“Language Processing and the Reading of Literature”
Reference, Coreference, and Attachment
IN the Introduction, I treated finding the referents of noun phrases as a part of comprehension rather than perception. My basic reason for doing so is that we have a propositional structure when we have identified a function in the structure for each phrase—we do not need to know what they refer to in terms of our knowledge as shaped by context. The ’scene' under discussion is part of the contextual frame, and linking the noun phrase to entities in the ’scene' integrates the material into it: hence we are in the domain of comprehension.1 It will happen in a stretch of text that the same individual is referred to by more than one noun phrase; hence those phrases will be coreferential, but comprehension may or may not proceed by identifying the coreference: one can find referents by noting that one noun phrase is coreferential to a preceding one (this I would call a structural strategy) and hence refers to the same individual, or by notingthat the noun phrase refers to an individual previously introduced into the scene and hence that the second noun phrase is coreferential to the one that earlier introduced the referent or last mentioned it. This second procedure I would call a referential strategy. It is quite different from the first in that no notation of coreference as such is necessarily involved. What distinguishes definite noun phrases and pronouns from indefinites is that they explicitly indicate that they refer to an individual previously introduced, so that one might suppose they would automatically trigger a process of the first type—call it a process of leftward search. However, pronouns and definite noun phrases are often used more loosely to refer to things which have not been introduced by a noun phrase—that is, there is no antecedent to be found. One may call such uses of pronouns ungrammatical if one chooses, but they are so common that we have some means of coping with them perceptually. Consider, for example, these lines from Paradise Lost:
(1) But he who reigns
Monarch in Heav'n, till then as one secure
Sat on his Throne, upheld by old repute,
Consent or custome, and his Regal State
Put forth at full, but still his strength conceal'd,
Which tempted our attempt, and wrought our fall.
[Milton, PL, I. 637-42]
Here the antecedent of which is, at the very least, “his concealing his strength,” and it may be all the preceding material. Which here is said to have a “sentential” antecedent—which is to say that we must take the entire action or 'thought' of the previous clause as the 'thing' referred to. Such uses of pronouns have sometimes been cited as evidence that sentences or clauses are noun phrases at some level of representation, but this is at best just one way of modeling the perceptual process involved. There are other uses that may force us to dig out or cobble together an 'antecedent':
(2)whoever hath an Ambition to be heard in a Crowd, must press, and squeeze, and thrust, and climb with indefatigable Pains, till he has exalted himself to a certain Degree ot Altitude above them. Now. in all Assemblies, tho' you wedge them ever so close. we may observe this peculiar Property, that, over their Heads there is Room enough....
[Swift, A Tale of A Tub, Quintaru Ed. 277]
Crowd and assemblies are said to function as anaphoric peninsulas here: one can find an antecedent in the paraphrase of their senses ("groups of people.. ."). Bever, Carroll, and Hurtig suggest that such sentences are easier to comprehend when the 'antecedent' is in a clause prior to the pronoun because the clause has been processed and its contents are accessible in the abstract form needed here.2 (Hence, by the way, the second them should be harder to process than the first, since the clause with the 'antecedent' [assemblies] has not yet been processed, though I think in this case that the previous them facilitates repeating the move.) Such a process of inventing an 'antecedent' is presumably more a referential than a structural process, and it is conceivable that referential processes might be more efficient in identifying referents, so that we would use a structural strategy of leftward search mainly for confirmation: “I think this is the referent—can I find an antecedent which introduces it?” If we cannot, we might pause briefly to consider whether the “ungrammatical” reading is in fact the best one. Consider, for example, the following passage from Spenser:
(3) They bene ymet, and both their points arriued,
But Guyon droue so furious and fell,
That seem'd both shield and plate it would haue riued;
Nathelesse it bore his foe not from his sell,
But made him stagger, as he were not well:
[Spenser, FQ, III.i. 6]
Citing only the second and third lines as context, Sugden identifies the referent of it as “the general idea of the preceding verb. . . i.e., the drive, the stroke,”3 but perhaps we could dig a singular point out of the plural since it is only Guyon’s point involved in his 'drive'. This example is reminiscent of (31) in Chapter One, where one can extract stroke as a 'cognate Object' from stroke as the referent of it, but helmet is a somewhat more likely referent. (Examples [11] and [12] of Chapter Five are similar.) The general rule of thumb should be the analogue of that sketched at the end of the previous chapter: take the grammatical reading (i.e., the one with an actual noun phrase antecedent) as the right one.
