Chapter 3: Research on Natural Syntactic Development

       Volumes of research have been published on natural syntactic development, but most of it, especially that of the last thirty years, is irrelevant to what goes on in our schools. David Crystal, for example, summarizes a famous experiment by Carol Chomsky in which she examines children’s ability to distinguish the meanings of "hard to see" and "easy to see":

What Carol Chomsky did was to present a group of children, aged from five upwards, with a blindfolded doll, and ask them "Is this doll easy to see or hard to see?" If the children said "Easy to see" then it is argued that they have learned the distinction; but if they said "Hard to see," and amplified their comment (upon request) by for instance, "Because it’s got a blindfold on," then it is argued that they had not learned it. The Chomsky results showed that before the age of six this distinction was hardly ever learned, whereas after seven it was known to nearly all children in the sample. (49)
The statement that such research is "pedagogically irrelevant" requires amplification. As Crystal quite correctly goes on to note: "it is possible to extract a general conclusion from this and similar experiments, namely, that there are matters of structural (and in this case also semantic) interpretation which it takes children many years to acquire." But once we accept the idea that syntactic development occurs over a period of years, perhaps even throughout an individual’s whole life, no further pedagogical conclusions can be drawn from such research: it suggests nothing about what we could or should teach. At age five, most children have trouble with this particular distinction; at age eight, most of them have mastered it. But they have figured it out for themselves -- no "instruction" was involved, no "pedagogy."
       The number of complex structures which children somehow figure out for themselves is awesome. I remember, for example, listening to my four-year-old son, speaking about a light, and saying, "Turn on it," instead of "Turn it on." Why children go through this stage with phrasal verbs, and how they work themselves through it, pose complicated linguistic and developmental questions. But these questions are not "pedagogical": children work their way out of them, all on their own, without any other instruction than access to conversations, just as they work their way through and out of "I readed a book" and "I cutted up the paper."
       There are many things we can’t teach, and shouldn’t try to. This does not mean that the developmental research of the last two decades is totally valueless: it has provided and confirmed some important general principles of language development, and, in such fields as teaching the learning disabled, it even provides many practical suggestions. But for the normal child, studies that involve linguistic development in pre-school children are irrelevant. For relevant studies, we must return to the sixties and seventies and the work of such people as Kellogg Hunt, Roy O’Donnell and Walter Loban. Fundamentally, what their studies demonstrated is that subordinate clauses naturally develop between seventh and ninth grades. Some constructions, such as participles and appositives, blossom after students have mastered subordinate clauses.

Hunt's "T-Unit"

       The work of these researchers centered on the child’s syntactic development between the ages of five and eighteen. The sentences of adults are obviously different than the sentences of children, and researchers had been looking for some way to measure the differences. A simple measure of words per sentence does not work because third, fourth, and some fifth graders create long sentences by compounding main clauses with "and."  Hunt wondered if a count per main clause (defined as including all subordinate clauses) would be a better measure. Perhaps because the grammarians cannot agree about their terms (See Chapter One.), Hunt used the term "T-unit" -- for "minimal terminable unit." (In doing so, of course, he simply added to the terminological chaos.) To denote what he was measuring, he used the term "maturity," but he was careful to define it:

In this study the word "maturity" is intended to designate nothing more than "the observed characteristics of writers in an older grade." It has nothing to do with whether older students write "better" in any general stylistic sense. [Grammatical p. 5]
Hunt convincingly demonstrated the validity of the T-unit as the "basic" unit of measurement. I emphasize "basic" because Hunt himself notes, as we will see, that it is not the only gauge. All of the subsequent major research studies have used Hunt's "T-unit" as their fundamental yardstick.
       Simply stated, Hunt counted the number of words per main clause in the writing of students in different grades and found that the average number increased steadily from grade to grade. Hunt’s study was corroborated by O’Donnell, to give the following figures, as summarized by O'Hare (22):
                                Grade Words / Main Clause
                                  3            7.67
                                  4            8.51
                                  5            9.34
                                  7            9.99
                                  8          11.34
                                  12         14.40
                                Superior Adults 20.30

The figure for "Adults" is from Hunt’s study of essays published in The Atlantic and Harper’s. Although there are numerous questions surrounding such statistical studies, the basic conclusion remains valid: if samples of writing are taken from students (of similar socio-economic status) at different grade levels, the average number of words per main clause will show an increase similar to that above. Likewise, a study of passages from a variety of published writers will often result in a count of approximately twenty words per main clause.

