diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index 60fa1d80..2be3dc85 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -6,8 +6,8 @@ Changelog
 0.8.1 (2014-07-23)
 +++++++++++++++++++
 
-* Bugfix release, the last release forgot to include files that improve
-    autocompletion for builtin libraries. Fixed.
+- Bugfix release, the last release forgot to include files that improve
+  autocompletion for builtin libraries. Fixed.
 
 0.8.0 (2014-05-05)
 +++++++++++++++++++
@@ -16,26 +16,27 @@ Changelog
   drastically. Loading times are down as well (it takes basically as long as an
   import).
 - REPL completion is starting to become usable.
-- Various small API changes. Generally this released focuses on stability and
+- Various small API changes. Generally this release focuses on stability and
   refactoring of internal APIs.
-- Introducing operator precedence, which makes calculating correct Array indices
-  and ``__getattr__`` strings possible.
+- Introducing operator precedence, which makes calculating correct Array
+  indices and ``__getattr__`` strings possible.
 
 0.7.0 (2013-08-09)
 ++++++++++++++++++
 
-- Switched from LGPL to MIT license
-- Added an Interpreter class to the API to make autocompletion in REPL possible.
-- Added autocompletion support for namespace packages
-- Add sith.py, a new random testing method
+- Switched from LGPL to MIT license.
+- Added an Interpreter class to the API to make autocompletion in REPL
+  possible.
+- Added autocompletion support for namespace packages.
+- Add sith.py, a new random testing method.
 
 0.6.0 (2013-05-14)
 ++++++++++++++++++
 
-- Much faster parser with builtin part caching
-- A test suite, thanks @tkf
+- Much faster parser with builtin part caching.
+- A test suite, thanks @tkf.
 
 0.5 versions (2012)
 +++++++++++++++++++
 
-- Initial development
+- Initial development.
diff --git a/docs/docs/features.rst b/docs/docs/features.rst
index 321cbf71..f921e13a 100644
--- a/docs/docs/features.rst
+++ b/docs/docs/features.rst
@@ -119,10 +119,11 @@ http://sphinx-doc.org/domains.html#info-field-lists
 
 ::
 
-    def myfunction(node):
+    def myfunction(node, foo):
         """Do something with a ``node``.
 
         :type node: ProgramNode
+        :param str foo: foo parameter description
 
         """
         node.| # complete here
diff --git a/jedi/evaluate/docstrings.py b/jedi/evaluate/docstrings.py
index 9a789a5d..02f83ba1 100644
--- a/jedi/evaluate/docstrings.py
+++ b/jedi/evaluate/docstrings.py
@@ -14,6 +14,7 @@ As an addition to parameter searching, this module also provides return
 annotations.
 """
 
+from ast import literal_eval
 import re
 from itertools import chain
 from textwrap import dedent
@@ -24,6 +25,7 @@ from jedi.common import indent_block
 
 DOCSTRING_PARAM_PATTERNS = [
     r'\s*:type\s+%s:\s*([^\n]+)',  # Sphinx
+    r'\s*:param\s+(\w+)\s+%s:[^\n]+', # Sphinx param with type
     r'\s*@type\s+%s:\s*([^\n]+)',  # Epydoc
 ]
 
@@ -35,26 +37,44 @@ DOCSTRING_RETURN_PATTERNS = [
 REST_ROLE_PATTERN = re.compile(r':[^`]+:`([^`]+)`')
 
 
-@memoize_default(None, evaluator_is_first_arg=True)
-def follow_param(evaluator, param):
-    func = param.parent_function
-    param_str = _search_param_in_docstr(func.raw_doc, str(param.get_name()))
-    return _evaluate_for_statement_string(evaluator, param_str, param.get_parent_until())
+try:
+    from numpydoc.docscrape import NumpyDocString
+except ImportError:
+    def _search_param_in_numpydocstr(docstr, param_str):
+        return []
+else:
+    def _search_param_in_numpydocstr(docstr, param_str):
+        """Search `docstr` (in numpydoc format) for type(-s) of `param_str`."""
+        params = NumpyDocString(docstr)._parsed_data['Parameters']
+        for p_name, p_type, p_descr in params:
+            if p_name == param_str:
+                m = re.match('([^,]+(,[^,]+)*?)(,[ ]*optional)?$', p_type)
+                if m:
+                    p_type = m.group(1)
+
+                if p_type.startswith('{'):
+                    types = set(type(x).__name__ for x in literal_eval(p_type))
+                    return list(types)
+                else:
+                    return [p_type]
+        return []
 
 
 def _search_param_in_docstr(docstr, param_str):
     """
-    Search `docstr` for a type of `param_str`.
+    Search `docstr` for type(-s) of `param_str`.
 