As with other perceptual rules-of-thumb, however, this rule cannot be taken categorically but must be adjusted to particular writers and texts. In Swift’s A Tale of A Tub, for example, (though not so much in his other writings) one must assume that a sentential antecedent is at least as likely as an actual noun phrase. Consider, for instance, the which's in the following passage:
(4) Now, it usually happens, that these active Spirits, getting Possession of the Brain, resemble those that haunt other waste and empty Dwellings, which for want of Business, either vanish, and carry away a Piece of the House, or else stay at home and fling it all out of the Windows. By which are mystically display'd the two principal Branches of Madness, and which some Philosophers not considering so well as I, have mistook to be different in their causes, over-hastily assigning the first to Deficiency, and the other to Redundance.
[Swift, TOT, Quintana Ed. 344]
The first which must refer to two things, but not the active spirits and the ghosts, or the simile collapses (it suits Swift’s satiric purpose, however, that this collapse should occur); rather, the two things are the two modes of operation of these ghosts (vanish ... and stay at home...). The second which, following and, seems coordinate to the first and hence coreferential to it, but the two modes of operation can scarcely be said to be caused by deficiency and redundance; rather, the second which must refer to the two principal branches of madness. On this reading, which I think is the best, the and is just a piece of craziness.
In this chapter, we will consider definite noun phrases and pronouns separately because there are differences which might lead one to prefer one type of strategy for definite noun phrases and the other for pronouns. Finally, we will take up the attachment of participials to the noun phrases they modify because the principles involved are very similar, reserving comprehension of appositives for the next chapter.
Although both definite noun phrases and (definite) pronouns convey the assumption that their referents have already been introduced into the scene, locating the referent with definite noun phrases is usually easier than with pronouns because the definite noun phrase contains more information about the referent. We can determine the referent of the father of our country, that is, without regard to verbal context, but not so for he, that, who, etc. 'Elegant variation' in the noun phrases one uses to refer to the same individual can present a problem slightly different from that with pronouns: pronouns do not introduce a new mode of reference (i.e., new properties of the referent). With elegant variation, we must identify an individual previously described one way in terms of a different (but not a more general) description. Seymour Chatman notes that James’s fondness for elegant variation creates considerable difficulty for the reader, who, for example, must in the first pages of The Ambassadors align the noun phrases The same secret principle, this principle, and this happier device as referring to the same thing.4 In a paper presented to the Linguistic Society of America (Winter, 1976), Robert Kantor pointed out that there is a contrast between noun phrases that are merely definite and those that are also demonstrative. The demonstrative article or pronoun (this/these, that/those, such), he observes, signals to the reader that some extra effort may be required to locate the referent, either because the entity referred to is not at the moment in the foreground of attention or because it is receiving a new description. This seems to be the function of this in the phrases just cited from The Ambassadors and explains several of the examples in Chapter One where a fronted Object is demonstrative (e.g., [36], [39-40]). To see what is meant by a “new description,” consider the contexts for examples (39) and (40):
(1.39)
All is not lost; the unconquerable Will,
And study of revenge, immortal hate,
And courage never to submit or yield:
And what is else not to be overcome?
That Glory never shall his wrath or might
Extort from me.
[Milton, PL, I. 106-11]
(1.40)
Then strait commands that at the warlike sound
Of Trumpets loud and Clarions be upreard
His mighty Standard; that proud honour claim'd
Azazel as his right, a Cherube tall:
[Milton, PL, I. 531-34]
In both of these, the demonstrative noun phrases bundle up the preceding propositions or action as single entities. Here we observe the link between a “new description” and a refocusing theme: the “new description” picks out certain aspects of what has been presented and hence constitutes a refocusing on what is already present in the scene. In other examples (e.g., I. 36, 34 below), the referent is hard to pick out because two are available or because the referent must be inferred (as in [34] below, where that birth must be inferred from the act of begetting, which is all that has been explicitly mentioned). Kantor noted that unskillful writers do not use these signals very effectively, but it appears that our writers do use them to indicate to the reader when special efforts to identify referents are necessary (see also [21, 23, 24] below).
