diff --git a/layout/generic/nsFrame.cpp b/layout/generic/nsFrame.cpp
index 3ba1093ae0eb..cea60a18bae9 100644
--- a/layout/generic/nsFrame.cpp
+++ b/layout/generic/nsFrame.cpp
@@ -1833,8 +1833,14 @@ nsFrame::HandlePress(nsPresContext* aPresContext,
   // NS_STYLE_USER_SELECT_TOGGLE, need to change this logic
   PRBool useFrameSelection = (selectStyle == NS_STYLE_USER_SELECT_TEXT);
 
-  if (!IsMouseCaptured(aPresContext))
-    CaptureMouse(aPresContext, PR_TRUE);
+  nsPoint pt = nsLayoutUtils::GetEventCoordinatesRelativeTo(aEvent, this);
+  ContentOffsets offsets = GetContentOffsetsFromPoint(pt);
+
+  if (!IsMouseCaptured(aPresContext) && offsets.content) {
+    nsIFrame* capturingFrame = shell->GetPrimaryFrameFor(offsets.content);
+    if (capturingFrame)
+      capturingFrame->CaptureMouse(aPresContext, PR_TRUE);
+  }
 
   // XXX This is screwy; it really should use the selection frame, not the
   // event frame
@@ -1863,9 +1869,6 @@ nsFrame::HandlePress(nsPresContext* aPresContext,
     return HandleMultiplePress(aPresContext, aEvent, aEventStatus);
   }
 
-  nsPoint pt = nsLayoutUtils::GetEventCoordinatesRelativeTo(aEvent, this);
-  ContentOffsets offsets = GetContentOffsetsFromPoint(pt);
-
   if (!offsets.content)
     return NS_ERROR_FAILURE;