Chapter Four -- Grammar in Motion -- A Psycholinguistic Model

       In Whither Music?, his brilliant survey of the history and theory of music, Leonard Bernstein develops an analogy between parsing sentences and parsing music. For both sentences and for music, Bernstein finds that parsing is pretty much useless, indeed "unbearable" and largely "meaningless" -- unless we have a theoretical context that gives the parsing meaning. Surprisingly, Bernstein builds his theoretical context for a "musical linguistics" by using basic concepts of Chomsly's transformational-generative grammar, primarily deletion and embedding. Using these concepts, Bernstein returns to parsing and finds it essential to his explanation for laymen of musical history and theory. I have emphasized "for laymen" because the grammar we teach in our schools should be a grammar for laymen, not for specialists in linguistics.
       The two transformational linguistic concepts that Bernstein adopts are fundamental to KISS grammar, and they can easily be explained in terms of traditional terminology. Deletion is, in part, comparable to the traditional concept of "ellipsis," the omission of words that are assumed to be understood. But it is much more powerful than is traditional ellipsis. According to transformational theory, all language is composed from very simple, deep-structure sentences, such as The jewel was in the drawer and The jewel was red. These two sentences can be combined into The red jewel was in the drawer by deleting The jewel was from the second sentence and embedding the adjective red into the first. Another way of looking at this is to say that the second sentence was reduced (to the adjective red) and then embedded into the first. As I will try to show in several different contexts later, deletion (reduction) and embedding enable us not only to simplify traditional grammar, but also to explain many aspects of natural syntactic development. Although they are crucial to putting grammar "in motion," these transformational concepts in themselves do not provide a good pedagogical context. A major boost toward developing such a context came in 1956, with George Miller's publication of "The Magic Number Seven, Plus or Minus Two."

Miller's Theory of Short-Term Memory

        Any pregnant new idea is subsequently developed, extended, and altered such that the meanings of terms become confused. "Short-term memory," for example, is now used by some psychologists and educators to refer to what we can remember next week from instruction that we receive today. This was not what Miller had in mind. Miller was concerned with what the mind can handle from one second to the next.  (Because of the confusion in terms, Miller's original concept is now often referred to as "working memory.")
        In his primary experiments, Miller put an assistant and a human subject in a quiet room. The assistant had a list of sequences of random numbers:

  1, 6, 9, 8, 3, 6, 8
  6, 8, 4, 3, 2
  7, 5, 4, 3, 8, 6, 5, 3, 1, 7, 8, 9, 0, 7
The list, of course, was longer than my example. The assistant would read a sequence, and the human subject was asked to repeat it. The assistant then recorded the response as either correct or incorrect. The experiment, naturally, was repeated over and over again, with different assistants and different subjects. The results produced a very nice bell curve. Most people were able to remember all the sequences that were seven digits in length. Some people had trouble with six-digit sequences; a few had problems with five digits. At the other end, some people managed the eight digit sequences very well, and a few managed those that were nine digits long. This gave Miller his "magic number seven, plus or minus two." Miller concluded that humans basically have a seven-slot working memory.
        The theory raises some interesting questions. If, for example, humans can only remember seven digits at a time, how can we remember phone numbers? The basic phone number is seven digits long, but if we put one digit in each slot of our STM, how do we remember whom we are calling and why? The obvious answer to that, of course, is that we do not remember phone numbers as individual digits -- we "chunk" them, usually into one chunk of three digits, and either a second chunk of four, or two following chunks of two each. Theoretically, if we did not chunk phone numbers, we would forget whom we were trying to call, and why.
        Although Miller started with numerical digits, the experiment was, of course, extended to words and sentences. The obvious question was if humans are limited to a seven-slot working memory, how can we understand and process sentences that are longer than seven words? Here again, the obvious answer is that we chunk them -- into phrases, etc.  The KISS web site includes a "psycholinguistic model" which demonstrates this chunking. In that model, the seven slots are represented at the top of the screen. At the bottom is the following sentence, which was written by an eighth grader: " If you want to go to Florida by car, you must plan to drive awhile."  The model takes 34 screens to move the words from the bottom sentence into an unoccupied slot and explain what is chunked with what until the entire sentence has been processed.
        Thirty-four diagrams are too many to be reproduced here, but we can look at the general principle, even without diagrams. "If" would go into the first slot, after which "you" would occupy the second. Readers expect to find a (grammatical) subject, and experienced readers would probably expect one even after the "if." Thus they would chunk the two words into one slot. (Note that inexperienced readers may not do so, and, as a result, have to use more slots to process the sentence.) With the "if you" in the first slot, "want" would then enter the now unoccupied second slot, but after subjects we expect verbs. Thus "want" would be chunked to the "If you" in the first slot. To simplify, we'll put "to" and "go" into the empty second and third slots, but having read (or heard) "want," the reader (or listener) will expect an answer to the question "want what?" Thus the ":to" and "go" are chunked together and join the "If you want" in the first slot as the direct object of "want." Likewise, "to Florida," a prepositional phrase, is chunked as an adverb to "to go." The model simply continues through the entire sentence to show students how the entire sentence is chunked using four slots.
       One nice thing about the model is that it can be presented to students as an introduction not only to some basic grammatical terms, but also to a good reason for why they might want to learn the terms. The students do not have to understand most of the terms to understand the model. They know, for example, that "by" chunks with the following "car" first, and not directly to the preceding "Florida." (We need to build on the excellent understanding of English syntax that all of our students have; instead, we usually ignore it.) The model is both an important teaching tool and the fundamental justification of all of KISS grammar. It validates both the rules we teach and the KISS theory of natural syntactic development.

