I

Robert Frost’s famous comment that “writing free verse is like playing tennis with the net down” makes a big rhetorical splash but badly misstates the logic of free verse composition. It makes no allowance for the poet’s ability to fashion his own challenging and effective rules for the creative “game.” The analogy functions far better if we employ it instead to combat the dangerous misconception that one might reasonably undertake a technical stylistic analysis without determining in advance the constraints proper to that particular activity. In a pursuit whose goal is, I have argued, an accurate as well as formally efficient description of a text’s language, the rules are “constitutive”; they do not “regulate” the discipline, to paraphrase Jonathan Culler, “so much as create the possibility” of its succeeding at all.¹ Altering or overriding the rules by brute force, therefore, could easily transform what began as a tennis match into a joust or even a purposeless melee.

In view of this danger, it is surprising to find that few books and articles in the recent literature offer clear methodological guidelines to the aspiring analyst. Many stylists merely assert, rather than demonstrate, the relevance of transformational or structuralist grammars, case grammar, or speech-act theory, to their particular texts. Where the language of those texts happens to include syntactically marginal or nonstandard forms, furthermore, they readily take a wide variety of theoretically extreme steps in their efforts to account for the perceived deviance.

Stylists who have shown any interest in justifying their analytical methods have usually claimed that their interpretive ends excused whatever extraordinary technical means they invoked. Too superficial a reading of such literary critics as John Reichert may appear at first to lend this laid-back approach a certain air of respectability. Reichert suggests that pragmatic justification must ultimately figure in our evaluation of any critical argument:

To regard criticism as aiming not just at true statements about, but at new perceptions of a literary work, is to admit among the others a purely pragmatic test for its success. . . . That is, one test of an interpretation is whether it actually provides the reader with the new way of perceiving that it is intended to provide him with.²

As will become clear in chapter 4, I have a great deal of sympathy for Reichert’s position when it comes to considering the critical (that is the technical or perceptual and the interpretive) act as a whole. But we must take great care not to confuse the task of interpreting a text with that of analyzing its linguistic form.³ Reichert’s suggestion that we aim “not just at true statements” about a text implies that truth as such does indeed have a place in literary studies; that place would seem to me to be, if anywhere, at the technical level of stylistic analysis. Here, objectivity and accuracy are at a premium, the careful formulation of adequate methodological “means” thus demanding the most careful attention.

I will not belabor this point, since Mary Louise Pratt has already argued it with great vigor. Employing an analogy from the workbench rather than the tennis court, she insists that the right to apply in stylistic research the methods of formal syntax brings with it attendant responsibilities. All too often, she complains,

the poetician is left with no theoretical obligations to linguistics. He is set free to raid the linguist’s toolbox at will and to use what he finds in whatever way he likes.⁴

In this chapter, then, I shall take as read Pratt’s arguments and those of others who have asserted the stylist’s “theoretical obligations to linguistics,” focusing instead on the logically subsequent task of defining those obligations as precisely as possible. I shall take as my point of departure a review of some specific practical problems that arise if one adopts a recklessly laissez-faire attitude toward the technical side of stylistics. My solutions to those problems will in turn motivate certain more general “rules of thumb” concerning proper and improper ways of developing stylistic analyses. Finally, I shall particularly emphasize the need for the stylist to develop a controlled sequence of analytical steps by which he will proceed in order to ensure the richest possible technical account of each text he examines.

II

The stylist’s first obligation as an analyst of technique is to make every possible effort to account for the syntactic form of his text as a sample of, rather than as an exception to, the language in which it was written. Such a description of our goal should not be taken to imply that stylistics never involves itself with devising nonstandard grammars; I shall turn to that topic in due course. It does entail that every effort should be made in the first instance to avoid such a step by analyzing a poetic text using only the methods and terms accepted by linguistic science for describing the standard language. As Pratt again points out, we should guard against any step that

obscures the real relation that holds between poetic utterances and the grammar of the language in which they are written,. . . the relation that all utterances in a given language hold with respect to the grammar.⁵

Whatever the temptation to innovate, the primacy of this fundamental relationship must be preserved.

Such a principle is fully consonant with Paul Kiparsky’s observations in his eminently reasonable essay “The Role of Linguistics in a Theory of Poetry.” From his study of linguistic behavior in languages around the globe that is considered “poetic,” Kiparsky concludes that “the linguistic [elements] which are potentially relevant in poetry are just those which are potentially relevant in grammar.”⁶ All that I now propose is the correlative procedural guideline: stylists should curb their more creative linguistic urges so that the analyses they propose will in fact fall within the scope of Kiparsky’s observational generalization. Indeed, I would further endorse the slightly narrower prescription that stylistic analysis initially employ not only linguistically salient “building blocks” (the metaphor is Kiparsky’s) but also the protocols and patterns for combining those blocks dictated by everyday usage. Where Kiparsky, then, detects “homologies” between standard and poetic syntax,⁷ I would mandate identity, at least where the literary data are compatible with such an approach.

The necessity for so strong a procedural requirement must, of course, be established on purely theoretical grounds by producing evidence that a failure to respect its provisions may result in major critical oversights. At the same time, we may note in passing that it corresponds to a highly common-sense view of the reader’s strategy as he approaches a text. Equipped virtually since birth with an acute facility for processing the language around him, why would the reader choose to reject that facility in reading poetry unless some aspect of its linguistic form actively forced him to do so? Put this way, the moral seems self-evident; yet as a methodological principle it is easily lost to view.

An important decision facing even the stylist who accepts my first point about technical analysis involves the choice of a specific syntactic theory within which to proceed. I remarked earlier that all empirical hypotheses, including hypotheses about human language, are vulnerable to scientific progress and destined for obsolescence in the long run.⁸ I noted too, however, that in the short run they still represent conscientious attempts to explain “the truth” about the phenomena they describe. A discredited theory is thus in an important sense “wrong” and the stylist has an obligation to rely for his technical material only on those theories of language that have not been disproved. I find it hard, for instance, to imagine a strong stylistic argument based on the once-important contrast in American structuralist linguistics between “exocentric” and “endocentric” constructions, at least as it was originally formulated by linguists such as Leonard Bloomfield.⁹ Advances in linguistic theory have made it difficult to accept as fundamental a binary categorization of syntactic structures that places the constructions in the house and if John ran away in one category, boys and girls and very fresh milk in the second. If he seeks an accurate and revealing account of the language of his poetic text, the stylist will certainly need to avoid such obvious anachronisms and work only with the most recent analytical tools.

For the present, I would contend, transformational generative grammars still offer the most widely accessible, linguistically respectable account of syntactic structure and syntactic processes. Most useful of these grammars, furthermore, is that version of transformational syntactic theory developed in papers and books by linguists such as Joan Bresnan, Noam Chomsky, Joseph Emonds, Ray Jackendoff, and John R. Ross.¹⁰ This is not to pretend that the set of assumptions from which all these authors may be said to have taken their lead and to which each has in turn contributed—the so-called Extended Standard Theory—has been without its critics. As Bresnan rightly points out in discussing “the familiar Aspects model,” “subsequent research has shown that each of [the] properties of the model is incorrect,” and, inevitably, competing theories—“Montague Grammar”; Gazdar, Pullum, and Sag’s “Generalized Phrase-Structure Grammar”; and Bresnan and Kaplan’s own “Lexical-Functional Grammar”—have arisen in response to these perceived inadequacies.¹¹ Nor, of course, can the theoretical linguist afford to undervalue even the most limited of these divergencies between theories:

How can we choose among these various theories? From a broad philosophical perspective, it is not necessary to do so. . . . But from a scientific perspective . . . we must find or construct the best theory.¹²

When we as stylists seek to apply linguistic theory to poetic texts, however, a second, more pragmatic consideration must also influence our choice. Ironically, it is again Bresnan who offers the most realistic rationale for adhering pro tempore to the Extended Standard Theory (EST), even though it has come under such heavy fire:

[T]his is still the picture that many linguists and psychologists [and critics] have of a transformational grammar, a fact that attests both to the intuitive appeal of the model and to its enormous fruitfulness in guiding linguistic research.¹³

As my argument in the preceding paragraphs dictates, should one or another of the theories now in open competition with the EST finally succeed in usurping its current position as the linguistic orthodoxy, stylists will have to adjust to its modes of analysis and description. For our present purposes, however, a very slightly modified version of the EST would seem still to offer an excellent blend of the twin virtues of easy familiarity and technical respectability.

