Basically, I looked at three things in the students' papers I collected. Length, surface errors, and use of evidence. So let's start with the simplest one, length.
Length is pretty straightforward. I counted the number of words in the papers. Some of the students who wrote the papers probably weren't familiar with this method of determining length, since they'd enlarged their margins and font size to make their papers physically larger. I did the word counts by hand, since I had hard copies of the papers. I didn't count words that were part of citations or headers. Only the actual body text. I did count words that were part of quotations.
I felt that length might not compare well across classes, because even though the three instructors were working with a pretty standardized syllabus, they might have had very different ideas of what length requirements meant. An instructor who automatically failed any paper which didn't meet a minimum length would likely receive longer papers than an instructor who considered length a guideline to help students figure out how much depth to go into. If such differences existed between the instructors of the classes I studied, I didn't want them to overshadow differences that might be associated with procrastination. So I found the median length of paper in each class in addition to an overall median length for all the papers. I considered a paper short if it fell in the first quartile of length in its class and long if it fell in the fourth quartile.