The Model as Validation of the Rules We Teach

       Given the model, students can understand why some rules are more important than others. Using the model, of course, students see one word chunking to another word, or to a group of other words. This "chunking" is the dynamic equivalent of the traditional concept "modification." Note, for example, that the model explains why dangling or misplaced modifiers are an "error." They are not an error because they will violate somebody's sense of etiquette (which is what I used to think); they are an error because the reader will have problems determining what they chunk to.
       Because the model is dynamic, it also explains why comma-splices, run-ons, and fragments are usually considered as errors. One of the things psycholinguists have done (and students can reproduce the experiment on their own) is to give "subjects" one sentence followed by a second sentence. Asked to repeat the first sentence, most people will reproduce the basic ideas, but they will use different syntax and vocabulary. This has led linguists to conclude that a main clause is generally processed in STM and then dumped to "long-term memory." The periods (or other ending punctuation marks) at the end of main clauses thus signal to readers that this sentence should be dumped to "long-term memory," thereby clearing STM for the next main clause. Comma-splices and run-ons can be errors because they do not signal the "dump." Readers, therefore, try to chunk words from the next main clause into the preceding one. That doesn't work. Processing crashes. Fragments cause crashes for a different reason. Readers become accustomed to the basic main-clause pattern -- every word in every sentence (with the exception of interjections) chunks to another word or construction until everything chunks back to the subject and finite verb pattern of a main clause. Fragments can confuse readers because a "dump" is signaled before a complete pattern has been established.
       In addition to punctuation errors, the model suggests the importance of some spelling errors. Learning to read is not just a matter of learning vocabulary words. It is a matter of learning how the syntax of English works, and then, using what one has learned to make educated guesses. As noted above, after "want," a reader expects an answer to "want what?" Similarly, "to" always evokes the question "to what?"; "too" never does. Using the wrong spelling leads readers either to expect a "what" when they shouldn't (because "too" was meant), or to get a "what" when they didn't expect it. ("She wanted to go too the store.") Similar problems occur with the misspelling of "its," "their," etc.
       Although an occasional error (of any of the types discussed above) may not cause a reader a problem, numerous errors cause very serious problems. To illustrate this to my students, I note that we can run on loose, rocky ground, trip, and usually keep running. But if we trip a few times in quick succession, sooner or later, we will end up on our behinds. And there are other sentences that are more comparable to running on ice -- one good slip and down we go. CRASH!
       What happens in a reader's mind during a serious crash? I processed our model sentence using only four of the seven slots of STM. I am not an expert in psycholinguistics, and I am sure that the situation is more complex than I present it, but I suggest to students that we use those other "free" slots to keep track of other, more global aspects of what we are reading -- the topic sentence of the paragraph, the thesis of the paper, etc. But when a crash occurs, the reader has to focus more slots on the crash site. Something has to get thrown out -- it may be the topic sentence of the paragraph, it may be the writer's thesis. The more errors there are, the more opportunities there are for the reader to crash.
       Although a lot more could be said about the psycholinguistic model and the question of errors, perhaps the most important point is the radical shift it creates within the pedagogical situation. The teacher is no longer imposing rules without reason. In fact, the teacher doesn't even have to impose the rules at all -- the model does it. The teacher, instead of being viewed as the policeman, becomes the "coach."

The Psycholinguistic Model as an Introduction to Normative Stylistics

       I don't remember reading much about syntax and the "texture" of sentences, but it can be an interesting and helpful concept to present to students. The question of "style" is itself interesting because many people consider "style" to be something positive -- "She did it with style." In a broader sense, however, "style" simply means a repeated pattern. And, since sentences all have to be based on an S/V/C pattern, every piece of writing has style. Part of this style results from syntactic texture, and readers sense that texture unconsciously, just as we sense any other texture. A KISS Approach, however, can make many aspects of syntactic texture very easy to see and discuss.
       One of the most basic aspects of style involves main-clause length. The theory suggests that the brain processes a main clause in STM and, when given an appropriate signal, dumps it to long-term memory.  This dumping creates a rhythm that is clearly perceptible to all readers, and it is a rhythm which students can investigate scientifically by counting the number of words in a passage, counting the number of main clauses, and dividing the total number of words by the total number of main clauses.  What they will have calculated, of course, is precisely Hunt's words per T-unit. If each member of a class analyzes a sample of his or her own writing in this way, and the results are averaged, the results will probably be close to the figures arrived at by Hunt, O'Donnell, Loban, etc.  When my college Freshmen do this, as a group they almost always average within a half word of fifteen words per main clause.
        What is important about this is that it sets a basic stylistic norm. I introduce students to the statistical results of the researchers, results which show the increasing length of main clauses with age. I also point out that their readers do not count words, but they definitely do sense the rhythm as a result of their "dumping" the contents of STM to LTM and clearing STM for the next sentence. I demonstrate this by simply using sequences of numbers::

1, 2, 3, 4, 5, 6, 7, 8, dump
1, 2, 3, 4, 5, 6, 7, 8, 9, dump
1, 2, 3, 4, 5, 6, 7, dump
clearly creates a different rhythm than does
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, dump
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, dump
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, dump.
In analyzing their own writing, some of the students found themselves at the low end; some, at the high; and most, in the middle.
       I suggest to students that, on this very basic measurement of style, they probably want to be somewhere in the middle. If their main clauses are, on average, too short, they may sound like middle school students. (No college Freshman wants that.) On the other hand, if their average is too high (and some of my students have averaged 25), then their sentences may overwhelm their readers, the length making the sentences too difficult to process through the seven slots of STM. In addition, longer main clauses require, in and of themselves, more slots for processing, thereby leaving fewer slots for other purposes. An error that might cause no problem in the context of shorter main clauses may well cause a critical crash in the context of a long main clause. Not all errors are created equal.
       We are approaching the question of the psycholinguistic model and the KISS theory of natural syntactic development, but before we turn to that, let's look, for a moment or two, at the difference between the KISS Approach and the typical application of sentence-combining in our classrooms. The KISS Approach, of course, assumes that students will learn to identify all the main clauses in their own writing, and that they will be able to count words and clauses  -- and do some simple averaging. In class, I usually have students work in small groups to check each other's work. Thus, in addition to seeing his or her own numbers (plus getting the average for the class) each student has seen the analyzed writing of some classmates and has participated in the "scientific" process. They have also listened to my argument for Aristotle's "golden mean." A student who is averaging eleven words per main clause sees and understands the situation. He now has a motive for increasing the length of his main clauses. At the other end, the student who is averaging 25 words per main clause realizes that he needs to shorten them. Most of the students, however, find themselves comfortably close to the norm -- their writing is "good." The first student can certainly use some appropriate sentence-combining exercises -- and he will probably appreciate them. The student at the high end does not need them -- and he knows it.
       Contrast this situation with the typical application of sentence combining in which the entire class is given exercises -- without anyone, often including the teacher, understanding where anyone stands on the continuum of average main-clause length. As noted by Hake and Williams (discussed in the section on research), sentence-combining probably doesn't help students until they are ready for it. And, in addition to possibly not being ready for it, the students on the low end of the continuum have no idea that they are there, and they have no special motivation for doing the exercises. The students on the high end, of course, are probably more than ready for such exercises. But, because they too have no sense of where they stand relative to the norm, they are likely to produce even longer main clauses, thereby making their writing even more difficult for readers to process.
       Words per main clause is only one part of normative stylistics, and I caution students that, for example, there are advanced constructions which reduce a writer's average number of words per main clause. We don't get to spend much time on these constructions, so I usually mark them in students' analyzed samples, thereby indicating to many students who find themselves in the middle or even at the low end that there are compensating constructions in their writing. We do, however, deal with subordinate clauses.
       As I tell my students, their history, forestry, automotive, and nursing instructors probably do not know what clauses are, but they can certainly "feel" them because as readers, they have to process two or more S/V/C patterns before "dumping" to long-term memory. Subordinate clauses are a crucial element in syntactic "texture." I start by telling students to forget "grammar" and simply explain "texture" -- what causes it. They all, of course, know what the word means, and they all can sense different textures. But what makes the texture of glass (on the windows) different from the texture of the walls, or of the rug on the floor? Four or five minutes of discussion usually leads to a consensus among the students that one of the most important "causes" of texture is the difference in surface structure. Things that are rough have a lot of bumps on the surface; things that are smooth don't. We sense the bumps. I point out to the class that, although they all understood the word, and they can all sense differences in texture, they had not really given the question much thought. It took a few minutes until some students led the class to what is now a consensus. Then I suggest that something similar occurs with subordinate clauses.
       According to the model, every word and construction eventually is chunked to the words in the S/V/C slots in the pattern of a main clause. According to the model, therefore, the words in the main clause slots get the focus as everything comes "back" to them. They are the "bumps." Ideas in subordinate clauses may be important, but they are structurally pushed into the background. And just as the students could distinguish textures without understanding what causes them, so too their teachers (in all disciplines) can sense differences in clause texture without being able to consciously identify clauses. A text that is written entirely with main clauses basically presents every idea as equal to every other idea, even though such is rarely the case in terms of meaning. Writers who use subordinate clauses can push supporting material into the background, thereby creating a sharper focus on the main ideas.
       As they analyze their own writing and work in small groups to check it, some of my students have called me over to note that "There are no subordinate clauses in mine." I check and sometimes have to respond, "Yes, that's true. You'll probably want to work on that." And because they have seen subordinate clauses in the writing of most of the other members of their group, they usually do indeed want to work on it. Then, of course, there are the students at the other end of the continuum, the writers who put a subordinate clause within a subordinate clause that is within another subordinate clause. They usually call me over to ask if that is "possible." Although I have learned to expect the question, I usually give the students a rather blank stare and ask if they know the definition of a clause. They do. Do they know how to analyze clauses? They do. Then I smile and ask, "Why are you asking me if it is 'possible'? It's there, and you see it." That having been said, I shift their attention to where I think it belongs by suggesting that such a depth of embedding may cause problems for readers -- a third level embedding like that presents readers with four S/V/C patterns to process, all within short-term memory.
       Whereas sentence-combining exercises usually attempt to impose a style on students, the normative stylistic aspect of the KISS Approach provides students with tools with which they can make their own intelligent choices. Thus the ability to analyze any sentence also frees students from the stylistic quirks of their teachers. I have seen, for example, some teachers who dislike gerundives -- "Having sold all my baseball cards, I stopped collecting them." These teachers prefer the subordinate clause -- "After I sold all my baseball cards, I stopped collecting them."  I have also seen a teacher of grammar state that she does not like noun absolutes. Such stylistic preferences are fine -- if one is talking about one's own writing, but should they be imposed on one's students? In a KISS Approach, such an imposition is rather difficult for the simple reason that students are taught to (and how to) analyze the style of anything they read.

