diff --git a/plugins/BadWords/config.py b/plugins/BadWords/config.py
index e66ea4602..8a3e9c225 100644
--- a/plugins/BadWords/config.py
+++ b/plugins/BadWords/config.py
@@ -44,15 +44,27 @@ def configure(advanced):
                          'spaces)'))
         conf.supybot.plugins.BadWords.words.set(words)
 
-class LastModifiedSetOfStrings(registry.SpaceSeparatedSetOfStrings):
+class LastModifiedSpaceSeparatedSetOfStrings(registry.SpaceSeparatedSetOfStrings):
     lastModified = 0
     def setValue(self, v):
         self.lastModified = time.time()
         registry.SpaceSeparatedListOfStrings.setValue(self, v)
 
+class LastModifiedCommaSeparatedSetOfStrings(registry.CommaSeparatedSetOfStrings):
+    lastModified = 0
+    def set(self, v):
+        if not v.strip():
+            self.setValue(set())
+        else:
+            super().set(v)
+
+    def setValue(self, v):
+        self.lastModified = time.time()
+        registry.CommaSeparatedListOfStrings.setValue(self, v)
+
 BadWords = conf.registerPlugin('BadWords')
 conf.registerGlobalValue(BadWords, 'words',
-    LastModifiedSetOfStrings([], _("""Determines what words are
+    LastModifiedSpaceSeparatedSetOfStrings([], _("""Determines what words are
     considered to be 'bad' so the bot won't say them.""")))
 conf.registerChannelValue(BadWords,'requireWordBoundaries',
     registry.Boolean(False, _("""Determines whether the bot will require bad
@@ -62,6 +74,9 @@ conf.registerChannelValue(BadWords,'requireWordBoundaries',
     false. After changing this setting, the BadWords regexp needs to be
     regenerated by adding/removing a word to the list, or reloading the
     plugin.""")))
+conf.registerGlobalValue(BadWords, 'phrases',
+    LastModifiedCommaSeparatedSetOfStrings([], _("""Comma-separated groups
+    of words that are considered to be 'bad'.""")))
 
 class String256(registry.String):
     def __call__(self):
diff --git a/plugins/BadWords/plugin.py b/plugins/BadWords/plugin.py
index cffaedccd..d19871927 100644
--- a/plugins/BadWords/plugin.py
+++ b/plugins/BadWords/plugin.py
@@ -53,6 +53,7 @@ class BadWords(callbacks.Privmsg):
         self.filtering = True
         self.lastModified = 0
         self.words = conf.supybot.plugins.BadWords.words
+        self.phrases = conf.supybot.plugins.BadWords.phrases
 
     def callCommand(self, name, irc, msg, *args, **kwargs):
         if ircdb.checkCapability(msg.prefix, 'admin'):
@@ -71,7 +72,7 @@ class BadWords(callbacks.Privmsg):
         self.filtering = True
         # We need to check for bad words here rather than in doPrivmsg because
         # messages don't get to doPrivmsg if the user is ignored.
-        if msg.command == 'PRIVMSG' and self.words():
+        if msg.command == 'PRIVMSG' and (self.words() or self.phrases()):
             channel = msg.channel
             self.updateRegexp(channel, irc.network)
             s = ircutils.stripFormatting(msg.args[1])
@@ -96,12 +97,14 @@ class BadWords(callbacks.Privmsg):
         return msg
 
     def updateRegexp(self, channel, network):
-        if self.lastModified < self.words.lastModified:
-            self.makeRegexp(self.words(), channel, network)
+        if self.lastModified < self.words.lastModified \
+                or self.lastModified < self.phrases.lastModified:
+            self.makeRegexp(self.words() | self.phrases(), channel, network)
             self.lastModified = time.time()
 
     def outFilter(self, irc, msg):
-        if self.filtering and msg.command == 'PRIVMSG' and self.words():
+        if self.filtering and msg.command == 'PRIVMSG' \
+                and (self.words() or self.phrases()):
             channel = msg.channel
             self.updateRegexp(channel, irc.network)
             s = msg.args[1]
@@ -124,7 +127,7 @@ class BadWords(callbacks.Privmsg):
 
         Returns the list of words being censored.
         """
-        L = list(self.words())
+        L = list(self.words() | self.phrases())
         if L:
             self.filtering = False
             utils.sortBy(str.lower, L)
@@ -134,29 +137,42 @@ class BadWords(callbacks.Privmsg):
     list = wrap(list, ['admin'])
 
     @internationalizeDocstring
-    def add(self, irc, msg, args, words):
+    def add(self, irc, msg, args, new_words):
         """<word> [<word> ...]
 
         Adds all <word>s to the list of words being censored.
         """
-        set = self.words()
-        set.update(words)
-        self.words.setValue(set)
+        words = self.words()
+        phrases = self.phrases()
+        for word in new_words:
+            if ' ' in word:
+                phrases.add(word)
+            else:
+                words.add(word)
+
+        self.words.setValue(words)
+        self.phrases.setValue(phrases)
+
         irc.replySuccess()
-    add = wrap(add, ['admin', many('somethingWithoutSpaces')])
+    add = wrap(add, ['admin', many('something')])
 
     @internationalizeDocstring
-    def remove(self, irc, msg, args, words):
+    def remove(self, irc, msg, args, old_words):
         """<word> [<word> ...]
 
         Removes <word>s from the list of words being censored.
         """
-        set = self.words()
-        for word in words:
-            set.discard(word)
-        self.words.setValue(set)
+        words = self.words()
+        phrases = self.phrases()
+        for word in old_words:
+            words.discard(word)
+            phrases.discard(word)
+        self.words.setValue(words)
+        self.phrases.setValue(phrases)
+
+
         irc.replySuccess()
-    remove = wrap(remove, ['admin', many('somethingWithoutSpaces')])
+    remove = wrap(remove, ['admin', many('something')])
 
 
 Class = BadWords
diff --git a/plugins/BadWords/test.py b/plugins/BadWords/test.py
index 06efc0d3a..753842877 100644
--- a/plugins/BadWords/test.py
+++ b/plugins/BadWords/test.py
@@ -76,7 +76,8 @@ class BadWordsTestCase(PluginTestCase):
         self.assertNotError('badwords list')
         self.assertNotError('badwords add shit')
         self.assertNotError('badwords add ass')
-        self.assertResponse('badwords list', 'ass and shit')
+        self.assertNotError('badwords add "fuck you"')
+        self.assertResponse('badwords list', 'ass, fuck you, and shit')
 
 # vim:set shiftwidth=4 softtabstop=4 expandtab textwidth=79: