diff --git a/spacy/ja/__init__.py b/spacy/ja/__init__.py
index 2f85406c0..26e39a593 100644
--- a/spacy/ja/__init__.py
+++ b/spacy/ja/__init__.py
@@ -16,6 +16,8 @@ from collections import namedtuple
 
 ShortUnitWord = namedtuple('ShortUnitWord', ['surface', 'base_form', 'part_of_speech'])
 
+DETAILS_KEY = 'mecab_details'
+
 def try_mecab_import():
     """Mecab is required for Japanese support, so check for it.
 
@@ -34,8 +36,12 @@ class JapaneseTokenizer(object):
         self.tokenizer = MeCab.Tagger()
 
     def __call__(self, text):
-        words = [x.surface for x in detailed_tokens(self.tokenizer, text)]
-        return Doc(self.vocab, words=words, spaces=[False]*len(words))
+        dtokens = detailed_tokens(self.tokenizer, text)
+        words = [x.surface for x in dtokens]
+        doc = Doc(self.vocab, words=words, spaces=[False]*len(words))
+        # stash details tokens for tagger to use
+        doc.user_data[DETAILS_KEY] = dtokens
+        return doc
 
 def resolve_pos(token):
     """If necessary, add a field to the POS tag for UD mapping.
@@ -91,7 +97,7 @@ class JapaneseTagger(object):
         # 1. get raw JP tags
         # 2. add features to tags as necessary for UD
 
-        dtokens = detailed_tokens(self.tokenizer, tokens.text)
+        dtokens = tokens.user_data[DETAILS_KEY]
         rawtags = list(map(resolve_pos, dtokens))
         self.tagger.tag_from_strings(tokens, rawtags)
 
@@ -112,8 +118,7 @@ class Japanese(Language):
     Defaults = JapaneseDefaults
 
     def make_doc(self, text):
-        words = [str(t) for t in self.tokenizer(text)]
-        doc = Doc(self.vocab, words=words, spaces=[False]*len(words))
+        jdoc = self.tokenizer(text)
         tagger = JapaneseDefaults.create_tagger(self.tokenizer)
-        tagger(doc)
-        return doc
+        tagger(jdoc)
+        return jdoc