diff --git a/babel/messages/extract.py b/babel/messages/extract.py
index 8fe3f606c..f8495b02b 100644
--- a/babel/messages/extract.py
+++ b/babel/messages/extract.py
@@ -461,6 +461,8 @@ def extract_javascript(fileobj, keywords, comment_tags, options):
     :param comment_tags: a list of translator tags to search for and include
                          in the results
     :param options: a dictionary of additional options (optional)
+                    Supported options are:
+                    * `jsx` -- set to false to disable JSX/E4X support.
     """
     from babel.messages.jslexer import tokenize, unquote_string
     funcname = message_lineno = None
@@ -472,7 +474,7 @@ def extract_javascript(fileobj, keywords, comment_tags, options):
     last_token = None
     call_stack = -1
 
-    for token in tokenize(fileobj.read().decode(encoding)):
+    for token in tokenize(fileobj.read().decode(encoding), jsx=options.get("jsx", True)):
         if token.type == 'operator' and token.value == '(':
             if funcname:
                 message_lineno = token.lineno
diff --git a/babel/messages/jslexer.py b/babel/messages/jslexer.py
index 22c6e1f9c..c00d8d40b 100644
--- a/babel/messages/jslexer.py
+++ b/babel/messages/jslexer.py
@@ -36,6 +36,7 @@
         ([eE][-+]?\d+)? |
         (0x[a-fA-F0-9]+)
     )''')),
+    ('jsx_tag', re.compile(r'<(?:/?)\w+.+?>', re.I)),
     ('operator', re.compile(r'(%s)' % '|'.join(map(re.escape, operators)))),
     ('string', re.compile(r'''(?xs)(
         '(?:[^'\\]*(?:\\.[^'\\]*)*)'  |
@@ -127,8 +128,11 @@ def unquote_string(string):
     return u''.join(result)
 
 
-def tokenize(source):
-    """Tokenize a JavaScript source.  Returns a generator of tokens.
+def tokenize(source, jsx=True):
+    """
+    Tokenize JavaScript/JSX source.  Returns a generator of tokens.
+
+    :param jsx: Enable (limited) JSX parsing.
     """
     may_divide = False
     pos = 0
@@ -138,6 +142,8 @@ def tokenize(source):
     while pos < end:
         # handle regular rules first
         for token_type, rule in rules:
+            if not jsx and token_type and 'jsx' in token_type:
+                continue
             match = rule.match(source, pos)
             if match is not None:
                 break
diff --git a/tests/messages/test_js_extract.py b/tests/messages/test_js_extract.py
index b355f32eb..d8c3ee806 100644
--- a/tests/messages/test_js_extract.py
+++ b/tests/messages/test_js_extract.py
@@ -1,3 +1,4 @@
+import pytest
 from babel._compat import BytesIO
 from babel.messages import extract
 
@@ -96,3 +97,27 @@ def test_misplaced_comments():
     assert messages[1][3] == [u'NOTE: this will show up', 'too.']
     assert messages[2][2] == u'no comment here'
     assert messages[2][3] == []
+
+
+JSX_SOURCE = b"""
+class Foo {
+    render() {
+        const value = gettext("hello");
+        return (
+            <option value="val1">{ i18n._('String1') }</option>
+            <option value="val2">{ i18n._('String 2') }</option>
+            <option value="val3">{ i18n._('String 3') }</option>
+        );
+    }
+"""
+EXPECTED_JSX_MESSAGES = ["hello", "String1", "String 2", "String 3"]
+
+
+@pytest.mark.parametrize("jsx_enabled", (False, True))
+def test_jsx_extraction(jsx_enabled):
+    buf = BytesIO(JSX_SOURCE)
+    messages = [m[2] for m in extract.extract_javascript(buf, ('_', 'gettext'), [], {"jsx": jsx_enabled})]
+    if jsx_enabled:
+        assert messages == EXPECTED_JSX_MESSAGES
+    else:
+        assert messages != EXPECTED_JSX_MESSAGES