diff --git a/scrapy/selector/lxmlsel.py b/scrapy/selector/lxmlsel.py
index 301858890..bb0a4e1ec 100644
--- a/scrapy/selector/lxmlsel.py
+++ b/scrapy/selector/lxmlsel.py
@@ -68,7 +68,12 @@ class XPathSelector(object_ref):
             return etree.tostring(self.root, method=self._tostring_method, \
                 encoding=unicode)
         except (AttributeError, TypeError):
-            return unicode(self.root)
+            if self.root is True:
+                return u'1'
+            elif self.root is False:
+                return u'0'
+            else:
+                return unicode(self.root)
 
     def register_namespace(self, prefix, uri):
         if self.namespaces is None:
diff --git a/scrapy/tests/test_selector.py b/scrapy/tests/test_selector.py
index 20672098f..e4197b084 100644
--- a/scrapy/tests/test_selector.py
+++ b/scrapy/tests/test_selector.py
@@ -63,12 +63,8 @@ class XPathSelectorTestCase(unittest.TestCase):
         body = "<p><input name='a'value='1'/><input name='b'value='2'/></p>"
         response = TextResponse(url="http://example.com", body=body)
         xs = self.hxs_cls(response)
-        true = xs.select("//input[@name='a']/@name='a'").extract()[0]
-        false = xs.select("//input[@name='a']/@name='n'").extract()[0]
-
-        # the actual result depends on the backend used
-        assert true in [u'1', u'True'], true
-        assert false in [u'0', u'False'], false
+        self.assertEquals(xs.select("//input[@name='a']/@name='a'").extract(), [u'1'])
+        self.assertEquals(xs.select("//input[@name='a']/@name='n'").extract(), [u'0'])
 
     @libxml2debug
     def test_selector_xml_html(self):