As detailed analyses of particular passages necessitate, I shall in the course of this study elaborate on specific aspects of the EST. A broad overview at this point, however, may help to keep such observations in their proper perspective. Modern syntactic theory, then, hypothesizes several components to a grammar. Taken together, these components describe the numerous and complexly interlocking regularities that constitute a native speaker’s syntactic competence. Every lexical item in the language, for instance, is assumed to belong to one of a small number of lexical categories each of which is in turn defined by the broadly similar behavior of its members in equivalent syntactic contexts. Each lexical category is assigned a label [Noun (or N); Verb (V); Adjective (ADJ); Adverb (ADV); Preposition (P)] and a set of phrase-structure rules then describes permissible combinations of these categories in the language. For them to achieve this description economically, it turns out to be necessary that the phrase-structure rules combine lexical categories into phrasal units [Noun Phrase (or NP); Verb Phrase (VP) etc.] before using those intermediate syntactic structures themselves to build clausal and sentential constructions (S). The phrase-structure rules thus generate (or define) a hierarchical structure for each sentence loosely equivalent to that assigned in some previous theories by “immediate constituent analysis.”¹⁴

Idiosyncratic information that determines for each individual word its appropriate syntactic and semantic context is contained in that word’s subcategorization and selection restrictions respectively. These formal conditions govern the insertion of lexical items into the abstract structures generated by the phrase-structure grammar (lexical insertion), ensuring, for example, that the verb put will appear only in sentences where it is followed syntactically by both a nominal object and a locative prepositional phrase (“Sherlock Holmes put a briar pipe into his mouth”) and that the verb enjoy will be used only when its subject displays the semantic property of being human or at least animate.

All of the rule types described in the preceding two paragraphs combine to generate theoretical constructs called deep or underlying syntactic structures. These structures may be portrayed diagrammatically in various formally equivalent ways. Linguists often use labeled bracketings, as in the notation

[S[NP[N Bill]] [VP[V sat] [PP[P down]]]].

But so-called tree-structures are often easier to appreciate at a glance and will be widely used in this study. A deep-structure tree for the sentence Bill sat down might look like Figure 1.

Deep structures in turn serve as input to a set of transformational rules which alter them in very narrowly restricted ways to derive further intermediate structures. Since both deep and intermediate structures are subject to transformational alteration, a full syntactic derivation may consist of a string of several structures each related pairwise with either one or two other structures by specific transformational rules. Thus to derive the sentence Down Bill sat! we need to apply at least one transformation to a deep structure identical to that for Bill sat down. The Prepositional-Phrase (PP) Fronting rule, to be precise, will move the post-verbal PP to the beginning of the sentence, giving the surface structure in Figure 2.¹⁵

The syntactician manipulates lexical, phrase-structure, and transformational components of the grammar to devise an overall system that is elegant in that it accurately and with maximum economy reflects relevant generalizations about the language’s structure. In many cases, this may involve positing deep or even intermediate structures that are noticeably “abstract,” far divorced from what we would immediately recognize as an English sentence. The permissibility of such abstraction, when linked with a number of other formal considerations, makes transformational grammars extremely powerful tools. Transformational syntacticians realized early in the evolution of this theory, therefore, that restrictions would need to be placed on the operation of transformations if they were not to “overgenerate” wildly, producing not only structures commonly attested in the language, but also a wide variety of highly counterintuitive constructions. First came the imposition of a specific ordering on transformational rules; it was shown that, unless applied in this invariant order, transformations would predict hopelessly unacceptable sentences. Next linguists discovered that, in deriving sentences that contained more than one clause, it was necessary not only to apply the rules in that same order, but also to repeat the full cycle of rules for each clause separately (beginning with the clause most deeply embedded).

FIG. 1

FIG. 2

Even this additional restriction, however, did not account for a wide variety of cases in which sentences predicted to be grammatical by the transformational component failed to satisfy common-sense criteria for acceptability. A set of constraints and conditions on transformations was therefore proposed in the late 1960s, the investigation of which quickly became, and has since remained, a major focus of syntactic research. These constraints and conditions formalize restrictions that the language imposes on all of its major transformational operations, and may be written so as to affect either the formal statement of the transformational rules themselves or their mode of application to particular kinds of syntactic structures.¹⁶ Some even define the acceptability of a given surface structure after it has been through the transformational cycle.¹⁷

The transformational syntactician’s ultimate goal in all of these theoretical additions and revisions has been to devise a set of rules, suitably constrained, that will apply to deep structures to derive a set of surface structures corresponding exactly to “all and only” the acceptable sentences of the language under study.

As may be clear from this summary, transformational grammar possesses enormous explanatory power, power whose existence worries the linguist, whose constant attention has been devoted to taming it.¹⁸ It offers a correspondingly strong temptation to his colleague in stylistics who is often unaware of the theoretical monster he is capable of fashioning. With its aid he can easily describe a wide variety of syntactic structures—real or imagined, thinkable or absurd, “poetic” or “prosaic.” Indeed, were he really free to “raid the linguist’s toolbox” as Pratt fears, the stylist could devise a hundred different ways to account for any given line of his text. Rather than dazzling us, then, the almost boundless capacities of the linguistic model that we rely on should instead alert us to the very real dangers of deviating from the relatively safe and well-trodden ground of a specific grammar of English when beginning the stylistic analysis of a poetic text.

III

The particular kind of blind spot that may develop if the stylist pays insufficient attention to the demands of the standard grammar when developing his argument is well illustrated in the extensive literature devoted to E. E. Cummings’ “anyone lived in a pretty how town.”¹⁹ (For those who may be unfamiliar with it, I have included a text of this ballad on a separate page.)

Any number of technical problems confront the stylistic analyst approaching this poem. Of these, one major concern will be the formal account to be given of Cummings’ use of the word anyone, which occurs in lines 1, 6, 16, 25, and 31. The specific cause for this concern will be the fact that a typical syntactic account of the English language bars the word any and its derivative compounds such as anyone and anywhere from appearing in certain clearly defined types of sentence. As Otto Jespersen explains,

Any indicates one or more, no matter which; therefore any is very frequent in sentences implying negation or doubt (questions, conditions).²⁰

Indeed, any and its derivatives turn out to be almost invariably unacceptable in simple declarative sentences unless those sentences contain a negative word such as not, noone, or never:

*I went out on any occasion.²¹
I did not go out on any occasion.

With these examples before him, the analyst finds himself forced to rule Cummings’ constructions in lines 1, 16, and 25 grammatically deviant by the standards of everyday usage:

anyone lived in a pretty how town

anyone lived in a pretty how town

(with up so floating many bells down)

spring summer autumn winter

he sang his didn’t he danced his did.

Women and men (both little and small) 5

cared for anyone not at all

they sowed their isn’t they reaped their same

sun moon stars rain

children guessed (but only a few

and down they forgot as up they grew 10

autumn winter spring summer)

that noone loved him more by more

when by now and tree by leaf

she laughed his joy she cried his grief

bird by snow and stir by still 15

anyone’s any was all to her

someones married their everyones

laughed their cryings and did their dance

(sleep wake hope and then) they

said their nevers they slept their dream 20

stars rain sun moon

(and only the snow can begin to explain

how children are apt to forget to remember

with up so floating many bells down)

one day anyone died i guess 25

(and noone stooped to kiss his face)

busy folk buried them side by side

little by little and was by was

all by all and deep by deep

and more by more they dream their sleep 30

noone and anyone earth by april

wish by spirit and if by yes.

Women and men (both dong and ding)

summer autumn winter spring

reaped their sowing and went their came 35

sun moon stars rain

Copyright 1940 by E. E. Cummings; renewed 1968 by Marion Morehouse Cummings. Reprinted from Complete Poems 1913–1962 by E. E. Cummings by permission of Harcourt Brace Jovanovich, Inc.

anyone lived in a pretty how town
anyone’s any was all to her
one day anyone died

Indeed, the problem runs deeper than this, since an additional constraint on the use of any- in day-to-day contexts requires that the negative word that forms so vital a part of its immediate context also appear to the left of (that is before, not after) the word any- itself:

I went out on each occasion.
On each occasion I went out.