The Natural Development of Subordinate Clauses

        Hunt, O'Donnell, and Loban all concluded that subordinate clauses "blossom" (to use Hunt's term) between seventh and eighth grades. The following table is based on O'Hare's summary (p. 22) of Hunt's and O'Donnell's studies.

Sub Clauses
Grade  / 100 Main Clauses Increase
                                  3   18
                                  4   29  + 11
                                  5   27  - 2
                                  7   30  + 3
                                  8   42  + 12
                                  12   68  + 26 (over four years)
                                Superior Adults 74  + 6

We need to be cautious with these results. The numbers for fourth, eighth, twelfth, and "Superior Adults" come from Hunt's study; the others, from O'Donnell's. Differences in the way constructions were defined and counted may make the results not totally comparable. The table, however, does suggest a major "blossoming" of subordinate clauses between seventh and eighth grades.
       Perhaps the most interesting study here, however, is Loban's. As its title suggests (Language Development: Kindergarten through Grade Twelve), Loban's study is the most comprehensive. In addition to its time span, Loban, like O'Donnell, studied both oral and written language. But he also divided the students whose sentences he studied into three groups, high, low, and random. The following graph (from page 39) clearly reflects the sharp increase in subordinate clauses between seventh and eighth grade for the low and random groups. Note that the "high" group shows a much slower increase during this grade level, and that, between eighth and ninth grades, all three groups show a decrease, the decrease being most significant for the high group. (I'll have more to say about this later.)

       Because even teachers of kindergarten may see what appear to be numerous subordinate clauses in the writing of their students, it is important to note that we are looking at a complicated question. O'Donnell, for example, wrote that "One of the most enigmatic features in the whole array of data collected in this study is the showing that kindergarten children used relative clauses more frequently than did children at any other stage, in either speech or writing." (60) It is "enigmatic," of course, because what appears to be a sign of increasing maturity decreases instead of increasing with age. It is highly possible that the relative clauses in the writing of kindergarten children are formulaic. (See below.)  All of the researchers attempted to break "subordinate clauses" into subcategories, but such distinctions become very tenuous unless the transcripts of the original data are available for review.
       Until we get that research, the only conclusive studies we have are those of Hunt, O'Donnell, and Loban, all of which clearly show that subordinate clauses naturally blossom in seventh and eighth grades. Attempts to get entire classes of students to use more subordinate clauses in their writing before seventh grade may, as Hake and Williams suggest, be counterproductive. (See Chapter Two.)

O'Donnell and the Concept of "Formulas"

       In his section on "Conclusions and Implications," O'Donnell parenthetically alludes to "formulas." Formulas create a serious problem, both for anyone doing statistical research on natural syntactic development, and for some teachers.  In the first paragraph of the following, O'Donnell comes as close as he ever does to suggesting a sequence of natural development beyond subordinate clauses. In the second, he mentions "formulas."

