diff --git a/furl/furl.py b/furl/furl.py
index cd6d710..67c275c 100644
--- a/furl/furl.py
+++ b/furl/furl.py
@@ -223,7 +223,7 @@ def is_valid_encoded_query_value(value):
     return is_valid_encoded_query_value.regex.match(value) is not None
 
 
-@static_vars(regex=re.compile(r'[a-zA-Z][a-zA-Z\-\.\+]*'))
+@static_vars(regex=re.compile(r'^[a-zA-Z][a-zA-Z0-9\-\.\+]*$'))
 def is_valid_scheme(scheme):
     return is_valid_scheme.regex.match(scheme) is not None
 
diff --git a/tests/test_furl.py b/tests/test_furl.py
index bc268c8..8c1167c 100644
--- a/tests/test_furl.py
+++ b/tests/test_furl.py
@@ -1290,6 +1290,17 @@ def test_scheme(self):
         assert furl.furl('+invalid$scheme://lolsup').scheme is None
         assert furl.furl('/api/test?url=http://a.com').scheme is None
 
+        # A prefix that looks like a scheme isn't one: the scheme regex must be
+        # anchored at both ends, so a ':' after invalid scheme characters
+        # doesn't get mis-parsed as a scheme separator.
+        f = furl.furl('a b:c')
+        assert f.scheme is None and str(f.path) == 'a%20b:c'
+        assert furl.furl('foo bar:baz').scheme is None
+
+        # Digits are valid in schemes (RFC 3986) and must still parse.
+        assert furl.furl('ws2://host').scheme == 'ws2'
+        assert furl.furl('a1b://host').scheme == 'a1b'
+
         # Empty scheme.
         f = furl.furl(':')
         assert f.scheme == '' and f.netloc is None and f.url == ':'