In the course of discussions on a listserver, someone said that if teaching grammar is effective, then providing research that demonstrates that it is effective should be easy. All we have to do is say that we are going to teach students to recognize prepositional phrases. Give them a pre-test. Teach them, and then give them a post-test. Unfortunately, it is not that easy. First of all, why do we want students to be able to recognize prepositional phrases? The anti-grammar research was almost all based on the effects of teaching grammar on students' writing. (That it was so itself makes that research questionable -- it ignores the effect on reading and thinking, but that is anothe question.) But to make a satistical research project convincing, there are several variables that must be taken into consideration.

     The first thing I want to do is to distinguish between a research project for the profession, and research projects in the classrooms. Eentually, I will suggest, the latter will be far more important. I will deal with them below, but before they begin to occur, the profession needs to be convinced.

     Although I do not agree with their conclusions, the research projects by Mellon, O'Hare, and Bateman & Zidonis did explore several important variables. Any research project which intends to convince the profession should at least consider them. [I have added some suggestions of my own.] I am not sure, however, that we are ready for research projects intended to convince the profession that teaching grammar is effective. As I note in my critique of the researchers who did comparative studies, they totally lacked a theoretical base; in effect, most of them simply assumed that longer main clauses means better writing. Before we undertake expensive, time-consuming research to "prove" that teaching grammar is effective, we need to do more small projects which explore the natural syntactic development. Such studies can give us an idea of what to look for, and why.

Theoretical Base

     Although the NCTE resolution condemned the teaching of grammar that is not supported by "theory and research," most of the studies that led to the resolution had little, if any, theory. Hunt admits that he had none. He was simply looking for some way to measure syntactic growth. Counts of words per sentence did not work. (Fourth graders build monster sentences by compounding with "and.") Mellon, O'Hare, and most of the other anti-grammarians simply built on Hunt's results by assuming that longer sentences are "better." (If you don't believe me, read their studies.) This is not exactly a theoretical base. But without such a base, one might as well count the number of times a writer uses the word "the." [I have attempted to explain my theoretical base in Teaching Grammar as a Liberating Art.]

Sample Size

     Most researchers agreed that a writing sample should be at least 250 words long. This allows the writer to "get into" the writing. Because they were doing the analysis by hand, most of the researchers limited sample size to just over 250 words, stopping with the last word in a sentence once they had counted to 250. (Some of the studies ignored this, and, if the sample was short, combined writing from several samples. Because there might be? a relationship between sentence structure and overall writing ability, and because I analyze passages with the assistance of a computer, I decided to analyze complete essays for my Spring 95 project.)

Sample Count

     Statistically speaking, the number of samples examined in most studies was very small. Thus far, it is only when several smaller studies reinforce each other (as in the work of Hunt, O'Donnell, and Loban) that their results can be considered significant.

In-Class, or Out?

     In-class writing samples have the disadvantages of being written in a hurry (no time to edit) and of often being illegible. One could, of course, as O'Hare did, have the transcriber edit out all the errors, but that begs the question. Samples written out of class have the disadvantage of contamination from outside help and/or plagiarism.

Availability of Original Samples and Transcripts

     As this site develops, I may be able to explain even more of the details involved in this type of analysis. One of the problems, for example, is that different researchers defined "main clause" (and "subordinate clause") differently. Such questions would be less serious if the original samples and transcripts of them were available for others to use.

Mode of Writing

     Some studies show that the mode of writing (narrative vs. expository) affects main-clause length. As most writing teachers would suggest, it also affects error-count. A lot of work could be done in this area.

Content of Instruction

     The comparative studies usually made a nod in the direction of this question, but much more could have been done. It is not uncommon for the studies to say that the grammar group did the "normal" or "regular" work on grammar, but what does that mean? Likewise, in some studies, the group that did not study "formal grammar" was taught grammar in writing conferences with the teacher.

Instructors' Influence

     As I understand them, several of the researchers who did comparative studies would not accept a research project which attempted to demonstrate that my approach to teaching grammar is effective if I did the teaching. They would claim that I am biased, that I am more knowledgeable than the average teacher, or both. There are, however, serious questions about who taught which groups (control and experimental) in their own studies. If the instructor does not like or understand the grammar that is being taught, instruction will certainly be affected.