       On the other hand, there was a group of items that appeared more than sporadically in kindergarten speech but were used from about three to ten times oftener by seventh graders. At various levels, there were significant increments in their use. These would appear good candidates for identification as generally later acquisitions. They were noun modification by a participle or participial phrase, the gerund phrase, the adverbial infinitive, the sentence adverbial, the coordinated predicate, and the transformation-produced nominal functioning as object of a preposition.
       Theoretically, it seems reasonable to suppose that these constructions (unless acquired as formulas) would be mastered relatively late. Transformational grammar derives them all by application of deletion rules, and some of them indirectly from their sources by way of strings that could more directly yield subordinate clauses. Thus, The man wearing a coat . . . may be more difficult than The man who was wearing a coat . . ., and A bird in the tree . . . more difficult than A bird that was in the tree . . . . Noting that noun clauses did not vary much in frequency after the first grade, while participial modifiers of nouns were used by seventh graders three times as often in speech and nearly eight times as often in writing as they were used by kindergarten children, we may contend that such clauses (The dove saw that the ant was drowning ) are easier to manage and earlier added to the child's repertory than is the reduction of them to a single participial modifier (The dove saw the ant drowning). (92)
The first paragraph supports the basic conclusions mentioned previously, and it extends them into areas that will not be covered in this book. Coordinated predication (She read a book and then took a nap), for example, would be an interesting area for research. Some researchers have counted this as two separate main clauses, and others have reported conflicting results about when and how often they appear. It should be obvious, however, that such coordination may have a significant affect on main-clause length.
        The most interesting point in the quotation, however, is the parenthetical allusion to "formulas." I have tried several times to find other references to this concept within O'Donnell's study, but with no success. This may be because the concept calls into question much of the statistical research of all three researchers. It is generally agreed (and a matter of common sense) that much of language is learned not as individual words, but rather as strings of words. This is obvious, for example, in idioms such as "It's raining cats and dogs." It is also apparent in many verbal phrases ("Wake up." "We get along.")  Formulas are thus an aspect of vocabulary, but they are also an aspect of syntax.
       Syntactically, some tenses are probably learned as formulas. Children repeatedly hear, for example, such things as "You are going to the store." "They are talking on the phone." "We are playing a game." What they learn from this is a syntactic string -- "___ are ___ ing" into which various pronouns (and, of course, nouns) and verbs can be inserted. We can also see it, for example, in young children's use of some "subordinate" clauses. These children hear "when ___ get(s) ___" dozens of times, and probably pick this up as a string, i.e., a formula.  If I understand O'Donnell correctly, he is implying that such formulaic expressions do not represent cognitive, syntactic mastery of the underlying construction.
       O'Donnell never discusses the possibility that formulas may account for most, if not all, of the "advanced" constructions in younger children's language, but he clearly does not see this problem as undercutting the basic conclusion that subordinate clauses blossom in seventh grade.  Although neither O'Donnell, nor any of the other researchers ever clearly discuss the question, we should also note that one cannot separate a purely syntactic description of children's writing from the semantic, or from their conceptual abilities. As Vygotsky observed,
The child may operate with subordinate clauses, with words like "because," "if," "when," and "but," long before he really grasps causal, conditional, or temporal relationships. He masters syntax of speech before syntax of thought. Piaget’s studies proved that grammar develops before logic and that the child learns relatively late the mental operations corresponding to the verbal forms he has been using for a long time. (46)
It is very possible, therefore, that the subordinate clauses, appositives, participles, etc. in the writing of most pre-seventh graders are formulaic.
       Formulaic expressions in students' writing and speech may thus often deceive teachers into thinking that their students understand advanced constructions when, in fact, they do not.  Many teachers, for example, claim to have seen participles, appositives, etc. in the writing of their young students, and thus claim that their students can understand and will be helped by grammatical explanations of these constructions. But when they are asked if what they saw might be formulas, these teachers have usually never even heard of the concept. Attempting to force advanced, "late-blooming" constructions into the writing of students who are not ready for them may cause more confusion than clarity.
       In this respect, Loban's study is particularly important. The graph given above is just one of many that demonstrates significant differences between advanced and slower students. The advanced constructions that teachers claim to see may just be in the writing of these students. If that is the case, the advanced students don't really need more help, especially if that help comes at the cost of confusing the less advanced students. If the constructions appear in the writing of the weaker students, they are probably formulas. As noted in the previous chapter, Hake and Williams suggest that sentence-combining exercises, for example, are effective only when students are ready for them.

"Late-Blooming" Constructions

        Throughout this chapter I have suggested that these researchers have shown that some grammatical constructions develop later than others. The general outline of that development appears to be 1) the development of basic sentences through sixth grade, 2) the blooming of subordinate clauses in seventh and eighth grades, and 3) the late blooming of constructions such as the gerundive and appositive. Hunt was the researcher who explored and expressed this sequence the most clearly and forcefully.
        It is tempting to say that Hunt even had a theory about why syntactic development occurs both in this sequence and as slowly as it does. From my perspective, 99% of the KISS theory of natural syntactic development (Part Two of this book) is clearly implied in Hunt's work. If Hunt didn't develop the theoretical aspects of his work, it was probably because he had started with the objective of finding a way of measuring syntactic growth, and, once he found it, he had to explain it. This is clear in his 1966 article, "Recent Measures in Syntactic Development." This eight-page article is an excellent introduction to Hunt's work. The basic objective of the article is to explain the "T-unit" and "subordinate clause index" as fundamental measures of increasing maturity. If we look at the article more closely, however, the concepts that Hunt explains, and the examples that he uses, contain almost everything one needs for a theory. Hunt is, we should note, very tentative in his presentation:

