Notes for 19 Samples of Fourth Graders' Writing from the  The 2004 North Carolina Writing Assessment  Grades 4, 7, and 10 Trainer Manual  (Summer, 2004)

Anthem

These samples were downloaded from the North Carolina Department of Education site at http://www.ncpublicschools.org/accountability/testing/writing/ (file name: trainingmanual2004.pdf) on January 25, 2005. This file will probably be replaced with newer versions in the future, and I have decided against presenting the original scans here. I can, however, send the document to anyone who wants to verify my transcriptions of the students' writing.      I would like to thank the North Carolina Department of Education for replying so quickly to my request to use these samples on this site. Their permission does not reflect an endorsement of the KISS Approach.

      As I understand it, samples G1 to G11 are used for discussion and training, and then samples TA-1 to TA-8 are used for practice and discussion. In the table below, the "Original" column presents my transcription of the students' writing and the accompanying evaluation notes on "Content" and "Conventions." (See below.) The originals have names blanked out, but because we are interested, among other things, in words per main clause, I have inserted names in the blanks (in brackets). In the exercise versions, these false names are kept without the brackets. The "C / C" column states the evaluation results. Errors in the originals have been corrected for the documents in the "Exercise" column so that you can print these to use as analysis exercises. (You may, by the way, want to copy the transcriptions of the students' writing in the originals, minus the evaluators' comments, and use them as exercises in which your students evaluate and correct the originals.

     The inclusion of "sentence fluency" under "Content" probably results in an overlap with "sentence formation" under "Conventions," but note that all the samples that have the highest grade for content (4) also have the highest grade for conventions (2).

     An enormous number of questions can be raised about statistical studies, but they do have their uses. When I first looked at these North Carolina samples, they seemed very sophisticated for fourth graders. They are not, I should note, presented as a statistical sample of the writing of fourth graders -- they are presented as exercises and samples in the Trainer Manual for North Carolina assessment. A statistical analysis, even a quick one like that given here, can, however, give us some perspective of what we are looking at.
     These nineteen samples average 9.9 words per main clause. [Note that each average is arrived at by taking the total number of words in the sample and dividing it by the number of main clauses ("/") in the analysis.] In his famous study, Walter Loban reported that fourth graders averaged 8.0 words per main clause; Kellogg Hunt reported 8.5. [For more on the statistics see the "Summary Statistics" in KISS's Cobweb Corner.] O'Donnell's fifth graders averaged 9.3, but Loban's sixth graders averaged only 9.0, and his seventh graders averaged 8.9. O'Donnell's seventh graders, on the other hand, averaged 10.0. In this context, the fourth grade samples at which we are looking here are comparable to the writing of seventh graders.
     A similar disparity exists in regard to subordinate clauses per main clause. These nineteen samples average .50, or one subordinate clause for every two main clauses. Loban's fourth graders averaged .19; Hunt's .29. In those studies, these numbers remain in the twenties until seventh grade, where they jump into the forties and fifties, and they remain in that range through eleventh grade. (Loban reported an average of .45 for eleventh graders.)
     Although we need to be very careful about drawing conclusions from these studies, there is one important one that we probably should make. The nineteen samples that we are looking at here are probably written, on average, by much better than average fourth graders. This is especially important in that we are making numerous exercises for fourth graders based on these samples. It was, I believe, Hunt who noted that natural syntactic development is "glacially slow." Put differently, we can use exercises based on these samples to help most fourth graders improve their reading and writing, but we need to keep our expectations in perspective.
     There is, however, a completely different way of using these statistics since the samples have been "graded" by the North Carolina DoE. Thus we can look at how the individual statistics compare to the scores for content and conventions. The following table reflects the average number of words per main clause, and subordinate clauses per main clause, according to the content and conventions scores:
 

Content W/MC SC/MC 1 (N=4) 8.6 .32 2 (N=6) 9.9 .56 3 (N=5) 9.7 .55 4 (N=4) 11.6 .52

Conventions  W/MC SC/MC 0 (N=5) 9.4 .44 1 (N=6) 9.3 .42 2 (N=8) 10.7 .60

     Perhaps I am overly influenced by the fact that, at the time I became interested in the teaching of sentence structure, sentence-combining was the rage, and the researchers and educators were pushing students to write longer and more complex sentences. The preceding tables suggest that length and complexity are important -- but not that important. The highest scoring students are well ahead of the rest, but they probably are so because they read and write a lot, and not because of conscious attempts to push them into writing more sophisticated sentences. Some simple sentence-combining exercises are probably appropriate for students who are at the low end of the scale, but even here there are questions. The writer of TA-3, for example, averaged 11.6 words per main clause and .68 subordinate clauses per main clause -- at or above the high end averages. But the sample is scored "2/0." The NC evaluators' comments are particularly interesting and probably highly relevant. The content grade suffers primarily from problems in the logical development and focus of the paper. It, as well as the conventions grade, suffers from run-on sentences, verb usage errors, spelling errors, and incorrect dialogue annotation. In essence, control of sentence structure and conventions is much more important than length and complexity.
     One of the things that fascinated me about this set of samples is the frequent appearance of what I have called subordinate clauses that function as interjections: