Also, I've done character, digraph and trigraph studies, without spaces, as
standard practice, all of which lead me to believe that the "word" is the
actual written entity. These studies peak with average word length, and
when spaces are observed the most repetitive structures appear. Going
beyond the "word" to include endings and/or beginnings of nearby words
yields interesting observations for a relatively few patterns, but starts to
break down the "cohesiveness" of the study, bringing us back to "word"
again. A detailed view of why this may not be so would be most interesting.