Precursors

In the decades before the first full system of diagrams, a number of authors experimented with various ways to visualize grammatical relationships. None of these systems were particularly influential, but they all show early instances of teachers grappling with the same problems that would later give rise to true diagrams.

---
(This page was last updated October 17, 2015.)

James Brown

The earliest, and certainly the most innovative, of these experimenters was James Brown, whose system of "American grammar," attempted a complete break with the terminology of English grammar—or indeed with that of any other system ever devised. A cursory glance at the pages of Brown's works provide the reader with definitions like the following:

II. THE CLAD ORDER,

Is a class of syllabanes, or ne-syllabanes, which is made up of those monos that have a branch dependence upon some other mono, or monos in the poetrone;
(J. Brown, 1847, p. 41).

It easy to see why Goold Brown (1862) dismissed the other Brown's work as "fantastical" (p. xiii). Such bizarre terminology, combined with his flagrant self-promotion and his captious attacks on rival grammarians can easily incline a reader to dismiss Brown as a mere crank. Yet Görlach's (1998) dry suggestion that "it would be worth studying Brown in case he is more than a productive eccentric" (p. 64) is well taken. Brown stands apart from the mainstream of nineteenth-century grammar, but some of his criticisms, such as the inadequacy of traditional definitions, are reminiscent of critiques by contemporary linguists.[4]

Most relevant to a consideration of diagrams, however, is Brown's treatment of constituency, or "scanning," as he calls it in J. Brown (1831).[5] Brown uses a system of brackets and numbers printed above words to show structural relationships among words. He divides up a sentence into "sections", as in the following example from J. Brown (1831), p. 36:

1   1 1 1     1 1 1  
[The sun shines] (upon all men) (who will receive his rays)
    1 1 1  
(which he sends) (from the heavens.)

Brown uses square brackets to denote "major sections," that is, word groups which he takes to be independent, and he uses parentheses to denote "minor sections," that is, word groups that depend on another unit. He uses numbers above individual words to distinguish "major words," which are left unmarked, and "minor words," that is, those that depend on another word. He also recognizes that these dependency relationships can be nested to an arbitrary extent, and uses increasing numbers to show the depth of the nesting:

4 3 2 1
Very much too cold weather.

In other words, with this notation, Brown indicates that very modifies much, which in turn modifies too, which modifies cold, which modifies weather. He uses similar numerical notations to show the dependency arrangements in the sections above: "the sun shines (upon all men1) (who will receive his rays2) (which he sends3) (from the heavens4.)"

As these examples illustrate, Brown was groping towards a way to illustrate syntactic structure. But his peculiar theory comports neither with grammar as it was understood in the nineteenth century nor with contemporary linguistic notions of phrase structure. The analysis here assumes that the root of language consists of nouns, and that all other words are
modifications of those fundamental terms. Thus, the major words in the constituents above are always nouns. (He groups pronouns and nouns in the same category.) For example, in the section "who will receive his rays," who and rays receive equal rank. In other words, the direct object, "his rays" is treated as a first-order constituent, on the same level as the subject. This treatment explicitly flips on its head the notion that the object of a verb is a dependent of that verb. A similar point can be made for his treatment of preposition phrases. Moreover, despite his use of numbers to annotate a degree of dependency, the way Brown deploys his brackets suggests that he remains stuck in a word-by-word way of thinking about structure. His sections simply divide up the sentence into groups one after the other rather than recognizing that one unit actually nests within another. In other words, he takes one section to depend on another section as a whole, rather than on a specific word in the section.

Brown does elaborate this system slightly to indicate the direction of the reference when there is more than one noun, and to handle compound relationships where a word refers to more than one other word at the same time, but the analytical shortcomings of the scheme, quite apart from his fondness for renaming all the conventional terms of grammar, ultimately doom this system to be a dead-end.

---
(This page was last updated Oct. 17, 2015.)

Notes

4. For an extensive discussion of Brown's overall work and its place in the intellectual currents of the time, see West (2000).

5. This work, although it has its own curious terminology, is relatively free of the dense neologisms of Brown's later works.

Barnard's Analytic Grammar

Brown was so caught up with revising the technical aspects of English grammar that he paid little attention to pedagogy. His methods are generally traditional, even if the scheme he taught was entirely new. Altogether more innovative in his use of new pedagogical techniques was another writer from the 1830s: Frederick A. P. Barnard, the tenth president of Columbia College (now Columbia University) and the person for whom Barnard College was named. In 1836, Barnard published his Analytic Grammar with Symbolic Illustrations, a work that can lay claim to be the first to employ graphical symbols as a way of recording and visualizing grammatical analysis.