I did not go out on any occasion.
*On any occasion I did not go out.

Although, as the first pair of sentences demonstrates, PP-Fronting can routinely prepose phrases structurally analogous to on any occasion, such movement is prohibited, probably by a simple filter operating at the level of surface structure, if it would result in a surface misalignment of any- and its associated negative element. Yet Cummings has chosen in lines 5 and 6 to overlook or transgress this constraint:²²

Women and men . . .
cared for anyone not at all.

This choice appears all the more significant upon reflection, since a syntactically more standard alternative involves only the most minor adjustment to the text:

Women and men . . .
did not care for anyone at all.

The task that confronts the stylist here is one that recurs constantly in subtly different forms whenever he involves himself with technical analysis. A formal statement needs to be made about a construction which the standard grammar will not generate and which we should therefore label unacceptable, deviant, or ungrammatical. In order to appreciate fully some of the many options available to the stylist under these circumstances, let us briefly examine the work of two scholars, Samuel R. Levin and James Peter Thorne, both of whom have mapped out plans of attack for the technical analysis of this particular poem that are prima facie worthy of serious consideration.²³

Levin proposes that, in analyzing a deviant syntactic construction from a poetic text, we should begin from the rules of the standard grammar, adjusting them as little as is necessary to generate the attested material:

[We] assume, first of all, that the grammar will not generate them. . . . We then ask how the grammatical rules can be fixed so as to generate the sentence in question.²⁴

He further suggests that such adjustment can usually be accomplished “in one of two ways: a new rule may be introduced; or items may be shifted from class to class.”²⁵ In his paper, Levin applies this method exclusively to line 4 of Cummings’ poem and in particular to the phrase “he danced his did.” But we can extrapolate from that discussion to determine the approach he probably would adopt to our own current problem, the appearance of anyone in syntactically nonstandard contexts.

One of Levin’s two proposed methods of adjusting the grammar would here involve eliminating from the syntactic features accompanying the word anyone all hints that it differs from other indefinite pronouns such as someone or everyone. This step would effectively “shift” anyone as an element in the language of this particular poem from a “class” of its own into a broader class of pronouns with fewer restrictions on their syntactic placement. Alternatively, the stylist might select Levin’s other option and adjust a “rule” of the standard syntax when writing a formal description of Cummings’ work. In this particular case it would not be appropriate to “introduce a new rule”; instead one would simply drop from the grammar the troublesome filter that, in standard English, determines the acceptability of sentences containing any- on the basis of certain negative words’ presence and relative position in the syntactic context. But whether we choose to adjust the class membership of anyone in the lexicon or to relax the constraint on its appearance in surface structure, it should be noted that the final outcome will be much the same: anyone will be ruled fully acceptable in this specific poetic context by virtue of Levin’s tampering with (“fixing”) standard syntactic conventions. It is this conclusion, with its implicit assumption that syntactic deviance should be explained away by the stylist, rather than simply observed, that seems to me a dangerous one. To be sure, Levin subsequently demands that the stylist compare the extended grammar that he has devised for some work of literature with the standard grammar to assess the effects of his adjustments. By then, however, as I shall shortly show, the damage will already have been done, since the risk of oversight exists as soon as the standard grammar is overridden in favor of an extraordinary syntactic analysis.

Thorne roundly rejects any approach to the technical analysis of poetic syntax which, like Levin’s, takes the standard grammar as its basis. He argues that a poetic text should instead be treated as pristine linguistic data, as “a sample of a different language, or a different dialect.”²⁶ The stylist, he contends, should devise categories and rules whose only justification is that they best characterize the language of the text currently at hand. The syntax that Thorne develops for Cummings’ poem thus involves lexical categories and, by implication at least, phrase-structure rules wholly foreign to the standard language. To give just one example, Thorne’s “Class A” words are characterized by three syntactic features: they appear only in object position; they are always preceded by possessive pronouns; and they never appear with adjectival modifiers. The class thus defined includes the words did and didn’t from line 4, joy and grief from line 14, and dream from line 20. Needless to say, such a classification has no application whatsoever to standard usage. Thorne’s ultimate goal is to write a syntax for this poem which both accounts for the sentences Cummings actually uses and, additionally, characterizes unattested sentences (such as Anyone kissed his children) which, Thorne claims, we feel instinctively to “ ‘belong’ to the same language as that in which the poem is written.”²⁷

We do not need to infer Thorne’s probable attitude towards our own problem with the syntactic analysis of this poem, since he discusses the question explicitly in his essay:

Another group of words which are most conveniently treated as nouns includes no one [sic], anyone, someones and everyones.²⁸

Thorne simply categorizes these words along with more standard nouns such as children (line 9) in his “Class C.” In so doing, he predicts that a sentence such as Women kissed their children will enjoy the same degree of acceptability within the context of this poem as the sentence Someones kissed their anyone . Since, in fact, his method assumes a priori that whatever appears in the text should be defined as acceptable by the specialized grammar for which it provides the only data, the appearance of anyone in lines 1, 6, 16, and 25 necessarily receives his approval despite the distinctly odd sound that I believe those sentences would have for the naive reader. There is in fact no mechanism in Thorne’s stylistic theory, no formal procedure, for recognizing ungrammaticality.

What previous approaches to the technical description of this poem have in common, then, is an incautious eagerness to discuss ways of extending (Levin) or totally recasting (Thorne) the grammar on the grounds of the deviant constructions it manifests. The authors cited here appear to have left out of consideration the possibility of analyzing the poem strictly according to the canons of the standard language. Significantly, both Levin and Thorne would, I suspect, miss what I take to be a crucial observation about the role of the word anyone in “anyone lived in a pretty how town”—its habit of appearing sometimes in standard, sometimes in distinctly nonstandard, syntactic environments.

To appreciate the importance of this aspect of Cummings’ technique, we must first restate one interpretively fundamental (and widely accepted) observation about the language of this poem: that the words anyone and noone have associated with them the personal pronouns he/him/his and she/her respectively.²⁹ The appearance of the form she in particular leads the reader to suppose a definite female individual (or character) referred to, an assumption which in turn suggests that the “indefinite” member of each pronominal pair may in fact be far from indefinite, constituting a “name,” Anyone and Noone respectively. A final logical step dictates that, as is always the case where such anaphoric relationships are found, we can paraphrase constructions in this poem that contain the relevant definite personal pronouns by substituting for each such occurrence the corresponding “name.” The clause in line 12 of this poem, for example, “noone loved him more by more,” translates by this method to the more fully defined Noone loved Anyone more by more. The only aspect of this procedure, as one must apply it to this particular text, that makes it at all remarkable is, of course, that the items used to make the crucial substitutions in this case are themselves pronominal and indeed would normally refer even more vaguely than the (singular personal) pronouns for which they substitute.

This complex situation enables the ironies of Cummings’ vision in this poem to emerge with great force. Narratively, after all, it is important that the deep affection of Noone for Anyone should blossom amid the aridity of the uncaring, imperceptive populace of the “pretty how town.” The poet himself, however, adds to this apparently simple Romeo-and-Juliet scenario a rather black comic twist, by arranging that both of the contrasted attitudes—the loving and the loveless, the personal and the impersonal—are equally well expressed in the key sentence “noone loved anyone.” Radically different interpretations of this simple clause result if we treat the subject and object NPs as proper nouns rather than as indefinite pronouns or vice versa. Again in line 26, Cummings mines the same rich vein of ambiguity, where, indeed, I find three rather than only two possible interpretations:

(a) “noone stooped to kiss anyone’s face”—with both indefinite pronouns interpreted generically, expressing the continuing emotional bankruptcy of the local population at large;

(b) “noone stooped to kiss Anyone’s face”—with Anyone used as a name, to capture the enduring hostility of this community towards the poem’s anti-hero; and

(c) “Noone stooped to kiss Anyone’s face”—Cummings’ reaffirmation of the central theme of his ballad, the love affair of the two named characters.

The importance of being able to substitute anyone for him and anyone’s for his in this way as a general interpretive strategy seems to me irrefutable. What may not have been recognized, I suspect, are the implications of this possibility for the technical analysis of the syntactic style of the poem.

If we substitute appropriately for third person singular anaphoric pronouns whenever they occur in this poem, we significantly increase the number of lines in which anyone may be said to “occur.” In the accompanying table, I list those lines in chronological order, using the letters “A” and “U” to indicate constructions in which the syntactic placement of anyone (and that aspect of the syntax alone) would be ruled acceptable or unacceptable respectively by the rules of the standard grammar.

This distribution of syntactically acceptable and unacceptable contexts for the word anyone in the poem corresponds strikingly with the narrative’s account of the fluctuating fortunes in love of the character Anyone.³⁰ As Anyone’s story opens, he is an outsider in the “pretty how town,” an outcast for whom “women and men / cared . . . not at all.” Cummings’ syntax early in the poem reflects this ostracism metaphorically, repeatedly stranding the word anyone in grammatically awkward surroundings, even when this requires, as we saw earlier, a certain amount of syntactic contortion (“. . . loved anyone not . . .”). The subsequent appearance upon the scene of noone/Noone (both the word, that is, and the character) provides anyone/Anyone for the first time with an appropriate linguistic/social context—a context that then remains intact throughout their courtship narrated in lines 11-24 the poem. In line 25, Anyone dies, his death severing the love-tie that had been his sole point of reference in society. At this point too, appropriately, the acceptability of anyone as a syntactic constituent again becomes problematical, as it appears in a nonstandard construction strongly echoic of the opening words of the poem (“anyone died”). In this case, however, a remedy for Anyone’s isolation is more quickly available in the shape of the death of Noone (line 26), and as the two protagonists are united in death, so too the syntax of the poem’s closing statements about them displays a return to harmony by conforming to standard acceptability judgments.

Any interpretation based on so abstract a syntactic dissection of this poem might at first seem far-fetched. We should note in passing, however, that Cummings himself points rather unambiguously to his own syntactic technique in the text of this poem. In my tabulation of the sentences in the text that contain the word anyone, I conspicuously failed to mark line 16 [“anyone’s any was all to her (noone)”] as either acceptable or unacceptable. This line poses considerable interpretive problems whatever the critical techniques one is employing. But it is at least somewhat comprehensible if read, so to speak, linguistically rather than narratively, as a syntactic commentary whose implications emerge only in light of Cummings’ use of associated syntactic devices elsewhere in the poem. Consider the following sentences, the second probably the most straightforward negation of the first:

Someone invited all of the guests here tonight.

Noone invited any of the guests here tonight.

Any in the second of these sentences constitutes the reflex, in a negative context created by the word noone, of all in the first sentence. Cummings’ statement in line 16 of his poem that (to paraphrase for a moment) the any- in anyone would be the equivalent of all to noone may thus be read as a linguistic clue to the poem’s technical syntactic structure, confirming the important role that syntactic restrictions on the placement of indefinite pronouns play in pressing home one of Cummings’ major themes.

The impact of my analysis of “anyone lived in a pretty how town” on the work of stylists such as Levin and Thorne should not be exaggerated. I endorse many of those writers’ conclusions; later in this chapter I shall adopt some parts of their respective methods. My own analysis weakens few if any of their assertions about the language of this poem. I certainly accept, for example, Thorne’s contention, mentioned earlier, that the sentence Anyone loved his children belongs to a stylistically important set of “possible-but-nonoccurring” sentences associated with this poem by the syntactic criteria he describes. I would add to his observation only the suggestion that, if it had been used, this sentence would have to have described an unpleasant, unsuccessful, or unfulfilling aspect of Anyone’s existence, since the word anyone itself appears here in a syntactically nonstandard context as the subject of a positive declarative sentence.

What is significant is that additional insights of this kind more often than not fail to be made at all, since they stem crucially from our having assessed the syntax of this poem by the standards applicable to everyday language, ruling each individual sentence acceptable or unacceptable solely by those standards. Whether he amends the standard grammar with Levin or recasts it with Thorne, the stylist who turns away too fast from the methods that linguists have already devised to account for day-to-day usage will forego the opportunity to crack these particular stylistic codes. Both of the methods that we have examined for adjusting the grammar to explain deviant constructions in poetry accept as “grammatical” a priori every one of the poem’s attested sentences. Nor will subsequent comparison of the grammar for a particular deviant text with that for standard English remedy the problem. Comparison will demonstrate only the general area in which the poet’s rejection of standard practice has introduced a peculiar syntactic flavor to the poem. It will never pinpoint the fact that this rejection may have been undertaken selectively and as a means for reinforcing some such nonlinguistic theme as narrative structure.

The moral to be drawn from this particular example, therefore, is not that there is no place whatsoever for nonstandard rules in stylistic analysis. Rather, I am arguing that a construction which strikes the analyst as nonstandard should first of all be assessed in precisely that light, with all the possible relevancies of nonstandard syntactic usage to the poet’s purpose being fully explored. Only after this crucial step has been completed should the stylist resort to adaptation or innovation, because both those radical procedures will distort his view of the structure of the text; they will effectively remove all possibility of discriminating between the acceptable and unacceptable constructions since they treat the poem as the only data to be explained, as itself the sole source of propriety judgments.

I have concentrated in the preceding discussion on Levin’s and Thorne’s analyses of “anyone lived in a pretty how town” as representing two particular schools of technical practice, but the literature abounds in parallel examples. Ann Banfield’s work on John Milton’s syntactic style seems to me to involve precisely the same procedural limitation.³¹ George Dillon’s insistence that the lines of poetry that he examines in his early essay “Inversions and Deletions in English Poetry” “do not differ greatly from the output of various optional rules of Modern English,”³² cannot offset his admission that his general purpose is to describe “relaxations of constraints on transformations” that, taken together, “characterize a kind of extra syntactic competence required of the reader.”³³ Nor does Dillon significantly modify this position in his later monograph, Language Processing and the Reading of Literature, where, by shifting the focus of his analysis towards heuristic, highly pragmatic “perceptual strategies,” he once more leaves the formal grammar essentially unconstrained, at least as it applies in literary contexts.³⁴ Even Richard Cureton’s admirable approach to the complexities of technical analysis in Cummings’ poetry allows the stylist the option of “mak[ing] various minimal changes in the order and substance of the words . . . in order to assign a structural description” to them before he has assessed the possibility that “the poet intends to communicate nonsense” or some other notion appropriately expressed in nonstandard syntactic form.³⁵ In view of this widespread acceptance by stylists of what I take to be mistaken priorities, and their general tolerance for poorly constrained, almost permissive theories in the technical subfield of poetic syntax, perhaps one further textual case study at this point would not be inappropriate.

For this purpose, I turn to “Adonais,” Percy Bysshe Shelley’s brilliant elegy for John Keats. Three consecutive stanzas of this poem contain lines that, by the standard grammar of English, one would almost certainly want to rule deviant.³⁶

Stanza V

And happier they their happiness who knew,

(CWS II, 390:39)

Stanza VI

Thy extreme hope, the loveliest and the last,

The bloom, whose petals nipt before they blew

Died on the promise of the fruit, is waste;

(CWS II, 390: 51-53)

Stanza IV

He died . . .

Blind, old, and lonely, when his country’s pride,

The priest, the slave, and the liberticide,

Trampled and mocked with many a loathed rite

Of lust and blood;

(CWS II, 390: 29, 31-34)

We must first take a few moments to pinpoint the source of syntactic infelicity in each of these passages individually. This technical exercise will itself demonstrate among other things the relevance to stylistic analysis of certain concepts at the very forefront of theoretical research in linguistics. Thereafter we may consider the implications of this particular text for the principal topic of discussion in this chapter—the proper treatment of deviant syntax in poetry.

A theoretical syntactician would probably explain the awkwardness that readers sense in line 39 of “Adonais” roughly as follows. Various transformations in English such as Passivization may shift syntactic material out of embedded (complement or relative) clauses into higher (or matrix) sentences. Thus from the underlying structure (a) in Figure 3, in which Piglet is the subject of S2, we can derive by Passivization structure (b), in Figure 4, in which that same NP is the subject of S1.

FIG. 3

FIG. 4

Such movement cannot take place altogether unrestrictedly, however. Any attempt to duplicate this situation when the embedded clause contains a finite verb instead of an infinitive invariably results in an unacceptable derived structure:

Pooh believed that Piglet had seen Christopher Robin.

* Piglet was believed by Pooh [S that ____ had seen Christopher
Robin].

To account for this restriction, which affects many movement rules besides Passivization, Chomsky has proposed a syntactic constraint, the Tensed S Condition, which he formalizes as follows:

No rule can involve X, Y in the structure

. . . X . . . [α . . . Y . . .]. . .

where a is a tensed sentence.³⁷

It is well known that not all syntactic rules that move material leftward are subject to this condition in its simplest form.³⁸ PP-Fronting, for example, can extract syntactic material from finite clauses with considerable freedom:

Up that path, I could imagine General Wolfe to have climbed at dead of night, (from “I could imagine . . . to have climbed up that path . . .”)

Up that path, I could imagine General Wolfe had climbed at dead of night, (from “I could imagine . . . had climbed up that path . . .”)

Even for these less heavily restricted transformations, however, cases still exist in which movement is prohibited, as, for example, when the embedded clause is introduced by an overt complementizer (roughly what used to be called a relative pronoun or a subordinating conjunction):

*Up that path I met a man who had climbed, (from “I met a man who had climbed up that path”)

Linguists attribute this secondary restriction to another syntactic condition, the Doubly-Filled Complementizer Constraint.³⁹ Together with the Tensed S Condition and various other constraints, this formal device severely limits the range within which transformations such as PP-Fronting are free to operate.

Let us return now to the relevant line of Shelley’s poem:

And happier they their happiness who knew.

It requires no linguistic skills to appreciate that a sentence structurally analogous to this one, such as Happy is the man his wife who meets (in town), would appear hopelessly ungrammatical in everyday conversation. The linguist need only study the relationship between Shelley’s line and its probable immediately antecedent structure, as shown in Figure 5, to detect the formal reason for this intuitive assessment.

FIG. 5

Whatever the movement rule responsible here for shifting their happiness from its underlying position as the post-verbal direct object of knew to its location in surface structure immediately before who, that rule has clearly applied in violation of the Tensed S Condition, of the Doubly-Filled Complementizer Constraint, or of both at once. As we shall see after I have discussed all three of the passages from “Adonais” relevant to my argument, this violation is by no means insignificant.

George Miller and Noam Chomsky, in an early paper that bears major implications for this whole discussion,⁴⁰ detected important limitations on the “depth” of certain types of syntactic structure. Where syntactic complexity is concentrated at the extreme left-or right-hand end of a construction, they noted, our ability to process it syntactically and semantically is limited only by comparatively trivial, nonlinguistic considerations such as the availability of sufficient “space” in our memory:

Left-Branching

[[[[[[Bill’s] aunt’s] chauffeur’s] mistress’] domineering father] mistook Bill for the chauffeur and broke his jaw.]

Right-Branching

[Grumpy mumbled [that Dozy had complained [that it was Snow White’s opinion [that the Wicked Queen knew [where they were all living.]]]]]

Where complexity occurs at the center of a given structure, by contrast, our abilities are far more limited. Thus while we accept as complex (but just interpretable) the first sentence below, the second, with only one greater layer of complexity, exceeds our parsing skills altogether (unless we help ourselves by using pen and paper):

Center-Embedded

? [The woman [whom the spider [whose web broke] startled] screamed.]

*[I’ll introduce the woman [who [when the spider [whose web broke] startled her] screamed] to my aunt tomorrow.]

In simple terms, this phenomenon is due to the fact that the person who attempts to interpret the first sentence faces the task of processing and memorizing three separate, incomplete syntactic units (the woman . . ., whom the spider . . ., and whose web . . .). He must then set about retrieving them in reverse order so as to match them correctly with their respective predicates, broke, startled, and screamed. If a retrieval task of this sort is extended beyond three to four or more matched pairs, as in the second sentence, psychologists predict that it will generally defeat us, regardless of the nature of the items to be memorized and recovered—be they numbers, letters, colors, or, as here, syntactic constituents. Our discomfort with such center-embedded syntactic structures thus represents merely a special case of a rather general limitation on our innate cognitive powers.

In light of the preceding discussion, we may now return to some lines quoted earlier from Stanza VI of “Adonais.” I repeat the passage below, inserting the relevant syntactic bracketing:

[Thy extreme hope, the loveliest and the last,

The bloom, [whose petals [nipt [before they blew]]

Died on the promise of the fruit,] is waste.]

It should now be clear where we must lay the blame for the intuitively obvious complexity of these lines. Shelley’s syntax here displays a classically center-embedded form and inevitably places an inordinate and quite perceptible strain on the reader’s language-processing capabilities by violating this crucial constraint.

The Center-Embedding Constraint, the Tensed S Condition, and the Doubly-Filled Complementizer Constraint have all been extensively analyzed by theoretical linguists. Less thoroughly explored is a constraint that explains, I believe, our uneasiness in reading Stanza IV of “Adonais.” At issue here is language’s inbuilt resistance to allowing nonstandard stylistic processes to add to the already heavy burden of potential ambiguity that inheres in syntactic surface structures. The basic insight was, once again, Chomsky’s:

[S]tylistic inversion . . . is tolerated up to ambiguity—that is, up to the point where a structure is produced that might have been generated independently by [other] grammatical rules.⁴¹

Banfield picked up on Chomsky’s lead in her dissertation, suggesting more concretely that, beyond the critical “point” that he had described, a syntactic condition (which I shall call the Up-to-Ambiguity Condition) actually prevented the generation of surface structures whose form might be attributed with equal plausibility to either of two syntactic processes.⁴²

English syntax permits leftward movement of simple NPs under various circumstances. Perhaps the simplest such case is Topicalization, which preposes a specified NP to sentence-initial position (see Figure 6). The surface word order that results from this operation, it may be noted, is object (O), followed by subject (S), then verb (V). Such sentences are reasonably common in everyday as well as in poetic language. But any account of syntactic practices in English poetry will have to account in addition for sentences whose surface structures display an SOV configuration, as in the following lines:

FIG. 6

And when we meet at any time again,

Be it not seen in either of our brows

That (NP we) (NP one jot of former love) retain.

(Michael Drayton, “Since there’s no help . . .,” lines 6-8; my
emphases)

In his paper on poetic inversion, Dillon makes a strong case for assuming that this configuration results from application of a stylistic rule of Verb Final, but he also astutely remarks that that rule is considerably complicated by the fact that “the position of the verb is normally crucial for determining grammatical relations”⁴³ Since the Verb Final transformation evades both of the standard SVO and OSV word orders by substituting an SOV surface structure, he notes in particular, the grammar places various additional ad hoc restrictions on the operation of that rule to ensure that overall complexity is kept to a minimum. (No additional dislocatory rules, for instance, seem able to apply once Verb Final has been selected.)

Of immediate interest to us is the closely related observation that the respective outputs of Topicalization and of Verb Final will often appear sufficiently similar to fall within the scope of the Up-to-Ambiguity Condition. We may take lines 32-34 of “Adonais” as a perfect case in point. Trampled and mocked being transitive verbs, the reader must try to determine which of the NPs preceding them is the subject and which is the object:

. . . when [NP1 his country’s pride,]

[NP2 The priest, the slave, and the liberticide,]

Trampled and mocked . . .

Has NP1 here been moved from object position by Topicalization? Or were the conjoined verbs moved to clause-final position by Verb Final, leaving NP1 and NP2 in their original deep-structure (S-O) order? The task of solving this problem—which is made no easier in practice, certainly, by Shelley’s prolific use of commas which mislead the reader into wondering whether NP2 might not even be appositive to NP1—cannot be accomplished if one relies on syntactic evidence alone. Only by semantic criteria, and even then only if he already knows of Shelley’s enduring political opposition to the more conservative and authoritarian elements of the British church and state, can the reader disentangle the structure of this passage correctly. (In actual fact, NP2 is the subject of the sentence, NP1 having indeed been fronted by Topicalization.) The problems that we experience in successfully determining the structure of this stanza of the poem are thus precisely those that the Up-to-Ambiguity Condition is supposed to preclude.⁴⁴

At first glance, the technical sketch that I have attempted in the preceding paragraphs may appear rather dry. Three consecutive stanzas of Shelley’s “Adonais,” I have argued, contain sentences that violate simple but supposedly absolute constraints on English syntactic processes; as a result, they strike the reader as unusual or awkward. Stated in this way, the analysis seems positively routine. Yet its implications for stylistic methods in general are by no means trivial, and involve both the issue of which linguistic model a stylist should work with and the question of what his priorities should be in conducting his technical analysis. We may summarize the most important points as follows.

