diff --git a/djangorestframework/compat.py b/djangorestframework/compat.py
index b7eedf85c..ab0d80f3c 100644
--- a/djangorestframework/compat.py
+++ b/djangorestframework/compat.py
@@ -374,7 +374,35 @@ else:
 # Markdown is optional
 try:
     import markdown
-    
+     
+    class CustomSetextHeaderProcessor(markdown.blockprocessors.BlockProcessor):
+        """
+        Class for markdown < 2.1
+
+        Override `markdown`'s :class:`SetextHeaderProcessor`, so that ==== headers are <h2> and ---- heade
+
+        We use <h1> for the resource name.
+        """
+        import re
+        # Detect Setext-style header. Must be first 2 lines of block.
+        RE = re.compile(r'^.*?\n[=-]{3,}', re.MULTILINE)
+
+        def test(self, parent, block):
+            return bool(self.RE.match(block))
+
+        def run(self, parent, blocks):
+            lines = blocks.pop(0).split('\n')
+            # Determine level. ``=`` is 1 and ``-`` is 2.
+            if lines[1].startswith('='):
+                level = 2
+            else:
+                level = 3
+            h = markdown.etree.SubElement(parent, 'h%d' % level)
+            h.text = lines[0].strip()
+            if len(lines) > 2:
+                # Block contains additional lines. Add to  master blocks for later.
+                blocks.insert(0, '\n'.join(lines[2:]))
+
     def apply_markdown(text):
         """
         Simple wrapper around :func:`markdown.markdown` to set the base level
@@ -384,7 +412,15 @@ try:
         extensions = ['headerid(level=2)']
         safe_mode = False,
 
-        md = markdown.Markdown(extensions=extensions, safe_mode=safe_mode)
+        if markdown.version < (2, 1):
+            output_format = markdown.DEFAULT_OUTPUT_FORMAT
+
+            md = markdown.Markdown(extensions=markdown.load_extensions(extensions),
+                               safe_mode=safe_mode,
+                               output_format=output_format)
+            md.parser.blockprocessors['setextheader'] = CustomSetextHeaderProcessor(md.parser)
+        else:
+            md = markdown.Markdown(extensions=extensions, safe_mode=safe_mode)
         return md.convert(text)
 
 except ImportError:
@@ -395,3 +431,4 @@ try:
     import yaml
 except ImportError:
     yaml = None
+
diff --git a/djangorestframework/tests/description.py b/djangorestframework/tests/description.py
index 17bb43306..c0424c24a 100644
--- a/djangorestframework/tests/description.py
+++ b/djangorestframework/tests/description.py
@@ -19,8 +19,22 @@ indented
 
 # hash style header #"""
 
-# If markdown is installed we also test it's working (and that our wrapped forces '=' to h2 and '-' to h3)
-MARKED_DOWN = """<h2 id="an-example-docstring">an example docstring</h2>
+# If markdown is installed we also test it's working 
+# (and that our wrapped forces '=' to h2 and '-' to h3)
+
+# We support markdown < 2.1 and markdown >= 2.1
+MARKED_DOWN_lt_21 = """<h2>an example docstring</h2>
+<ul>
+<li>list</li>
+<li>list</li>
+</ul>
+<h3>another header</h3>
+<pre><code>code block
+</code></pre>
+<p>indented</p>
+<h2 id="hash_style_header">hash style header</h2>"""
+
+MARKED_DOWN_gte_21 = """<h2 id="an-example-docstring">an example docstring</h2>
 <ul>
 <li>list</li>
 <li>list</li>
@@ -92,4 +106,6 @@ class TestViewNamesAndDescriptions(TestCase):
     def test_markdown(self):
         """Ensure markdown to HTML works as expected"""
         if apply_markdown:
-            self.assertEquals(apply_markdown(DESCRIPTION), MARKED_DOWN)
+            gte_21_match = apply_markdown(DESCRIPTION) == MARKED_DOWN_gte_21
+            lt_21_match = apply_markdown(DESCRIPTION) == MARKED_DOWN_lt_21
+            self.assertTrue(gte_21_match or lt_21_match)