The Psycholinguistic Model and the KISS Theory of Natural Syntactic Development

       As we look at the psycholinguistic model and the KISS theory of natural syntactic development, we need to keep in mind both O'Donnell's allusion to "formulas" and the question of general cognitive development. The latter includes the level of abstraction of the writing. The student who writes "Mother said we can go if it doesn't rain" uses a conditional "if then" construction, but the elements being discussed are all simple and concrete. This "if then" should not be considered the same as the "if then" in "If train A starts from Washington at 40 mph  . . ." We can't simply count words or constructions. That having been said, I would suggest that our psycholinguistic model helps explain the research results of Hunt, O'Donnell, and Loban.
       The KISS web site includes a rather lengthy essay on "Language as a Stream of Meaning." It suggests that the stream originates from the river of ideas that is in the writer's head. Dostoevsky is reported to have said that only 10% of what was in his head ever got down on paper, but the same is true even of the weakest of our student writers. Lots of ideas flow into and out of our brains, and as we try to write, we have no choice but to select some of those ideas and attempt to embody them in words and sentences. The basic sentence pattern, of course, is the main clause -- a subject, finite verb, and possible complement. This is, of course, also the basic sentence pattern of fourth graders.
       Let's assume, for a few moments, that subordinate clauses, appositives, and gerundives are beyond the ability of the fourth grader -- that any appearances of such constructions are the result of formulas. In Vygotsky's terms, we are assuming that they are beyond these students' "zone of proximal development."  Vygotsky visualized the zone as that area between two concentric circles. The inner circle represents those ideas and concepts which the child has mastered. The area between the two circles, the zone of proximal development, represents those concepts which the student currently has the ability to learn. The area outside the outer circle represents those ideas which are currently beyond the range of the students' ability to comprehend. In math, for example, the concept of multiplication is beyond the zone of any student who cannot yet add and subtract. Once the student understands addition and subtraction, the circles grow and multiplication comes into the zone of proximal development.
       Even if subordinate clauses, appositives, and gerundives are beyond the fourth graders zone, as Hunt suggested, there is still a lot within that zone. In addition to compounding subjects, finite verbs, and complements, Hunt suggested, for example, the embedding and deletion that result in 1) a predicate adjective in one clause being embedded in another -- "The red jewel was in the drawer." 2) a prepositional phrase from one main clause being embedded in another -- "The jewel in the drawer was red" and 3) a clause with "have" being reduced to a possessive -- "I like to ride my new bicycle."  And Hunt's list was meant to be suggestive, not exhaustive. Since the research shows that subordinate clauses do not blossom until seventh grade, we can probably conclude that between fourth and the end of sixth, most students are mastering a variety of other constructions such as those described by Hunt. The point here is that even though we do not yet know enough about what students are learning, learning is certainly going on. Forcing these students to "learn" what we want them to (clauses, gerundives, appositives) probably interferes with their natural learning.
       The development of the subordinate clause appears to us to be a major expansion of the zone, or in Piaget's terms, an ascent to a higher plateau, primarily because it is easier for us to see it. When it does occur, of course, it is a major expansion because it requires the student to process at least two S/V/C patterns, as S/V/C patterns (and not as formulas), within the student's short-term memory. If we compare the processing in STM with juggling, the number of things juggled at one time at least doubles. It should be obvious that it will take some time to master this new skill. The research suggests seventh and eighth grades, but as I noted previously, a few of my college Freshmen have still not mastered it. It is not, as Hunt suggested, "the language" which "sets a practical limit on the number of subordinate clauses" (1965, p. 69). For young students, it is their linguistic, and psycholinguistic experience. (Note that, as suggested above, at the upper end of development, the limit is imposed by the psycholinguistic processing abilities of intended readers.)
       Once subordinate clauses have been mastered (in Piaget's sense of the term), natural development, as suggested in the previous chapter, turns to the appositive and gerundive, Hunt's "late" bloomers. Why are they late bloomers? Hunt perceived most of the late blooming constructions as deriving from reductions of clauses, and he was probably basically right.

