It is easier to perceive nonwords that follow the rules of the English language $($e$.$g$.,$ "HICE"$),$ compared to a random string of letters that do not look like English words $($e$.$g$.,"$RSFK$")$ because

they are encountered more often.

bigram detectors for more$-$common letter combinations fire more readily.

they are more distinctive.

word detectors can be quickly created for strings that are nearly words.