diff --git a/docdiff.rb b/docdiff.rb
old mode 100644
new mode 100755
diff --git a/docdiff/charstring.rb b/docdiff/charstring.rb
index 1673f84..9beac1e 100644
--- a/docdiff/charstring.rb
+++ b/docdiff/charstring.rb
@@ -13,9 +13,20 @@ module CharString
 #    @encoding = CharString.guess_encoding(string)
 #    @eol     = CharString.guess_eol(string)
 =end unnecessary
+    @fullstring = string
     super
   end
 
+  # Fullstring includes non-graphical characters like newlines and spaces.
+  # These are (optionally) kept separate, because users comparing documents
+  # don't care about whitespace changes.
+  attr_reader :fullstring
+  attr_writer :fullstring
+
+  def to_s()
+    @fullstring
+  end
+
   def encoding()
     @encoding
 #     if @encoding
@@ -289,7 +300,8 @@ module CharString
     count_latin_blank_char() + count_ja_blank_char()
   end
 
-  def split_to_word()
+  # splits a string into words, treating whitespaces as words.
+  def split_to_word_verbose()
     raise "Encodings[encoding] is #{Encodings[encoding].inspect}: encoding not specified or auto-detection failed." unless Encodings[encoding]
     # raise "EOLChars[eol] is #{EOLChars[eol].inspect}: eol not specified or auto-detection failed." unless EOLChars[eol]
     scan(Regexp.new(Encodings[encoding]::WORD_REGEXP_SRC, 
@@ -298,6 +310,21 @@ module CharString
     )
   end
 
+  # splits a string into words, storing whitespaces in an attribute of
+  # the preceeding word.
+  def split_to_word()
+    words = split_to_word_verbose()
+    word_regex = Regexp.new(Encodings[encoding]::PURE_WORD_REGEXP_SRC,
+                    Regexp::MULTILINE, 
+                    encoding.sub(/ASCII/i, 'none'))
+    result = words.collect { |word|
+      word_text = word.scan(word_regex).select{|x| x != ""}[0]
+      word_text.extend CharString
+      word_text.fullstring = word
+      word_text
+    }
+  end
+
   def count_word()
     split_to_word().size
   end
diff --git a/docdiff/encoding/en_ascii.rb b/docdiff/encoding/en_ascii.rb
index 6bab6ae..96f760e 100644
--- a/docdiff/encoding/en_ascii.rb
+++ b/docdiff/encoding/en_ascii.rb
@@ -67,9 +67,10 @@ module CharString
     PRINT.replace(Regexp.quote(PRINT)) # kludge to avoid warning "character class has `[' without escape"
     GRAPH.replace(Regexp.quote(GRAPH)) # kludge to avoid warning "character class has `[' without escape"
 
-    WORD_REGEXP_SRC = ["(?:[#{GRAPH}]+[#{BLANK}]?)", 
-                       "|(?:[#{SPACE}]+)", 
-                       "|(?:.+?)"].join
+    WORD_REGEXP_SRC = ["(?:[#{PUNCT}]+[#{SPACE}]*)",
+                       "|(?:[#{ALNUM}]+[#{SPACE}]*)",
+                       "|(?:[^#{GRAPH}]+)"].join
+    PURE_WORD_REGEXP_SRC = "[#{GRAPH}]+"
 
     # override default method, as ASCII has no Japanese in it
     def count_ja_graph_char()
diff --git a/docdiff/view.rb b/docdiff/view.rb
index 6535090..e15282b 100644
--- a/docdiff/view.rb
+++ b/docdiff/view.rb
@@ -50,20 +50,27 @@ class View
     str.gsub(tags[:outside_escape_pat]){|m| tags[:outside_escape_dic][m]}
   end
 
+  def format_block(block)
+    if block == nil
+        ""
+    else
+        block.collect {|x| x.to_s}.to_s
+    end
+  end
+
   def apply_style(tags, headfoot = true)
     result = []
     @difference.each{|block|
       operation = block.first
+      source = format_block(block[1])
+      target = format_block(block[2])
       if block_given?
-        source = yield block[1].to_s
-        target = yield block[2].to_s
-      else
-        source = block[1].to_s
-        target = block[2].to_s
+        source = yield source
+        target = yield target
       end
       case operation
       when :common_elt_elt
-        result << (tags[:start_common] + escape_outside(source, tags) + tags[:end_common])
+        result << (tags[:start_common] + escape_outside(target, tags) + tags[:end_common])
       when :change_elt
         result << (tags[:start_before_change] + escape_inside(source, tags) + tags[:end_before_change] +
                    tags[:start_after_change] + escape_inside(target, tags) + tags[:end_after_change])