diff --git a/src/tokenized-line.coffee b/src/tokenized-line.coffee
index 45af81e57..8b4c40ff1 100644
--- a/src/tokenized-line.coffee
+++ b/src/tokenized-line.coffee
@@ -41,6 +41,9 @@ class TokenizedLine
     firstNonWhitespaceColumn = null
     lastNonWhitespaceColumn = null
 
+    substringStart = 0
+    substringEnd = 0
+
     while bufferColumn < @text.length
       # advance to next token if we've iterated over its length
       if tokenOffset is @tags[tokenIndex]
@@ -66,7 +69,7 @@ class TokenizedLine
         firstNonWhitespaceColumn ?= screenColumn
         lastNonWhitespaceColumn = screenColumn + 1
 
-        text += @text.substr(bufferColumn, 2)
+        substringEnd += 2
         screenColumn += 2
         bufferColumn += 2
 
@@ -78,14 +81,21 @@ class TokenizedLine
       # split out leading soft tabs
       else if character is ' '
         if firstNonWhitespaceColumn?
-          text += ' '
+          substringEnd += 1
         else
           if (screenColumn + 1) % @tabLength is 0
             @specialTokens[tokenIndex] = SoftTab
             suffix = @tags[tokenIndex] - @tabLength
             @tags.splice(tokenIndex, 1, @tabLength)
             @tags.splice(tokenIndex + 1, 0, suffix) if suffix > 0
-          text += @invisibles?.space ? ' '
+
+          if @invisibles?.space
+            text += @text.substring(substringStart, substringEnd) if substringEnd > substringStart
+            substringStart = substringEnd
+            text += @invisibles.space
+            substringStart += 1
+
+          substringEnd += 1
 
         screenColumn++
         bufferColumn++
@@ -93,6 +103,9 @@ class TokenizedLine
 
       # expand hard tabs to the next tab stop
       else if character is '\t'
+        text += @text.substring(substringStart, substringEnd) if substringEnd > substringStart
+        substringStart = substringEnd
+
         tabLength = @tabLength - (screenColumn % @tabLength)
         if @invisibles?.tab
           text += @invisibles.tab
@@ -100,6 +113,9 @@ class TokenizedLine
           text += ' '
         text += ' ' for i in [1...tabLength] by 1
 
+        substringStart += 1
+        substringEnd += 1
+
         prefix = tokenOffset
         suffix = @tags[tokenIndex] - tokenOffset - 1
         splitTokens = []
@@ -122,12 +138,17 @@ class TokenizedLine
         firstNonWhitespaceColumn ?= screenColumn
         lastNonWhitespaceColumn = screenColumn
 
-        text += character
+        substringEnd += 1
         screenColumn++
         bufferColumn++
         tokenOffset++
 
-    @text = text
+    if substringEnd > substringStart
+      unless substringStart is 0 and substringEnd is @text.length
+        text += @text.substring(substringStart, substringEnd)
+        @text = text
+    else
+      @text = text
 
     @firstNonWhitespaceIndex = firstNonWhitespaceColumn
     if lastNonWhitespaceColumn?