(a) It is essential that the technical analysis of poetic syntax be conducted using the most modern linguistic methods available. Earlier in this chapter, I discussed this claim from a theoretical standpoint; I now reaffirm it for entirely practical reasons. The all-important link just uncovered among the three passages quoted from “Adonais” could never have been isolated had I not been employing a technical framework that recognized the existence of both syntactic transformations, and, more particularly still, conditions governing their operation. Such concepts have been developed within linguistic theory itself only quite recently. They thus demonstrate well the need for stylists to stay abreast of even the most current advances in linguistic science.

(b) Even purely technical analyses, stylistic accounts that contain absolutely no interpretation of the passages they describe, may nevertheless interest and assist the literary scholar. In the present instance, for example, my conclusions could be used to add focus and definition [not to mention objective support—see point (c) below] to previous commentators’ essentially intuitive and subjective remarks about Shelley’s use of the English language.

Shelley’s style in general has often come under critical fire for its “syntactic disorders.”⁴⁵ Remarks on this aspect of his work have varied in their specificity. Deborah Rosenfelt, in fact, came quite close to identifying syntactically at least those problems that we would now attribute to Shelley’s reliance on center-embedded constructions; she complained that his poetry too often exhibited “lengthy interruptions in the main line of the sentence.”⁴⁶ But the usefulness even of comments about Shelley’s poetic style as insightful and precisely stated as this one has been seriously compromised by the fact that their technical basis has generally been insufficiently sophisticated. The “length” of the “interruptions,” we now realize in particular, is far less important than their structural location and their overall “depth.”

Armed with this new information, the stylist may move forward without fearing repeated challenges to his basic technical premises. Such additional research might for example take the form of comparing Shelley’s syntactic complexity with that of Milton or Cummings.

(c) Not all technical features isolated by stylistic analysis are mere reflections of the analyst’s preconceived notions of what should (interpretively speaking) be the major emphases within the text. This is a charge repeatedly leveled by Barbara Herrnstein Smith and by Stanley Fish, who accuse stylists of maintaining the naive belief that there exist some “formal features that one can pick out independently of any interpretation of them.”⁴⁷ Any such pretense of separating technical from interpretive judgments, Fish maintains, is either downright dishonest or self-deluding.

The present case, I suggest, disproves his claim. Whether or not a given syntactic structure within “Adonais” is center-embedded or violates the Doubly-Filled Complementizer Constraint is a factual issue, altogether unaffected by the interpretations that a stylist may subsequently decide to place upon it. No competing derivations, no shadings or revisions of the grammar are involved here; the issue is, for once, black and white.

Indeed, I might well go further and argue that even a statement about readers’ responses to these constraint violations could well remain objective and noninterpretive. Linguists have claimed increasingly often in the past decade that the existence of syntactic conditions at all and even certain aspects of their formal shape might be determined by cognitive strategies relied on by hearers and readers for interpreting sentences.⁴⁸ To oversimplify the argument for a moment, it is suggested that all syntactic constraints may exist to ensure that surface structures generated by the grammar do not positively mislead the language-analyzer in our heads as it goes about its task of decoding sentences.

Within such a framework, we may note, a poet’s decision to violate a given condition necessarily puts the reader’s mind under a certain kind of cognitive pressure.⁴⁹ To assess the relevance of this fact to our understanding of the text as a whole represents by all means an interpretive act; but simply recognizing it as a fact involves us only in an empirical claim within the technical field, a claim certainly subject to subsequent confirmation or disproof.

All three of these conclusions constitute useful substantiation for claims I advanced earlier in this study. Equally significant, though, is the fact that my technical analysis of these lines from “Adonais” provides important confirmation for a fourth point I made in the immediately preceding portion of this chapter about preferred strategies for describing deviant syntactic constructions in poetic texts:

(d) Stylistic insights may be lost wherever technical analysis proceeds with the principal aim of showing how all lines of a given text may be encompassed (ruled acceptable) by a custom-designed poetic syntax. On the one hand, it is unclear, given all that has been said about syntactic conditions above, how a stylist’s specialized grammar for a text such as “Adonais” could arbitrarily rule fully acceptable a series of structures which, linguists would maintain, our cognitive faculties simply fail adequately to process. Such an exercise in futility would reduce the stylist’s accountability to linguistic theory to zero and enclose the technical analysis of poetic syntax within a hermetically sealed bubble.

On the other hand, a “grammar-fixing” approach to the opening section of “Adonais” would also, presumably, have to proceed by assuming Shelley’s style in that poem to be uniformly unheedful of syntactic conditions and constraints. The conditions in question would, in effect, be struck altogether from the grammar formulated for this poem. This approach unfortunately overlooks a crucial, uneven distribution of syntactic deviations in this text. It turns out that only Stanzas IV, V, and VI of the poem contain such marked syntactic aberrations (a fact which is highly significant, as I shall show in detail when I return in chapter 4 to a discussion of how we set about interpreting this elegy). But a stylistic analysis that reacts to each violation merely by recasting the grammar for this poem by deleting the relevant constraint altogether loses thereby its sensitivity to the presence or absence of similar violations elsewhere in the text.

As in the discussion of Cummings’ “anyone lived in a pretty how town,” so too here we discover the danger caused by premature capitulation to the temptation to rule-monger. Discriminating the acceptable from the unacceptable within a single text becomes impossible unless deviations are treated as precisely that. Only reluctantly, then, if at all, should the analyst move away from the standard grammar in the direction of a more powerful, less restricted syntax for his text.

IV

The act of stepping outside the standard grammar to complete a technical stylistic analysis is itself fraught with various theoretical and practical dangers. From the very beginning of modern stylistic research, the vast power of transformational grammars again loomed as the major obstacle to uncontroversial analyses; a deviant syntactic construction, far from lacking any adequate formal account, usually admitted of several. Naturally, therefore, if a little sluggishly, stylists began the search for suitable criteria by which to judge the relative merits of competing technical analyses.

Perhaps the most popular such criterion—he calls it a “methodological moral”—is that advocated by Dillon:

[O]ne should assume the minimum of difference between poetic syntax and that of ordinary language.⁵⁰

As we have already seen, Levin would almost certainly endorse this recommendation and in this particular respect would find a strong ally in Thorne:

[T]here will be a tendency to make the grammar of the literary dialect as nearly isomorphic with the grammar of the language . . . as possible.⁵¹

For a long time, the terms “different” and “isomorphic,” clearly vital to an adequate assessment of this principle, went completely undefined, as did the methodological steps by which one should proceed if one wished to discover in practice the maximally “similar” syntactic account desired. John Lipski justly complained in 1977 that seldom in the literature did one “find explicit criteria for establishing the transitions necessary” to define a stylistic grammar on this principle.⁵² Within this context, Cureton’s 1980 dissertation represented a major breakthrough, making a number of highly significant contributions toward remedying this deficiency in the technical subfield of stylistic theory. For those interested in a clearer understanding of the principle of isomorphism, I strongly recommend chapter 3 of that valuable work.

Vital as such studies may be, I wish to plot a somewhat different course. It seems to me by no means a foregone conclusion that isomorphism with the standard grammar, however that may ultimately be defined, should in fact constitute our primary goal in describing poetic syntax. As a syntactician, I find it most disturbing that such a principle, based as it is on the external orientation of the grammar proposed, should have been adopted as a methodological a priori. In recent linguistic theory, by contrast, the assumption has generally been that analysts’ efforts will be directed initially towards establishing a range of individual grammars on the basis of purely internal criteria of adequacy. Only after a number of such grammars have achieved a substantial measure of descriptive power in their own right does the linguist induce any generalizations about properties that those grammars may have in common. Thus, for instance, virtually every attempt to establish so-called linguistic universals has grown out of, rather than preceded, detailed investigation of a number of specific natural languages. Only in this way, it is felt, can the theory being developed avoid the trap of becoming self-defeatingly prescriptive.

