diff --git a/sphinx_gallery/py_source_parser.py b/sphinx_gallery/py_source_parser.py
index d397087f9..6d072e735 100644
--- a/sphinx_gallery/py_source_parser.py
+++ b/sphinx_gallery/py_source_parser.py
@@ -47,20 +47,41 @@ def get_docstring_and_rest(filename):
     if not isinstance(node, ast.Module):
         raise TypeError("This function only supports modules. "
                         "You provided {0}".format(node.__class__.__name__))
-    if node.body and isinstance(node.body[0], ast.Expr) and \
-       isinstance(node.body[0].value, ast.Str):
-        docstring_node = node.body[0]
-        docstring = docstring_node.value.s
-        if hasattr(docstring, 'decode'):  # python2.7
-            docstring = docstring.decode('utf-8')
-        # This get the content of the file after the docstring last line
-        # Note: 'maxsplit' argument is not a keyword argument in python2
-        rest = content.decode('utf-8').split('\n', docstring_node.lineno)[-1]
-        return docstring, rest
-    else:
+    try:
+        # in python 3.7 module knows it's docstring
+        docstring = node.docstring
+        # grab the rest of the file
+        # in python 3.7 module knows it's docstring
+        docstring = node.docstring
+        # count the lines in the docstring.  We have to do it this way instead
+        # of looking at the line of the first element of the AST because there may
+        # be comments before the first code that SG needs to capture.
+        ds_lines = len(docstring.split('\n'))
+        # if there is a coding line, this will be off by one
+        if content.startswith(b'# -*- coding:'):
+            ds_lines += 1
+        # grab the rest of the file
+        rest = '\n'.join(content.decode('utf-8').split('\n')[ds_lines:])
+
+    except AttributeError:
+        # this block can be removed when python 3.6 support is dropped
+        if node.body and isinstance(node.body[0], ast.Expr) and \
+           isinstance(node.body[0].value, ast.Str):
+            docstring_node = node.body[0]
+            docstring = docstring_node.value.s
+            if hasattr(docstring, 'decode'):  # python2.7
+                docstring = docstring.decode('utf-8')
+            # This get the content of the file after the docstring last line
+            # Note: 'maxsplit' argument is not a keyword argument in python2
+            rest = content.decode('utf-8').split('\n', docstring_node.lineno)[-1]
+        else:
+            docstring, rest = '', ''
+
+    if not docstring :
         raise ValueError(('Could not find docstring in file "{0}". '
                           'A docstring is required by sphinx-gallery')
                          .format(filename))
+    return docstring, rest
 
 
 def split_code_and_text_blocks(source_file):