diff --git a/adam-core/src/main/scala/org/bdgenomics/adam/rdd/read/MarkDuplicates.scala b/adam-core/src/main/scala/org/bdgenomics/adam/rdd/read/MarkDuplicates.scala
index 32d6b1364d..7153ae4403 100644
--- a/adam-core/src/main/scala/org/bdgenomics/adam/rdd/read/MarkDuplicates.scala
+++ b/adam-core/src/main/scala/org/bdgenomics/adam/rdd/read/MarkDuplicates.scala
@@ -20,13 +20,10 @@ package org.bdgenomics.adam.rdd.read
 import org.bdgenomics.utils.misc.Logging
 import org.apache.spark.rdd.RDD
 import org.bdgenomics.adam.instrumentation.Timers._
-import org.bdgenomics.adam.models.{
-  RecordGroupDictionary,
-  ReferencePosition
-}
+import org.bdgenomics.adam.models.{ RecordGroupDictionary, ReferencePosition }
 import org.bdgenomics.adam.rdd.ADAMContext._
 import org.bdgenomics.adam.rdd.fragment.FragmentDataset
-import org.bdgenomics.formats.avro.{ AlignmentRecord, Fragment }
+import org.bdgenomics.formats.avro.{ AlignmentRecord, Fragment, Strand }
 
 private[rdd] object MarkDuplicates extends Serializable with Logging {
 
@@ -96,19 +93,24 @@ private[rdd] object MarkDuplicates extends Serializable with Logging {
                           recordGroups: RecordGroupDictionary): RDD[SingleReadBucket] = {
     checkRecordGroups(recordGroups)
 
+    def positionForStrand(defaultPos: Option[ReferencePosition], otherPos: Option[ReferencePosition], strand: Strand) = {
+      defaultPos.filter(_.strand == strand).orElse(otherPos.filter(_.strand == strand).orElse(defaultPos))
+    }
+
     // Group by library and left position
     def leftPositionAndLibrary(p: (ReferencePositionPair, SingleReadBucket),
                                rgd: RecordGroupDictionary): (Option[ReferencePosition], String) = {
+      val leftPosition = positionForStrand(p._1.read1refPos, p._1.read2refPos, Strand.FORWARD)
       if (p._2.allReads.head.getRecordGroupName != null) {
-        (p._1.read1refPos, rgd(p._2.allReads.head.getRecordGroupName).library.getOrElse(null))
+        (leftPosition, rgd(p._2.allReads.head.getRecordGroupName).library.orNull)
       } else {
-        (p._1.read1refPos, null)
+        (leftPosition, null)
       }
     }
 
     // Group by right position
     def rightPosition(p: (ReferencePositionPair, SingleReadBucket)): Option[ReferencePosition] = {
-      p._1.read2refPos
+      positionForStrand(p._1.read2refPos, p._1.read1refPos, Strand.REVERSE)
     }
 
     rdd.keyBy(ReferencePositionPair(_))
@@ -166,4 +168,5 @@ private[rdd] object MarkDuplicates extends Serializable with Logging {
       scoreBucket(x._2) - scoreBucket(y._2)
     }
   }
+
 }
diff --git a/adam-core/src/test/scala/org/bdgenomics/adam/rdd/read/MarkDuplicatesSuite.scala b/adam-core/src/test/scala/org/bdgenomics/adam/rdd/read/MarkDuplicatesSuite.scala
index d8db373be0..37f02e455c 100644
--- a/adam-core/src/test/scala/org/bdgenomics/adam/rdd/read/MarkDuplicatesSuite.scala
+++ b/adam-core/src/test/scala/org/bdgenomics/adam/rdd/read/MarkDuplicatesSuite.scala
@@ -309,4 +309,17 @@ class MarkDuplicatesSuite extends ADAMFunSuite {
     assert(nonDups.size == 2 && nonDups.forall(p => p.getReadName.toString == "best"))
     assert(dups.forall(p => p.getReadName.startsWith("poor")))
   }
+
+  sparkTest("inverse pairs") {
+    val firstPair = createPair("0", 100, 251, "0", 1100, 1251, "pair1")
+    val secondPair = createPair("0", 1100, 1251, "0", 100, 251, "pair2")
+    secondPair.head.setReadNegativeStrand(true)
+    secondPair.head.setMateNegativeStrand(false)
+    secondPair(1).setReadNegativeStrand(false)
+    secondPair(1).setMateNegativeStrand(true)
+
+    val marked = markDuplicateFragments(firstPair ++ secondPair: _*)
+    val (dups, nonDups) = marked.partition(_.getDuplicateRead)
+    assert(dups.size == 2)
+  }
 }