My thesis in the discussion that follows, then, will be that, in designing a syntactic model for stylistically deviant language, the need to reflect major syntactic generalizations about the language of a given text should claim the stylist’s attention above and beyond all other methodological considerations. In adopting this position, I shall, in a sense, be accepting the basic philosophy of Thorne’s paper as cited earlier in this chapter, advocating with him that each text be treated as a separate block of data demanding a maximally simple and elegant account of its own idiosyncratic regularities. I shall differ from him only in rejecting his essentially unsupported assumption that the most descriptively adequate account of a text within this framework may be determined equally effectively simply by demanding isomorphism with the standard grammar. To take another point of reference, I shall bypass Dillon’s first “methodological moral,” which I quoted earlier, in favor of a second, which he advances almost as an afterthought: “that one must consider the grammar of the poet as a whole.” ⁵³

If one were interested in elaborating what Dillon at one point calls “a taxonomy of difficulty” in the works of various English poets,⁵⁴ one might contrast the syntactic complexity of Shelley’s style with that found in the poetry of Alexander Pope, which may be attributed not to any violation of syntactic conditions or constraints, but to the many coordinate sentence structures in his poems from which constituents have been deleted by processes unacceptable in the standard language.⁵⁵ The following examples are typical:⁵⁶

While Fish in Streams, or Birds delight in Air,

(“The Rape of the Lock”: PAP, 231: 163)

Who writes a Libel, or who copies out:

(“Epistle to Dr. Arbuthnot”: PAP, 607: 290)

Favours to none, to all she Smiles extends,

(“The Rape of the Lock”: PAP, 223: 11)

The material deleted from each of these lines can, of course, be inferred by the reader upon reflection. He can, that is, reconstruct the presumably underlying clausal structures:

While Fish delight in Streams, (. . .)

(. . .) or who copies out a Libel,

and

She extends Favours to none, (. . .).

Still, precisely analogous surface structures in everyday contexts would, as I shall shortly demonstrate in detail, be judged unacceptable. Thus the stylist is faced with the task of accounting technically for the success of a series of structures not explicable within the framework of the standard grammar.

Furthermore, constructions of the kind cited in the preceding paragraph pervade Pope’s poetry. As a result of such widespread occurrence, they cannot be expected to correlate with specific themes that might then justify positing strategically confused or ungrammatical syntactic expression of the kind we discussed in connection with Cummings and Shelley above. Tentatively, therefore, we may begin instead to explore ways of extending the syntax of standard English so that it will actually generate such examples, watching at the same time for means by which we might ultimately evaluate each such formal extension as a more or less desirable contribution toward a custom-designed grammar of Pope’s poetry.

Let us begin this exploration by analyzing line 163 of “The Rape of the Lock,” considering in turn two possible technical accounts of the surface form it manifests. Let us assume, first, that deviant structures in poetic texts always bear maximally “transparent” relations to their deep-structure sources.⁵⁷ This hypothesis will lead us to assign line 163 some such underlying structure as that in Figure 7.

Deriving line 163 as it actually appears in the text of the poem from such a deep structure will require that we delete Va (delight) from the left conjunct, presumably on the basis of its identity with the corresponding verb in the structurally parallel right conjunct. Generally speaking, such an approach would not represent a particularly extreme proposal. Coordinate structure deletions of this kind are common in English as in many other languages, and the specific process involved here, usually termed Gapping, has generated extensive theoretical discussion among linguists as a bona fide transformation in the standard grammar.⁵⁸ One aspect of Pope’s application of the Gapping rule in this particular context does, however, demand further comment. In standard English, Gapping deletes material from the right and not the left of two parallel conjuncts, as can be clearly seen from the following sentences:

FIG. 7

Pope lived near London and Swift lived in Ireland.

Pope lived near London and Swift, ___ in Ireland.

* Pope ___ near London and Swift lived in Ireland.

The sentence that Pope employs in his poem, of course, conforms precisely to the usually unacceptable pattern illustrated in the third of these sentences. Were we to adopt the general approach to the technical analysis of this passage that I have just outlined, therefore, we would have to take the important step of extending the grammar of standard English so as to permit leftward or “backward” Gapping.

At this point I want briefly to set alongside this technical analysis of line 163 a second, initially plausible, approach, but one which we shall eventually find evidence to reject. This second approach would abandon the deep structure for this sentence given above in favor of a more abstract one in which S2 preceded rather than followed S1 (as in Figure 8). Gapping can of course apply to this structure in a perfectly straightforward, left-to-right manner, correctly deleting delight from S1. Now, however, we shall have to posit a new rule of Conjunct Scrambling to account for the fact that, in surface structure, S1 appears to the left of S2. While simple enough to devise, such a rule is nonstandard, and its addition represents the “cost” of this second approach to the overall economy of the grammar being proposed.

FIG. 8

Our discussion thus far has led us to formulate two hypotheses concerning the derivation of our problematical line from “The Rape of the Lock.”⁵⁹ Each involves making a single clearly delimited adjustment to some aspect of standard English syntax. Under the first hypothesis, it will be necessary to apply the Gapping transformation backward; the formal statement of that rule will therefore have to be altered accordingly. In the second, we shall need instead to introduce a completely new rule, Conjunct Scrambling, to ensure the correct surface order of the coordinate clauses involved. It is extremely important to remember also that any decision we make to prefer one of these hypotheses over the other will do far more than just provide a preferred technical account of one line. It will affect the treatment of numerous parallel cases in Pope’s poetry, of which the following are merely a representative sample:

Now Leaves the Trees, and Flow’rs adorn the Ground;

(“Pastorals: Spring”: PAP, 125: 43)

Some few in that, but Numbers err in this,

(“Essay on Criticism”: PAP, 144: 5)

A Belt her Waste, a Fillet binds her Hair,

(“Windsor-Forest”: PAP, 201: 178)

We should therefore proceed with some caution when determining what would constitute legitimate arguments in favor of one or the other of our two hypotheses.

Early evidence on the side of what we could call the “Conjunct Scrambling Hypothesis” might include the observation that not all of the examples of irregular coordinate structure deletion in Pope’s work involve just two conjoined clauses; in several instances, deletion results in far more obviously “scrambled” patterns. Let us briefly return, for example, to line 163 of “The Rape of the Lock.” That line, we now see, constitutes only the first part of a construction that as a whole spans two lines, a construction in which delight has been deleted from the third, as well as from the first, of three conjuncts:

While Fish _____ in Streams, or Birds delight in Air,

Or in a Coach and Six _____ the British Fair,

(PAP, 231: 163-164)

It is thus in actual fact the verb of a medial clause that apparently controls two deletions (one leftward and the other rightward) in this passage. Nor does this pattern represent the only possibility even where just three clauses are involved. In my next example, deletion sites occur in clauses one and two, the verb supply (and incidentally the direct object him, also missing in the earlier clauses) appearing only in the third conjunct:

Happy the man, . . .

Whose herds with milk, whose fields with bread,

Whose flocks supply him with attire,

(“Ode on Solitude”: PAP, 265: 1, 5-6)

And a similar case from Pope’s “Epistle to Cobham” exhibits a still more complex distribution of deletions, with the only overt mention of the crucial verb turn appearing in the second of four conjoined clauses:

Manners with Fortunes, Humours turn with Climes,

Tenets with Books, and Principles with Times.

(PAP, 555: 166-167)

It might well be argued that in such cases uniform rightward application of Gapping, to be followed by fairly random scrambling of the various clauses involved, represents the elegant derivational path. By comparison, simultaneous forward and backward Gapping of the kind that would be required within an extended grammar that had no scrambling rule looks distinctly clumsy.

Against such formal syntactic arguments in favor of the conjunct scrambling solution one must weigh reasons of an altogether different kind for preferring the “Backward Gapping Hypothesis.” Suppose, for example, that we adopt temporarily the assumption that the preferred stylistic grammar should always be the one that generates the smallest possible set of nonstandard sentences beyond those already contained in the text under examination. (While almost impossible to realize in terms of “hard” statistics, such a doctrine has at least the distinct merit of providing an explicit, theoretically quantifiable criterion for measuring stylistic isomorphism.) In the present case, Conjunct Scrambling will potentially apply to a very wide range of coordinate structures. Indeed, unless carefully constrained, its addition to the grammar will make extremely difficult the task of determining unambiguously the precise deep-structure order of any pair of conjoined clauses. Permitting backward Gapping, by contrast, will affect only a subset of conjoined clauses, those that display strict structural congruity and a certain degree of local referential identity.⁶⁰ By this standard, therefore, the Backward Gapping Hypothesis emerges as the preferred, because the more restrictive, account of Pope’s poetic syntax.

