diff --git a/_pytest/skipping.py b/_pytest/skipping.py
index 6ddbb6553..aee98ffa8 100644
--- a/_pytest/skipping.py
+++ b/_pytest/skipping.py
@@ -26,11 +26,13 @@ def pytest_configure(config):
         "http://pytest.org/latest/skipping.html"
     )
     config.addinivalue_line("markers",
-        "xfail(condition, reason=None, run=True): mark the the test function "
+        "xfail(condition, reason=None, run=True, raises=None): mark the the test function "
         "as an expected failure if eval(condition) has a True value. "
         "Optionally specify a reason for better reporting and run=False if "
-        "you don't even want to execute the test function. See "
-        "http://pytest.org/latest/skipping.html"
+        "you don't even want to execute the test function. If only specific "
+        "exception(s) are expected, you can list them in raises, and if the test fails "
+        "in other ways, it will be reported as a true failure. "
+        "See http://pytest.org/latest/skipping.html"
     )
 
 def pytest_namespace():
@@ -60,6 +62,12 @@ class MarkEvaluator:
     def wasvalid(self):
         return not hasattr(self, 'exc')
 
+    def invalidraise(self, exc):
+        raises = self.get('raises')
+        if not raises:
+            return
+        return not isinstance(exc, raises)
+
     def istrue(self):
         try:
             return self._istrue()
@@ -171,7 +179,11 @@ def pytest_runtest_makereport(__multicall__, item, call):
         if not item.config.option.runxfail:
             if evalxfail.wasvalid() and evalxfail.istrue():
                 if call.excinfo:
-                    rep.outcome = "skipped"
+                    if evalxfail.invalidraise(call.excinfo.value):
+                        rep.outcome = "failed"
+                        return rep
+                    else:
+                        rep.outcome = "skipped"
                 elif call.when == "call":
                     rep.outcome = "failed"
                 else:
diff --git a/doc/en/assert.txt b/doc/en/assert.txt
index 6d9f60448..1dd7ba0c0 100644
--- a/doc/en/assert.txt
+++ b/doc/en/assert.txt
@@ -95,6 +95,22 @@ asserts that the given ``ExpectedException`` is raised.  The reporter will
 provide you with helpful output in case of failures such as *no
 exception* or *wrong exception*.
 
+Note that it is also possible to specify a "raises" argument to
+``pytest.mark.xfail``, which checks that the test is failing in a more
+specific way than just having any exception raised::
+
+    @pytest.mark.xfail(raises=IndexError)
+    def test_f():
+        f()
+
+Using ``pytest.raises`` is likely to be better for cases where you are testing
+exceptions your own code is deliberately raising, whereas using
+``@pytest.mark.xfail`` with a check function is probably better for something
+like documenting unfixed bugs (where the test describes what "should" happen)
+or bugs in dependencies.
+
+
+
 .. _newreport:
 
 Making use of context-sensitive comparisons
diff --git a/doc/en/example/xfail_demo.py b/doc/en/example/xfail_demo.py
index 75cb7bea1..c6e147e4d 100644
--- a/doc/en/example/xfail_demo.py
+++ b/doc/en/example/xfail_demo.py
@@ -23,3 +23,8 @@ def test_hello5():
 
 def test_hello6():
     pytest.xfail("reason")
+
+@xfail(raises=IndexError)
+def test_hello7()
+    x = []
+    x[1] = 1
diff --git a/doc/en/skipping.txt b/doc/en/skipping.txt
index c896d404d..ab6c838ba 100644
--- a/doc/en/skipping.txt
+++ b/doc/en/skipping.txt
@@ -149,6 +149,11 @@ on a particular platform::
     def test_function():
         ...
 
+If you want to be more specific as to why the test is failing, you can specify
+a single exception, or a list of exceptions, in the ``raises`` argument. Then
+the test will be reported as a regular failure if it fails with an
+exception not mentioned in ``raises``.
+
 You can furthermore prevent the running of an "xfail" test or
 specify a reason such as a bug ID or similar.  Here is
 a simple test file with the several usages:
diff --git a/testing/test_skipping.py b/testing/test_skipping.py
index d85a5d635..45652703a 100644
--- a/testing/test_skipping.py
+++ b/testing/test_skipping.py
@@ -330,6 +330,26 @@ class TestXFail:
             "*1 xfailed*",
         ])
 
+
+    @pytest.mark.parametrize('expected, actual, matchline',
+                             [('TypeError', 'TypeError', "*1 xfailed*"),
+                              ('(AttributeError, TypeError)', 'TypeError', "*1 xfailed*"),
+                              ('TypeError', 'IndexError', "*1 failed*"),
+                              ('(AttributeError, TypeError)', 'IndexError', "*1 failed*"),
+                              ])
+    def test_xfail_raises(self, expected, actual, matchline, testdir):
+        p = testdir.makepyfile("""
+            import pytest
+            @pytest.mark.xfail(raises=%s)
+            def test_raises():
+                raise %s()
+        """ % (expected, actual))
+        result = testdir.runpytest(p)
+        result.stdout.fnmatch_lines([
+            matchline,
+        ])
+
+
 class TestXFailwithSetupTeardown:
     def test_failing_setup_issue9(self, testdir):
         testdir.makepyfile("""
@@ -575,7 +595,7 @@ def test_default_markers(testdir):
     result = testdir.runpytest("--markers")
     result.stdout.fnmatch_lines([
         "*skipif(*condition)*skip*",
-        "*xfail(*condition, reason=None, run=True)*expected failure*",
+        "*xfail(*condition, reason=None, run=True, raises=None)*expected failure*",
     ])
 
 def test_xfail_test_setup_exception(testdir):
@@ -617,7 +637,6 @@ def test_imperativeskip_on_xfail_test(testdir):
         *2 skipped*
     """)
 
-
 class TestBooleanCondition:
     def test_skipif(self, testdir):
         testdir.makepyfile("""