diff --git a/Lib/test/test_stringprep.py b/Lib/test/test_stringprep.py
new file mode 100644
index 000000000..d4b4a13d0
--- /dev/null
+++ b/Lib/test/test_stringprep.py
@@ -0,0 +1,92 @@
+# To fully test this module, we would need a copy of the stringprep tables.
+# Since we don't have them, this test checks only a few code points.
+
+import unittest
+
+from stringprep import *
+
+class StringprepTests(unittest.TestCase):
+    def test(self):
+        self.assertTrue(in_table_a1("\u0221"))
+        self.assertFalse(in_table_a1("\u0222"))
+
+        self.assertTrue(in_table_b1("\u00ad"))
+        self.assertFalse(in_table_b1("\u00ae"))
+
+        self.assertTrue(map_table_b2("\u0041"), "\u0061")
+        self.assertTrue(map_table_b2("\u0061"), "\u0061")
+
+        self.assertTrue(map_table_b3("\u0041"), "\u0061")
+        self.assertTrue(map_table_b3("\u0061"), "\u0061")
+
+        self.assertTrue(in_table_c11("\u0020"))
+        self.assertFalse(in_table_c11("\u0021"))
+
+        self.assertTrue(in_table_c12("\u00a0"))
+        self.assertFalse(in_table_c12("\u00a1"))
+
+        self.assertTrue(in_table_c12("\u00a0"))
+        self.assertFalse(in_table_c12("\u00a1"))
+
+        self.assertTrue(in_table_c11_c12("\u00a0"))
+        self.assertFalse(in_table_c11_c12("\u00a1"))
+
+        self.assertTrue(in_table_c21("\u001f"))
+        self.assertFalse(in_table_c21("\u0020"))
+
+        self.assertTrue(in_table_c22("\u009f"))
+        self.assertFalse(in_table_c22("\u00a0"))
+
+        self.assertTrue(in_table_c21_c22("\u009f"))
+        self.assertFalse(in_table_c21_c22("\u00a0"))
+
+        self.assertTrue(in_table_c3("\ue000"))
+        self.assertFalse(in_table_c3("\uf900"))
+
+        self.assertTrue(in_table_c4("\uffff"))
+        self.assertFalse(in_table_c4("\u0000"))
+
+        self.assertTrue(in_table_c5("\ud800"))
+        self.assertFalse(in_table_c5("\ud7ff"))
+
+        self.assertTrue(in_table_c6("\ufff9"))
+        self.assertFalse(in_table_c6("\ufffe"))
+
+        self.assertTrue(in_table_c7("\u2ff0"))
+        self.assertFalse(in_table_c7("\u2ffc"))
+
+        self.assertTrue(in_table_c8("\u0340"))
+        self.assertFalse(in_table_c8("\u0342"))
+
+        # C.9 is not in the bmp
+        # self.assertTrue(in_table_c9(u"\U000E0001"))
+        # self.assertFalse(in_table_c8(u"\U000E0002"))
+
+        self.assertTrue(in_table_d1("\u05be"))
+        self.assertFalse(in_table_d1("\u05bf"))
+
+        self.assertTrue(in_table_d2("\u0041"))
+        self.assertFalse(in_table_d2("\u0040"))
+
+        # This would generate a hash of all predicates. However, running
+        # it is quite expensive, and only serves to detect changes in the
+        # unicode database. Instead, stringprep.py asserts the version of
+        # the database.
+
+        # import hashlib
+        # predicates = [k for k in dir(stringprep) if k.startswith("in_table")]
+        # predicates.sort()
+        # for p in predicates:
+        #     f = getattr(stringprep, p)
+        #     # Collect all BMP code points
+        #     data = ["0"] * 0x10000
+        #     for i in range(0x10000):
+        #         if f(unichr(i)):
+        #             data[i] = "1"
+        #     data = "".join(data)
+        #     h = hashlib.sha1()
+        #     h.update(data)
+        #     print p, h.hexdigest()
+
+if __name__ == '__main__':
+    unittest.main()