diff --git a/src/calibre/ebooks/txt/markdownml.py b/src/calibre/ebooks/txt/markdownml.py
index 2f7960e94c..116561f355 100644
--- a/src/calibre/ebooks/txt/markdownml.py
+++ b/src/calibre/ebooks/txt/markdownml.py
@@ -28,16 +28,36 @@ class MarkdownMLizer(object):
 
     def mlize_spine(self):
         output = [u'']
+        
         for item in self.oeb_book.spine:
             self.log.debug('Converting %s to Markdown formatted TXT...' % item.href)
+            
             html = unicode(etree.tostring(item.data, encoding=unicode))
+            
             if not self.opts.keep_links:
                 html = re.sub(r'<\s*a[^>]*>', '', html)
                 html = re.sub(r'<\s*/\s*a\s*>', '', html)
             if not self.opts.keep_image_references:
                 html = re.sub(r'<\s*img[^>]*>', '', html)
                 html = re.sub(r'<\s*img\s*>', '', html)
-            output += html2text(html)
+            
+            text = html2text(html)
+        
+            # Ensure the section ends with at least two new line characters.
+            # This is to prevent the last paragraph from a section being
+            # combined into the fist paragraph of the next.
+            end_chars = text[-4:]
+            # Convert all newlines to \n
+            end_chars = end_chars.replace('\r\n', '\n')
+            end_chars = end_chars.replace('\r', '\n')
+            end_chars = end_chars[-2:]
+            if not end_chars[1] == '\n':
+                text += '\n\n'
+            if end_chars[1] == '\n' and not end_chars[0] == '\n':
+                text += '\n'
+            
+            output += text
+            
         output = u''.join(output)
 
         return output