The science of measuring syntactic maturity is barely emerging from the stages of alchemy. It scarcely deserves to be called a science at all. But we do know a few things.
        For the last thirty years we have known at least three things about the development of language structure. First, as children mature they tend to produce more words on any given subject. They have more to say. Second, as children mature, the sentences they use tend to be longer. Third, as children mature a larger proportion of their clauses are subordinate clauses.
        In the last two years it has been possible to add a few more measures, and I will come to them later. But first let me turn back to the statement about subordinate clauses and try to make clear its significance for the teaching program. It would be worse than useless for a fourth-grade teacher to say to her students, "Now if you will go back to your last paper and add more subordinate clauses to the main clauses, you will be writing like Miss Hill's wonderful sixth graders or Miss Summit's wonderful eighth graders instead of my own miserable fourth graders." Such an approach would be worse than useless. (732, my emphasis)
Hunt's focus is on "measuring," but he is clearly cautioning against rushing natural syntactic development.
        Having explained that, among subordinate clauses, adjective clauses are the best measure of maturity, he continues:
        But of course subordinating clauses is not all there is to syntactic development. In every pair of examples I have given so far, it would have been possible to reduce one of the clauses still further so that it is no longer a clause at all, but merely a word or phrase consolidated inside the other clause. In this fashion two clauses will become one clause. The one clause will now be one word or one phrase longer than it was before, but it will be shorter than the two clauses were together. By throwing away some of one clause we will gain in succinctness. The final expression will be tighter, less diffuse, more mature. (734)
Hunt here clearly proposes a basic outline of natural syntactic development. Children develop mastery of the basic main clause, then expand it by embedding one clause into another (subordination), and then learn to tighten their sentences by reducing subordinate clauses to less-than-clause constructions.
        We should note that Hunt's conclusion clarifies and supports O'Donnell's previously noted "enigma": "kindergarten children used relative clauses more frequently than did children at any other stage . . ." (60) As O'Donnell himself suggested, children may more easily produce A bird that was in the tree . . .  than A bird in the tree . . . . (92) Hunt is suggesting that we need to look at this in terms of full clauses. Thus the "kernel" sentences will be A bird was chirping and The bird was in the tree. The first stage of syntactic development is to embed the entire second clause into the first, thereby producing a relative clause: A bird that was in the tree was chirping.  After the child has cognitively mastered that construction (which may take years), the subordinate clause is reduced to a simple prepositional phrase: A bird in the tree was chirping.  As this development proceeds, the relative clauses that O'Donnell found to be so frequent and so enigmatic in the writing of kindergarten children simply disappear.
       In 1977, Hunt published "Early Blooming and Late Blooming Syntactic Structures." Like "Recent Measures," it is short, easy to understand, and very important for anyone interested in natural syntactic development. Unlike his earlier and most famous study, the study reported in "Blooming" involves what Hunt calls "rewriting" as opposed to "freewriting."  Whereas for the earlier study he had simply collected samples of whatever the students were writing about for school, in this "rewriting" study participants were given the "Aluminum" passage and asked to rewrite it. The "Aluminum" passage is written in very short sentences. Hunt includes the entire passage at the end of the article, but he discusses the first six sentences: 1) "Aluminum is a metal." 2) "It is abundant." 3) "It has many uses." 4) "It comes from bauxite." 5) "Bauxite is an ore." 6) "Bauxite looks like clay."  In explaining the advantages of the passage as a research tool, Hunt notes that "since all students rewrite the same passage, all students end up saying the same thing -- or almost the same thing. What differs is how they say it. Their outputs are strictly comparable." (92) The passage can be an excellent research tool. Like all these research reports, this one is crammed with data and ideas which fascinate those of us who are interested in the problem. Here, however, I have to limit myself to two of what Hunt considers the middle and later blooming constructions.
       The first of these is the appositive. For adults, appositives appear to be relatively simple constructions in which the syntactic connection results from an identity of meaning, and not from a preposition or from a subordinate conjunction -- Winchester, a city in Virginia, is a beautiful place to live in. From the perspective of natural syntactic development, however, appositives are much more complicated. Fourth graders, for example, tend to write lists that appear to be appositives -- Our house has five rooms a kitchen, a living room, a bathroom, and two bedrooms. Such lists, which usually appear at the ends of sentences, are probably produced as afterthoughts, i. e., they are probably not generated by embedding one sentence into another and dropping the connecting words.
       Hunt states that "Ability to write appositives was in full bloom by grade eight, but not by six or four." (98) A much more technical version of this chapter (on the KISS web site) explores six objections to Hunt's conclusion that "appositives was in full bloom by grade eight." There are probably major differences between giving eighth graders combining exercises in which they create subordinate clauses, and giving them exercises that necessitate appositives. There is certainly a difference between following and supporting Mother Nature, as opposed to forcing her.
        Although I have suggested caution about Hunt's conclusion relative to eighth graders and appositives, his data can be interpreted more forcefully in a different way -- fourth and sixth graders are clearly not ready for instruction in appositives. Given 200 invitations each (50 students times 4 invitations per student), they produced only one and eight appositives respectively, and even here, without the transcripts, we cannot be sure that the appositives that they did produce were not formulas.
        The other important late-blooming construction is the gerundive, also known as a participle. Interestingly, Hunt does not give it a name, referring to it instead as "this construction," and identifying it with an example from Christensen, who took it from E.B. White:
We caught two bass, hauling them in briskly as though they were mackerel, pulling them over the side of the boat in a businesslike manner without any landing net, and stunning them with a blow on the back of the head.
Hunt reports that of the 300 people who rewrote "Aluminum," "not one of them produced this construction" (my emphasis). Perhaps because no one used it for him, Hunt here switches to a discussion of the "Chicken" passage. He states:
Out of 10 fourth graders who rewrote "The Chicken," not even one produced it. By 10 eighth graders who rewrote it, it was produced once:
       She slept all the time, laying no eggs.
By 10 twelfth graders this construction was produced twice. Here are both examples.
       The chicken cackled, waking the man.
       Blaming the chicken, he killed her and ate her for breakfast.
But the university students produced 14 examples. In fact, 9 out of 10 university students studied produced at least one example, whereas only 1 out of 10 twelfth graders had done so. In the little time between high school and the university, this construction suddenly burst into bloom." (100, my emphasis)
The gerundive is a late-blooming construction -- currently, perhaps as late as college. We also need to keep in mind Loban's clear demonstration that students do not progress through the natural sequence of syntactic development simultaneously. The fact that one eighth grader produced a gerundive does not mean that all eighth graders should be forced to study, or, worse yet, try to use, gerundives.

