diff --git a/src/register/splitreg.c b/src/register/splitreg.c
index 01a1b0a65a..10abf8b96b 100644
--- a/src/register/splitreg.c
+++ b/src/register/splitreg.c
@@ -664,11 +664,9 @@ xaccConfigSplitRegister (SplitRegister *reg, int newtype)
 {
    if (!reg) return;
 
-   /* before reconfiguring, clean up any messy state that may be lying about */
-   xaccVerifyCursorPosition (reg->table, -1, -1);
-
    reg->type = newtype;
    configLayout (reg);
+   configTraverse (reg);
    configCursors (reg);
 }
 
diff --git a/src/register/table-motif.c b/src/register/table-motif.c
index dda847406b..12d95e78c9 100644
--- a/src/register/table-motif.c
+++ b/src/register/table-motif.c
@@ -447,6 +447,22 @@ traverseCB (Widget mw, XtPointer cd, XtPointer cb)
       }
    }
 
+   /* Don't do a thing unless we verify that the row and column
+    * are in bounds. Ordinarily, they are always in bounds, except 
+    * in an unusual, arguably buggy situation: If the table has 
+    * been recently resized smaller, then the Xbae code might report
+    * a traverse out of a cell that was in the larger array, but not
+    * in the smaller array.  This is probably an Xbae bug. It 
+    * will core dump array access.
+    */
+   if ((row >= table->num_phys_rows) || 
+       (col >= table->num_phys_cols)) {
+
+      table->prev_phys_traverse_row = cbs->next_row;
+      table->prev_phys_traverse_col = cbs->next_column;
+      return;
+   }
+
    xaccVerifyCursorPosition (table, row, col);
 
    /* compute the cell location */