This script will iterate through all the machine/section pairs, printing the first line of tokenised data.

#!/usr/bin/env bash

for i in `seq 1 78` ; do 
  HOST=`paste /home/cluster/mnt/madsen /home/cluster/mnt/madsen | tr '\t' '\n' | head -$i | tail -1` ;
  echo "Host $HOST, Section $i" ;
  sleep 2 ;
  ssh $HOST "head -1 /tmp/data/wikipedia/tok/section$i.tok" ;
done