     >>> _search_param_in_docstr(':type param: int', 'param')
-    'int'
+    ['int']
     >>> _search_param_in_docstr('@type param: int', 'param')
-    'int'
+    ['int']
     >>> _search_param_in_docstr(
     ...   ':type param: :class:`threading.Thread`', 'param')
-    'threading.Thread'
-    >>> _search_param_in_docstr('no document', 'param') is None
-    True
+    ['threading.Thread']
+    >>> bool(_search_param_in_docstr('no document', 'param'))
+    False
+    >>> _search_param_in_docstr(':param int param: some description', 'param')
+    ['int']
 
     """
     # look at #40 to see definitions of those params
@@ -63,9 +83,10 @@ def _search_param_in_docstr(docstr, param_str):
     for pattern in patterns:
         match = pattern.search(docstr)
         if match:
-            return _strip_rst_role(match.group(1))
+            return [_strip_rst_role(match.group(1))]
 
-    return None
+    return (_search_param_in_numpydocstr(docstr, param_str) or
+            [])
 
 
 def _strip_rst_role(type_str):
@@ -126,6 +147,17 @@ def _evaluate_for_statement_string(evaluator, string, module):
     return list(chain.from_iterable(it)) or definitions
 
 
+@memoize_default(None, evaluator_is_first_arg=True)
+def follow_param(evaluator, param):
+    func = param.parent_function
+
+    return [p
+            for param_str in _search_param_in_docstr(func.raw_doc,
+                                                     str(param.get_name()))
+            for p in _evaluate_for_statement_string(evaluator, param_str,
+                                                    param.get_parent_until())]
+
+
 @memoize_default(None, evaluator_is_first_arg=True)
 def find_return_types(evaluator, func):
     def search_return_in_docstr(code):
diff --git a/test/completion/docstring.py b/test/completion/docstring.py
index 1d7bff7f..f80fea31 100644
--- a/test/completion/docstring.py
+++ b/test/completion/docstring.py
@@ -3,13 +3,14 @@
 # -----------------
 # sphinx style
 # -----------------
-def f(a, b, c, d):
+def f(a, b, c, d, x):
     """ asdfasdf
     :param a: blablabla
     :type a: str
     :type b: (str, int)
     :type c: threading.Thread
     :type d: :class:`threading.Thread`
+    :param str x: blablabla
     :rtype: dict
     """
     #? str()
@@ -22,23 +23,28 @@ def f(a, b, c, d):
     c.join
     #? ['join']
     d.join
+    #? ['lower']
+    x.lower
 
 #? dict()
 f()
 
 # wrong declarations
-def f(a, b):
+def f(a, b, x):
     """
     :param a: Forgot type declaration
     :type a:
     :param b: Just something
     :type b: ``
-    :rtype: 
+    :param x: Just something without type
+    :rtype:
     """
     #? 
     a
     #? 
     b
+    #?
+    x
 
 #? 
 f()
diff --git a/test/test_evaluate/test_docstring.py b/test/test_evaluate/test_docstring.py
index daa2d331..341e2c6d 100644
--- a/test/test_evaluate/test_docstring.py
+++ b/test/test_evaluate/test_docstring.py
@@ -6,6 +6,13 @@ from textwrap import dedent
 import jedi
 from ..helpers import unittest
 
+try:
+    import numpydoc
+except ImportError:
+    numpydoc_unavailable = True
+else:
+    numpydoc_unavailable = False
+
 
 class TestDocstring(unittest.TestCase):
     def test_function_doc(self):
@@ -50,6 +57,16 @@ class TestDocstring(unittest.TestCase):
         names = [c.name for c in jedi.Script(s).completions()]
         assert 'start' in names
 
+    def test_docstrings_param_type(self):
+        s = """
+                def func(arg):
+                    '''
+                    :param str arg: some description
+                    '''
+                    arg."""
+        names = [c.name for c in jedi.Script(s).completions()]
+        assert 'join' in names
+
     def test_docstrings_type_str(self):
         s = """
                 def func(arg):
@@ -87,3 +104,49 @@ class TestDocstring(unittest.TestCase):
         assert 'a' in names
         assert '__init__' in names
         assert 'mro' not in names  # Exists only for types.
+
+    @unittest.skipIf(numpydoc_unavailable, 'numpydoc module is unavailable')
+    def test_numpydoc_docstring(self):
+        s = dedent('''
+        def foobar(x, y):
+            """
+            Parameters
+            ----------
+            x : int
+            y : str
+            """
+            y.''')
+        names = [c.name for c in jedi.Script(s).completions()]
+        assert 'isupper' in names
+        assert 'capitalize' in names
+
+    @unittest.skipIf(numpydoc_unavailable, 'numpydoc module is unavailable')
+    def test_numpydoc_docstring_set_of_values(self):
+        s = dedent('''
+        def foobar(x, y):
+            """
+            Parameters
+            ----------
+            x : {'foo', 'bar', 100500}, optional
+            """
+            x.''')
+        names = [c.name for c in jedi.Script(s).completions()]
+        assert 'isupper' in names
+        assert 'capitalize' in names
+        assert 'numerator' in names
+
+    @unittest.skipIf(numpydoc_unavailable, 'numpydoc module is unavailable')
+    def test_numpydoc_alternative_types(self):
+        s = dedent('''
+        def foobar(x, y):
+            """
+            Parameters
+            ----------
+            x : int or str or list
+            """
+            x.''')
+        names = [c.name for c in jedi.Script(s).completions()]
+        assert 'isupper' in names
+        assert 'capitalize' in names
+        assert 'numerator' in names
+        assert 'append' in names