidate(base.TestCase):
           This is a single string.
         '''))
         print(type(note_bodies['issues']))
-        self.assertIsInstance(note_bodies['issues'], six.string_types)
+        self.assertIsInstance(note_bodies['issues'], str)
         ldr = self._make_loader(note_bodies)
         parse_results = ldr.parse_note_file('note1', None)
         self.assertIsInstance(parse_results['issues'], list)
diff --git a/reno/tests/test_utils.py b/reno/tests/test_utils.py
index cb76cc7..8b4bc48 100644
--- a/reno/tests/test_utils.py
+++ b/reno/tests/test_utils.py
@@ -13,7 +13,6 @@
 # under the License.
 
 import mock
-import six
 
 from reno.tests import base
 from reno import utils
@@ -28,7 +27,7 @@ class TestGetRandomString(base.TestCase):
         randrange.return_value = ord('a')
         actual = utils.get_random_string()
         expected = '61' * 8  # hex for ord('a')
-        self.assertIsInstance(actual, six.text_type)
+        self.assertIsInstance(actual, str)
         self.assertEqual(expected, actual)
 
     @mock.patch('random.randrange')
@@ -38,5 +37,5 @@ class TestGetRandomString(base.TestCase):
         randrange.return_value = ord('a')
         actual = utils.get_random_string()
         expected = '62' * 8  # hex for ord('b')
-        self.assertIsInstance(actual, six.text_type)
+        self.assertIsInstance(actual, str)
         self.assertEqual(expected, actual)
diff --git a/reno/utils.py b/reno/utils.py
index 88cd098..12aa327 100644
--- a/reno/utils.py
+++ b/reno/utils.py
@@ -23,11 +23,11 @@ LOG = logging.getLogger(__name__)
 def get_random_string(nbytes=8):
     """Return a fixed-length random string
 
-    :rtype: six.text_type
+    :rtype: str
     """
     try:
         # NOTE(dhellmann): Not all systems support urandom().
-        # hexlify returns six.binary_type, decode to convert to six.text_type.
+        # hexlify returns binary, decode to convert to str.
         val = binascii.hexlify(os.urandom(nbytes)).decode('utf-8')
     except Exception as e:
         print('ERROR, perhaps urandom is not supported: %s' % e)
diff --git a/requirements.txt b/requirements.txt
index 079e796..b97c1c8 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -4,5 +4,4 @@
 
 pbr
 PyYAML>=3.10
-six>=1.9.0
 dulwich>=0.15.0  # Apache-2.0