diff --git a/akka-cluster/src/main/scala/akka/cluster/Cluster.scala b/akka-cluster/src/main/scala/akka/cluster/Cluster.scala
index 17842453bb..07712d8ed9 100644
--- a/akka-cluster/src/main/scala/akka/cluster/Cluster.scala
+++ b/akka-cluster/src/main/scala/akka/cluster/Cluster.scala
@@ -240,8 +240,8 @@ case class Gossip(
     val mergedMembers = Gossip.emptyMembers ++ reduceHighestPriority(this.members.toSeq, that.members.toSeq).
       filterNot(mergedUnreachable.contains)
 
-    // 5. merge seen (FIXME is this correct?)
-    val mergedSeen = this.overview.seen ++ that.overview.seen
+    // 5. fresh seen table
+    val mergedSeen = Map.empty[Address, VectorClock]
 
     Gossip(GossipOverview(mergedSeen, mergedUnreachable), mergedMembers, mergedMeta, mergedVClock)
   }
diff --git a/akka-cluster/src/test/scala/akka/cluster/GossipSpec.scala b/akka-cluster/src/test/scala/akka/cluster/GossipSpec.scala
index 985b6d5a89..8c790cf159 100644
--- a/akka-cluster/src/test/scala/akka/cluster/GossipSpec.scala
+++ b/akka-cluster/src/test/scala/akka/cluster/GossipSpec.scala
@@ -73,5 +73,17 @@ class GossipSpec extends WordSpec with MustMatchers {
 
     }
 
+    "start with fresh seen table after merge" in {
+      val g1 = Gossip(members = SortedSet(a1, b1, c1, d1)).seen(a1.address).seen(b1.address)
+      val g2 = Gossip(members = SortedSet(a2, b2, c2, d2)).seen(b2.address).seen(c2.address)
+
+      val merged1 = g1 merge g2
+      merged1.overview.seen.isEmpty must be(true)
+
+      val merged2 = g2 merge g1
+      merged2.overview.seen.isEmpty must be(true)
+
+    }
+
   }
 }