        Near the end of the article, Hunt does suggest the application of the "blooming" sequence to teaching. "The kind of information given previously as to which structures bloom early and which bloom late would be preliminary to actual measures of teachability at a given level" (102).  This sentence clearly implies that certain constructions should be taught at later grade levels, but it was never developed and has been ignored. Thus we have sentence combining exercises for second graders in which they are required to use appositives, and even, in some cases, gerundives. If, instead of trying to rush instruction into earlier grade levels, we could follow the natural syntactic development suggested by the research, wouldn't we have a much more effective and theoretically justified approach to teaching grammar?
       Hunt, O'Donnell, and Loban were more interested in research than in theory. As a group, they convincingly demonstrated that the ability to generate subordinate clauses (as opposed to using formulas) blossoms between seventh and ninth grades. Appositives and gerundives bloom after students have mastered subordinate clauses, probably because they are generally reductions of such clauses. Hunt was, as I suggested at the beginning of this section, very close to the KISS theory of natural syntactic development. In fact, he was extremely close. In Part Two of this book, we will see just how close.
 

[The longer, web version of this chapter is in the section on "Essays on Grammar." In addition to discussing some of the points discussed here (such as formulas and late-blooming constructions) in more detail, it also explores why the work of Hunt, O'Donnell, and Loban was not pursued, and what problems will be faced by further research.]