They live in Endicott, which is a city in New York.
can be reduced to the appositive
They live in Endicott, a city in New York.
When they were going to the park, they saw an accident.
can be reduced to the gerundive
 Going to the park, they saw an accident.
If, forgetting formulas for a moment, we assume that all of these constructions derive from the reductions of clauses, then they obviously must develop after the student masters the subordinate clause.
       When we put formulas back into the equation, we see still another reason for the delay in their development. A formula, remember, is basically a string of words which students learn as a whole. But how many appositives and gerundives do students see? Although the figures are still open to question, it is generally believed that appositives and gerundives are more typical of writing than of speech. Whereas students can "pick up" subordinate clauses in speech, they are much less likely to assimilate appositives and gerundives in the same way. And if we look at what they read, subordinate clauses are much more frequent than are appositives and gerundives. In my web site analysis of the openings of six novels, the writers average 72 subordinate clauses for every 100 main clauses, but only 16 appositives and 18 gerundives. Presented with many more examples of the subordinate clause, students obviously learn to develop it first.
       If we return now to our psycholinguistic model, we can see still another reason for the late development of appositives and gerundives. By sixth grade, students are masters of the basic S/V/C pattern. They can elaborate it in a large number of ways, but almost all of this conceptual elaboration (as opposed to formulaic) is in the form of compounding, or of adding adjectives, adverbs, and prepositional phrases. The mastery of subordinate clauses involves taking one whole S/V/C pattern and embedding it in another with the use of a subordinate conjunction. (Note that in sentences such as "He said she was pretty," the subordinate clause is probably formulaic -- children hear "He said " "Daddy said .." all the time.)  I do not mean to underrate their achievement, but what is involved here is taking one fully mastered construction (a main clause) and embedding it, in its entirety, in another fully mastered construction (another main clause) by adding a connecting word.
       But appositives and gerundives involve exactly the opposite mental process. Not only do they involve reduction in addition to embedding, but they also result from the elimination of connecting words. In an appositive, all "syntactic" connections (prepositions, conjunctions) are eliminated, including the pronoun in the "contributing" sentence: "They live in Endicott, which is a city in New York." In the appositive, and only in the appositive, the syntactic connection is created entirely by the meaning of the words. Compared to everything that the students have already mastered, the appositive presents a radically different type of embedding. Gerundives present students with a comparable, but perhaps less radical type of problem. The reduction involved in the gerundive includes the connecting words, and part of the original finite verb essentially changes its function, becoming an adjective: "When they were going to the park, they saw an accident."
        As reductions of clauses, appositives and gerundives save writing time. When the mind is faster than the hand, the hand has to try to catch up, especially when STM is full. When STM is not full, there is no reason to rush the hand. Again we need to think about the river of ideas in the writer's head and the stream of meaning (sentences) that the writer is trying to embed selected ideas into. All of this is going on within the writer's seven-slot STM. (Studies, including some by my own students, have shown that STM does not grow with age.) As Hunt, and all the other researchers have noted, older students write more -- they have more to say. The river in their heads gets deeper and faster. More and more words get packed into main and subordinate clauses, and, as these clauses grow, they reach a point at which it is difficult to hold all the words in STM while the hand puts them onto the paper. At this point in their development, students find appositives and gerundives to be useful tools; before this point, students are probably much more comfortable when they include all the connecting words that these constructions delete.
        If we step back and look at a general view of the psycholinguistic model and natural syntactic development, we can probably say that there are three natural laws that govern the process:

1. As students grow older (and have more to say), there is a natural tendency to combine more ideas into one main clause.

2. Much of that combining involves embedding of ideas from one clause into another, and, at about seventh grade, the development extends to embedding one entire clause (as a subordinate clause) into another main clause.

3. Throughout the process, the combining and embedding involve reductions.  Before the development of the subordinate clause, most of the embedding and reduction results in the addition of simple adjectives, adverbs, and prepositional phrases; after the development of the subordinate clause, this process extends to appositives and gerundives.

There is, of course, still a lot that we need to learn about natural syntactic development, particularly in the area of formulas. The process is complex, and we need to be careful in the way that we teach. The first rule of doctors, however, is to do no harm. We need to be sure that, in intervening in this process of natural development, we too do no harm.

        As I suggested earlier in this chapter, the model presents a dynamic view of grammar that we can present to students. The model justifies the "rules," and it can be used to give students some basic insights into style. The model also gives the study of grammar a clearly defined, relatively easily attainable goal -- the students' objective is to understand and be able to discuss how any word in any sentence is "processed" by readers. The grammatical concepts that students need to learn in order to understand and use the model are the subject of Part Four. Taken all at one time, they may appear to be complicated, but if they are studied in a KISS approach, they are not as difficult as they may appear.