Objective measure of how 'English-like' a string is without understanding English. Based on a table (~389,000 entries in Louis's version) of every observed four-letter sequence in English and its corpus frequency, e.g. 'TION' at ~13M points, 'GSCI'/'DSSS' at ~7K points, 'AACX'/'AEYY' at 1 point. To score a candidate decryption, sum the quadgram score of every length-4 window. Used together with English-letter frequency seeding and iterative letter-swap search to break substitution ciphers.