Barnard's grammar achieved little popularity outside the community of educators for the Deaf. Whereas popular grammar books of the day would go through dozens, or even hundreds, of editions, Barnard's book exists only in a single edition.[6] Today, it has become entirely obscure, and the schemes of sentence diagramming that come to dominate the later half of the nineteenth century appear to be uninfluenced by his work. Nevertheless, the Analytic Grammar is notable not merely because it precedes other diagramming systems but because it also reflects the transition in grammatical pedagogy that was occurring at the time. In it we can see Barnard not only attempting to provide visual representations for the old accidence-based parsing but also groping his way towards a hierarchical conception of language, expressed in visual terms. And along the way he proves to be a leader in other pedagogical techniques, such as linking the study of grammar to composition.

Barnard's method involved annotating sentences by placing a symbols above each word in the sentence. These symbols convey information not only about parts of speech but also about concepts such as case and tense, as well as certain semantic relationships.

Example of Barnard's Symbols

Earlier grammarians had used interlinear glosses to convey certain grammatical information. Lowth (1764), for example, used numbers over words to indicate the parts of speech in the eighteenth century. But Barnard credits the idea of representing the grammatical features of a word visually to the inspiration of Roche-Ambroise Sicard.

Sicard ran the the first public school for the deaf, the Institution Impérial des Sourds-Muets, as it was known in 1808, when he published his Théorie des Signes pour l'Instruction de Sourds-Muets, describing a system of signs that was used at his school.[7]

Sicard's signs were logical source of inspiration for Barnard, who himself had grown increasingly deaf as a young adult and who worked, at the time he wrote the Analytical Grammar, as a teacher at the New York Institution for the Instruction of the Deaf and Dumb (now known as the New York School for the Deaf). These signs, though, served more as a general inspiration for Barnard than a specific road-map.

Sicard's system is not actually a record of French Sign Language. Rather, it is a method of teaching deaf students how to read and write French when they may possess only a rudimentary home sign language. The bulk of the Théorie des Signes constitutes a dictionary, but the keywords in that dictionary are simply French words. The gestures which constitute definitions are closer to pantomimes meant to convey a definition than true signs in a full sign language. Sicard decomposes words into semantic primitives, each of which is conveyed through a gesture that is meant to be iconic. The result can be an extensive series of signs—far too many for practical, day-to-day conversation—that include gestures not only for a word's core meaning but also for a wide variety of grammatical features, including signed equivalents for both derivational and inflectional morphology.

The process of semantic decomposition and the effort at iconicity can both be seen in Sicard's treatment of grammatical concepts. For example, Sicard's method for expressing what he calls the absolute past (the passé composé) is as follows:

on jette la main droite par-dessus l'épaule, derrière soi, pour indiquer que l'action qu'on énonce n'est ni devant soi, ce qui annonceroit qu'elle est présente; ni au devant de soi, à quelque distance de soi, ce qui annonceroit
qu'elle est future.

(the right hand is thrown back over the shoulder, indicating that the action described is neither immediately in front of us, which indicates that it is the present, nor ahead of us at some distance, which indicates that it is the future.)
Sicard (1808), v. 2, p. 574.

His signs for the passé simple, which he calls the "first relative past," are the following:

Le signe du 1er passé relatif se compose de deux signes : 1°. Du signe du passé absolu (qui est aussi le signe de tous les passés). 2°. De l'indication de l'époque, qui se fait en avançant l'index devant soi, comme pour indiquer un objet quelconque.

(The sign for the first relative past consists of two signs: (1) The sign of the absolute past, which is also the sign of all the pasts. (2) The indication of the time, which is done by advancing the index finger in front of oneself, as if to indicate some object or other.)
Sicard (1808), v. 2, p. 575.

This method of dealing with tense clearly shows that Sicard bases his signs on a semantic analysis of time relationships, even if he takes the grammatical distinctions made available in French as his starting point. Notably, his signs bear no direct correspondence to language-specific morphology. Hence, the absolute past, a two-word verb form in French (e. g.,
j'ai parlé), is taken as the semantic base from which to build other expressions of past time, whereas the first relative past is treated as a derived form, even though its French expression is a one-word form (e. g., je parlai). Nor does Sicard distinguish between the different verb classes. Signs for verbs like parler, venir, and rendre all receive the same additional gestures to mark tense, mood, person, and number.

It is the primarily the concept of representing abstract grammatical features in an iconic form that Barnard takes from Sicard's work. Barnard does not attempt to draw Sicard's gestures, but like Sicard, he attempts to find symbols that bear some transparent semantic relationship to the concepts that he is trying to annotate. Also following the lead of Sicard, he builds up his symbols compositionally, starting with basic elements and gradually adding more and more features to encode other concepts, a point that he is at pains to stress in his preface:

Should the variety of forms which the symbolic characters assume, appear, at the first glance, to render the subject complicated, it may be worth while to suggest that all these forms spring from the combination of elements, exceedingly simple in their nature, and in number only six, viz. :---

Barnard's Basic Forms

These elements are separately introduced, at intervals, and gradually modified and combined, to meet the growing wants of the learner.
Barnard, 1836, p. iii.

How Barnard elaborates upon these basic elements, and his attempt at iconicity, can be seen in his treatment of nouns. The resemblance to stick figures of people is not accidental. His basic representation for a substantive, a vertical line, is first expanded by adding little feet to the bottom, "[t]o give it an appearance of stability" (Barnard, 1836, p. 24). The little arms extending from the nominative and objective forms of the noun symbols are meant to suggest action proceeding from the subject and moving towards the object. The possessive form adds the basic symbol for the attributive function to that for a noun. In other words, it conveys Barnard's analysis of the possessive noun as a word that is both a noun and a modifier. Plural forms are indicated by adding two ticks to the basic forms.

Barnard's Paradigm for Nouns

The number of features that Barnard encodes in his symbols can become quite elaborate, particularly when he treats the verbs. Unlike modern theories that also enumerate grammatical features explicitly, for example head-driven phrase structure grammar, Barnard normally encodes only the forms he regards as marked. Thus plurality can be indicated for both nouns and verbs, but only the plural forms modify the basic symbol.

Barnard's explicit discussion of gender treats both masculine and feminine nouns as marked. He distinguishes between masculine, feminine, neuter, and common-gender nouns. In terms of his symbols, the symbol for a neuter noun is the most basic, consisting of a single vertical line. The masculine noun is shown by the addition of the two feet, as shown above. For feminine nouns (e. g. actress), the feet are replaced by a horizontal line. The symbol for common nouns joins the modifiers for both masculine and feminine nouns, forming a triangle beneath the vertical line (Barnard, 1836, p. 85). In practice, however, Barnard regards these distinctions as of little value and uses the masculine form as the default for all nouns. He also does not distinguish gender in third-person singular pronouns. He, she, and it all are given
the same symbol.

The most intriguing moments in Barnard's work are when he starts to consider how to use his system to represent different levels of constituent structure. In the diagram reproduced below, Barnard represents structure in a series of bracketed groupings that look, at least superficially, as if they anticipate tree diagrams. The numbers at the left of the figure indicate successive stages of grouping in which words are progressively combined into larger and larger units, building from the
bottom up.

Barnard's Constituent Structure

One important reason that Barnard's figure reminds a modern reader of a tree diagram is that it maintains the linear order of the words in the sentence, and the figures are grouped together on that basis. But once we start to look at the details, there are many differences. To better appreciate the ways in which his drawing does and does not anticipate a phrase-structure diagram, I have translated it into format will be more familiar to readers who are used to reading tree diagrams.

Barnard's Constituent Structure Translated

In the figure above, I have replaced Barnard's symbols with labels that reflect the part of speech and the marked features that the symbol encodes. Because he has no symbol for a clause, Barnard stops once he gets to the level of what we would call the subject and the predicate, although his symbols suggest that he actually thinks of these constituents as an expanded noun and verb, respectively. Indeed, Barnard does not have any consistent concept of a phrase as found in modern linguistics. As these symbols suggest, he treats part-of-speech labels as equally applicable to individual words as well as to larger units. Thus, for example, "in disposition" is labeled an adverb, because it modifies the adjective mild. This analysis is more or less identical to that adopted by later schoolbook grammars, even when they do use the term phrase.

In many other ways, however, the details of this translation show that Bernard was not thinking of phrase structure in any rigorously hierarchical way. For example, he marks the relative clause "who is mild in disposition" as an adjective. This distinction is in keeping with the mainline tradition of nineteenth-century grammar. But he shows this clause as consisting of two components: "who is mild," an adjective, and "in disposition," an adverb. In reality, "mild in disposition" is a single constituent, the complement of is, and "in disposition" should truly be shown as a dependent of mild. The lack of any symbol for a clause may be part of the problem here. Similarly, "very many" and "friends" are not shown as forming a single constituent, the object of make.

With such issues, it's difficult to decide if Barnard has simply made a mistake by his own standards or if the figure truly reflects his own best understanding of constituency. Given the paucity of discussion of the issue in previous grammar books, it is unsurprising that he would struggle to achieve a clear conception of such details of sentence structure.

---
(This page last updated Oct. 18, 2015.)

Notes:

6. Within schools for the Deaf, however, Barnard's methods had a more lasting impact. His system was used in New York through at least the 1860s, and is mentioned as being taught in Michigan in 1873.

7. This system is typically credited to Sicard's predecessor at the institute, Charles-Michel de l'Épée. Sicard, however, suggests that it underwent major elaborations after Épée's death, extensions that made it much more effective. So the system that Sicard presents is not merely the one that Épée devised. For simplicity, I will refer to this system as Sicard's, but it would be most accurate to say that it was the result of long collaboration among many teachers and students.