Definitions and Examples of What Is Being Counted -- And Why

     The comparativists were fairly good at this, but not good enough. As O'Hare noted:

"Mellon counted clauses of condition, concession, reason, and purpose as separate T-units [i.e., main clauses] because he believed that logical conjunctions behave much like coordinate conjunctions. In addition, he discarded clauses with repeating predicate phrases because he claimed they were elliptical and therefore vacuous. This experimenter remained unconvinced by Mellon's reasoning in either case and, therefore, retained Hun and O'Donnell's simpler and more convincing methodology." [Sentence Combining, NCTE, 1973. 48.]

O'Hare claimed that "The seventh grade was selected as the level on which to conduct this experiment simply because Mellon chose seventh graders." (37) His statement is highly suspect. Hunt and O'Donnell had already demonstrated that seventh grade is the beginning of the blossoming of subordinate clauses. There is also the quesiton of when, in the process of developing their writing skills, students begin to master, in Piaget's sense, questions of condition, concession, reason, and purpose. Seventh grade sounds like a prime candidate. O'Hare, in other words, should have had much better reasons for choosing seventh graders than just the fact that Mellon chose them.
      There is also the quesiton of what, exactly, counts for what. Among the documents at this site, there are several on which I have, and will continue to, collect examples of questions. Consider, for example, the sentence

Does "After playing baseball" count as a subordinate clause? Or is it a prepositional phrase with a gerund as object? On one side of the argument, the construction is similar to "When playing baseball, the children forget their problems." Because "when" is not considered a preposition, the latter example is usually defined as a subordinate clause with its subject and part of its verb ellipsed -- "When they are playing baseball, the children...." But on the other side of the argument, the construction is similar to "The children were talking about playing baseball." The latter is clearly a prepositional phrase -- one cannot simply reinsert the subject and part of the verb: "The children were talking about they were playing baseball."
     The problems in defining exactly what one is counting are one reason for the need of the original transcripts of the samples.

     Whereas research to convince the professionals involves numerous problems, statistical syntactic research can be done in most classrooms beginning as early as even third grade. I have suggested elsewhere that third graders could learn to identify all the prepositional phrases in their own writing. What third graders write is usually not very long. Thus they could count the total number of words, count the words in the prepositional phrases, divide the number of words in prepositional phrases by the number of total words, and see what the results are -- their own, and the average for the class. Then they could discuss what those words in prepositional phrases do in their writing.
     Many English teachers tell students to use strong verbs. That is funny once you realize that most students cannot identify the verbs, much less "strong" ones, in what they write; apparently many English teachers like to talk to themselves, out loud, in front of their students. The KISS approach, on the other hand, teaches fifth graders to identify all the subjects and verbs in their own writing. Fifth graders could write a passage, find all the verbs, count them, DEFINE THE "WEAK" AS OPPOSED TO THE "STRONG, count them, and calculate one as a percentage of the other.
     Seventh graders could start doing what I teach my college students to do. Identify all the main and subordinate clauses in a passage of their own writing, count them, count the number of words in the passage, and calculate the average number of words per main clause, and the average of subordinate clauses per main clause. They could then calculate the average for the class to see how their writing matches up. Such an exercise can be extremely important for some students. In my classes, I have had students who had NO subordinate clauses. When they saw that almost everyone else had subordinate clauses in their writing, these students WANTED to learn how to write subordinate clauses. On the other hand, students at the other extreme were cautioned that their writing could become too complex.
     Statistical syntactic research in the classroom also has other advantages. For one, it engages them in the scientific method -- the collection and analysis of samples, and the reaching of conclusions therefrom. [Although students have supposedly been taught the scientific method in middle and/or high school, I am biannually dismayed by their lack of understanding of it when they reach me as Freshmen in college. Somewhere, something is wrong.] Statistical research would also integrate the teaching of grammar and math. In addition to the work with division and percentages, students could, for example, make graphs -- of class averages, of the (anonymouos) results of each class member, or even of their own progress from year to year. (The latter presupposes that, for example, they analyzed the number of words per main clause in a selection of their own writing each year.)
     Statistical syntactic research in the classroom -- why isn't it being done?