Neither of these arguments, though, strikes me as irrefutable. Calculating the potential of some rule for generating nonstandard sentences represents only one possible scale on which to measure isomorphism with the standard grammar; the total number of rules added or dropped or the overall “naturalness” of the new grammar might be equally plausibly proposed. In any case, the theoretical cost of introducing a rule such as Conjunct Scrambling could only be finally assessed in light of its usefulness both here and elsewhere in accounting for stylistic phenomena in Pope’s works. In the end, such a debate is not so much unresolvable in my view as irrelevant, the whole picture being altered fundamentally by the observation that there exists a second class of nonstandard coordinate surface structures in Pope’s works.

A representative member of this second class, contained in a line from “The Epistle to Dr. Arbuthnot,” was cited earlier in this discussion:

Who writes a Libel, or who copies out.

Copy out, a transitive verb, appears in this line without a specified direct object. As the context makes perfectly clear, that object must in deep structure have been identical to the object NP of the syntactically parallel preceding clause, a Libel, yet deletion from such a configuration is never admissible in the standard language:

*The man who loves his wife and who respects _____ should be universally admired.

A pronominal object (her) must appear in the second clause if the whole construction is to be judged acceptable. Not that English never permits the deletion under identity of rightmost constituents in conjoined clauses. Indeed, where the item to be deleted is the rightmost constituent of a lefthand conjunct, as in sentence (b) below, deletion is highly probable. Only where the deletion works from left to right are problems encountered, as illustrated in sentence (c).

(a) Nixon made the tapes, and Miss Woods erased the tapes.

(b) Nixon made ___ ______, and Miss Woods erased, the tapes.

(c)^* Nixon made the tapes, and Miss Woods erased ___ ______ .

The transformation that derives the acceptable (b) sentence here, usually referred to as Right Node Raising, has never to my knowledge been satisfactorily described in formal terms. Its various properties, however, have been closely examined and its characteristic proscription of rightward deletion noted.⁶¹ What, then, are we to make of Pope’s line in his “Epistle,” where it is the supposedly illegal option that the poet has evidently invoked?

One approach would be simply to adjust the standard syntax by relaxing the prohibition on rightward Right Node Raising. This would permit an intuitively direct derivation to connect the attested sentence with the plausible deep-structure source shown in the tree-structure in Figure 9. It would also require, however, that we effectively double the potential output of the Right Node Raising transformation by allowing it to apply freely in either direction.

An altogether different approach would again involve use of the Conjunct Scrambling rule and would indeed closely parallel that proposed for Gapping cases in the preceding section of this argument. S2, that is, would be generated before S1 in deep structure; the direct object of copies out would be deleted by standard (leftward) Right Node Raising; and Conjunct Scrambling would finally invert S1 and S2 to yield the attested line. At first glance, certainly, this second approach appears quite elegant, especially in view of the possibility discussed earlier that Conjunct Scrambling might be included in our syntax for Pope’s poetry on independent grounds—the need to handle cases of apparent backward Gapping.

FIG. 9

Any premature hopes, however, that the Conjunct Scrambling Hypothesis might offer an integrated solution to all of the coordinate structure deletion irregularities encountered so far are promptly dashed when one comes to consider the following line from Pope’s “Essay on Man”:

Since but to wish more Virtue, is to gain.

(PAP, 545: 326)

We find here a construction involving two sentences that behave, at least from the point of view of syntactic deletion, very much like a pair of conjoined clauses. The matrix sentence in this case is, however, copular, and semantically the sentence as a whole connects the two clauses in a virtual cause-and-effect relationship. To posit in this case an underlying sentence

To gain more virtue is but to wish more virtue,

as the Conjunct Scrambling Hypothesis would dictate, appears radically counterintuitive. Worse yet, such a step would grant Conjunct Scrambling enormous latitude, permitting it to reverse fundamental semantic predications and compounding the problems noted earlier that are created by this rule for the unambiguous determination of underlying syntactic structure. Once again, then, we appear to be faced with two equally imperfect hypotheses to describe a specific set of nonstandard constructions.⁶²

Let us pause a moment to take stock of our position. We began this section of the chapter by studying constructions in Pope’s poetry in which the results of syntactic deletion most closely resembled those usually associated with the Gapping transformation. We noted, however, that the particular sentences Pope wrote were nonstandard in that deletion occurred in the left-rather than the right-hand conjunct. Two formal descriptions of this situation (the Conjunct Scrambling Hypothesis and the Backward Gapping Hypothesis) seemed equally plausible prima facie. We then turned to cases in which deletion appeared to have resulted from some form of Right Node Raising.⁶³ Again we encountered irregularities and were forced to consider various means of accounting for them. Conjunct Scrambling, on this occasion, proved unable to account for at least one important case, in which the copular verb be rather than a conjunction linked the relevant clauses in a nonreversible semantic predication. In the case of this second class of examples, I now want to stress, only one formal extension of the grammar could effectively handle all of the examples that I adduced. Only one, that is to say, which we might call the “Rightward Right Node Raising Hypothesis,” achieved even “observational adequacy” on Chomsky’s hierarchy of theoretical adequacy.⁶⁴ At one level, of course, this discovery allows us to assert a firm preference for the Rightward Right Node Raising Hypothesis where that hypothesis is in fact directly applicable. At another level, though, the same result may be brought to bear on the parallel and more delicately balanced problems of how to approach cases of apparent backward Gapping in Pope’s poetry.

Evidence in this area, it will be recalled, seemed inconclusive when discussion was confined solely to the lines with Gapping-style deletion patterns. Now, however, from a broader perspective, we can see that the situation is far less uncertain. Both Right Node Raising and Gapping may be classified as coordinate structure deletion transformations. Both share significant characteristics and restrictions on their operation. If, therefore, we accept both rightward Right Node Raising and backward Gapping as the appropriate ways in which to account for their respective classes of nonstandard constructions in Pope’s poetry, then we shall be able to infer an important higher-level generalization about Pope’s syntactic style: that Pope characteristically applies coordinate structure deletion transformations without regard to any restrictions imposed on the direction in which they may operate by the syntax of the standard language.

It is to this end that my whole discussion in the preceding pages has tended. I have sought a formal description of Pope’s syntactic deviance that could stand, as any linguistic hypothesis should stand, the requirement that it characterize “significant generalizations about the structure” of the data it purports to describe.⁶⁵ Such a requirement appears to me altogether crucial to the proper technical analysis of poetic style. In some cases, perhaps in many, the syntactic choices that it dictates will indeed conform to those that would result from applying Thorne’s doctrine of maximal isomorphism with the standard grammar. But where these two methodological morals conflict, I maintain that there can be no doubt which should be the stylist’s “Golden Rule.”

V

Despite the length and detail of the three discussions that have occupied the greater part of this chapter, I have only begun to explore a daunting mass of material in the subfield of technical analysis that stylists urgently need to consider. Listing all the material not discussed would of course be futile, but we may usefully consider for a moment just a few of the other broad subject areas that await adequate treatment:

(a) How should stylistics approach texts in which no deviant syntactic constructions as such occur but in which at the same time unusually heavy usage of a particular syntactic process (or group of allied processes) results in a unique stylistic flavor?⁶⁶

(b) What relationship, if any, obtains between the methods of technical stylistics and those used in other branches of so-called applied linguistics? Can all nonstandard dialects, whether literary, geographical, or social, be handled according to the same principles of analysis?

(c) More generally still, what can the theoretical syntactician himself learn from the ways in which poetic syntax differs from that of the standard language? Can evidence from poetic practice be used to support one hypothesis about the standard language over another?

All these, like the issues discussed (however inadequately) in the course of this chapter, are empirical questions. They are also, to be sure, complex and elusive. The theoretical conclusions to which they may lead us, however, are by no means trivial. The careful and accurate formulation of a technical analysis of each individual work offers the beleaguered discipline of stylistics its best shot at general credibility; and “accuracy” in matters such as these is inseparable from thoughtful and closely reasoned discussion of proper procedures by which to derive those analyses. Harold Whitehall’s assertion that “no criticism can go beyond its linguistics”⁶⁷ is clearly too strong, too territorially aggressive; the corollary that “no stylistics can survive poor linguistics” is by contrast one that every stylist would do well to take to heart.