diff --git a/Lib/fontTools/ttLib/tables/otConverters.py b/Lib/fontTools/ttLib/tables/otConverters.py
index a9bc578c9..8f96af77b 100644
--- a/Lib/fontTools/ttLib/tables/otConverters.py
+++ b/Lib/fontTools/ttLib/tables/otConverters.py
@@ -571,11 +571,12 @@ class AATLookup(BaseConverter):
 
 	def write(self, writer, font, tableDict, value, repeatIndex=None):
 		glyphMap = font.getReverseGlyphMap()
-		# TODO: Also implement format 4 and 8.
+		# TODO: Also implement format 4.
 		formats = list(sorted(filter(None, [
 			self.buildFormat0(font, value),
 			self.buildFormat2(font, value),
 			self.buildFormat6(font, value),
+			self.buildFormat8(font, value),
 		])))
 		# We use the format ID as secondary sort key to make the output
 		# deterministic when multiple formats have same encoded size.
@@ -651,6 +652,24 @@ class AATLookup(BaseConverter):
 			writer.writeUShort(key)
 			writer.writeUShort(value)
 
+	def buildFormat8(self, font, values):
+		mapping = list(sorted([(font.getGlyphID(glyph), val)
+		                       for glyph, val in values.items()]))
+		minGlyphID, maxGlyphID = mapping[0][0], mapping[-1][0]
+		if len(mapping) != maxGlyphID - minGlyphID + 1:
+			return None
+		encodedValueSize = 2
+		return (4 + len(mapping) * (2 + encodedValueSize), 8,
+		        lambda writer: self.writeFormat8(writer, font, mapping))
+
+	def writeFormat8(self, writer, font, mapping):
+		firstGlyphID = mapping[0][0]
+		writer.writeUShort(8)
+		writer.writeUShort(firstGlyphID)
+		writer.writeUShort(len(mapping))
+		for _, value in mapping:
+			writer.writeUShort(font.getGlyphID(value))
+
 	def readFormat0(self, reader, numGlyphs):
 		data = reader.readUShortArray(numGlyphs)
 		return {k:v for (k,v) in enumerate(data)}
diff --git a/Tests/ttLib/tables/otConverters_test.py b/Tests/ttLib/tables/otConverters_test.py
index 2640966c8..c7796d7c2 100644
--- a/Tests/ttLib/tables/otConverters_test.py
+++ b/Tests/ttLib/tables/otConverters_test.py
@@ -307,6 +307,24 @@ class AATLookupTest(unittest.TestCase):
             "FFFF FFFF "    # entries[4]=<END>
         ))
 
+    def test_writeFormat8(self):
+        writer = OTTableWriter()
+        font = FakeFont(".notdef A B C D E F G H".split())
+        self.converter.write(writer, font, {}, {
+            "B": "B",
+            "C": "A",
+            "D": "B",
+            "E": "C",
+            "F": "B",
+            "G": "A",
+        })
+        self.assertEqual(writer.getData(), deHexStr(
+            "0008 "                          # format=8
+            "0002 "                          # firstGlyph=B
+            "0006 "                          # glyphCount=6
+            "0002 0001 0002 0003 0002 0001"  # valueArray=[B, A, B, C, B, A]
+        ))
+
     def test_xmlRead(self):
         value = self.converter.xmlRead({}, [
             ("Substitution", {"in": "A", "out": "A.alt"}, []),