Shuffle phase replication factor

[John Lilley]

When MapReduce enters "shuffle" to partition the tuples, I am assuming that it 
writes intermediate data to HDFS.  What replication factor is used for those 
temporary files?
john