As with pronouns, the referent of a definite noun phrase may not have been specifically introduced:
My car failed inspection. The muffler is shot.
Here we make use of 'knowledge of the world' in establishing the link: introducing a car is tantamount to introducing a muffler. Violations of the principle that the referent of a definite noun phrase has been introduced are a source of difficulty in Wallace Stevens. Consider, for example, the definite noun phrases in the following poem, including those of the title:
(5) The Sense of the Sleight-of-Hand Man
One’s grand flights, one’s Sunday baths,
One’s tootings at the weddings of the soul
Occur as they occur. So bluish clouds
Occurred above the empty house and the leaves
Of the rhododendron rattled their gold,
As if someone lived there. Such floods of white
Came bursting from the clouds. So the wind
Threw its contorted strength around the sky.
Could you have said the bluejay suddenly
Would swoop to earth? It is a wheel, the rays
Around the sun. The wheel survives the myths.
The fire eye in the clouds survives the gods.
To think of a dove with an eye of grenadine
And pines that are cornets, so it occurs,
And a little island full of geese and stars;
It may be that the ignorant man, alone,
Has any chance to mate his life with life
That is the sensual, pearly spouse, the life
That is fluent in even the wintriest bronze.
[Stevens, CP, 222]
Some of the definite noun phrases may be taken as given in a scene (the leaves of the rhododendron, the clouds, the wind, the sky, the bluejay, the fire eye in the clouds); others may appear as drawn from a stock of general types (the ignorant man) or conventional entities (the gods). (Later in his career Stevens makes ever more frequent definite references to types and figures in his own system.) The effect of positing a scene in this way (the empty house) rather than describing it, and of drawing on general types and figures, is to treat the scene as an instance of something else, not the locus of attention: this is one source of the often remarked abstractness in Stevens' poetry. The title itself is a definite noun phrase—obviously without antecedent—but one may assume by a general convention governing titles that it will be specified by the poem which follows. There remain two definite noun phrases, however, that are difficult to identify if one is unfamiliar with Stevens. The sleight-of-hand man might be taken as coreferential to one, but one could not be sure of this without having some idea of what each phrase refers to. It is easier to be relatively sure of this if one is familiar with Stevens’s (self-)deprecatory ways of referring to the poet and his poetizing impulse, a familiarity requiring some knowledge of his poetry beyond that available in the poem. Finally, the sensual pearly spouse is unexpectedly definite. The problem is that mate and weddings of the soul appear to be tossed in and are read metaphorically so that they do not introduce spouse into the scene, though as surely as cars have mufflers, weddings and matings have spouses (though not necessarily pearly, sensual ones). The whole noun phrase gives the impression that the reader is overhearing the poet in meditation rather than being addressed by him.5
The effect produced by Stevens with 'unintroduced definites' is in some ways similar to the effect obtained by certain of James’s narrators: one enters their stories in mid-telling, as if one is in a conversation which has been going on for some time. Here is the first sentence of “The Real Thing”:
(6) When the porter’s wife (she used to answer the house-bell), announced “A gentleman—with a lady, sir,” I had, as I often had in those days, for the wish was father to the thought, an immediate vision of sitters.
[James, Tales, VIII. 229]
The definite the thought precedes its 'antecedent', which is indefinite (an immediate vision), and we must infer from certain oblique references in the pages that follow what the wish refers to. One does not have to wait long, to be sure, to get some further information, but the definite article has just the opposite of its normal function here: instead of signaling that the referent has already been introduced, it signals that the reader must seek more information as he reads on. We will discuss this process of 'rightward search' in regard to pronouns below. The next example is even more unconventional:
(7) These things were true, but it was not less true (I may confess it now—whether because the aspiration was to lead to everything or to nothing I leave the reader to guess), that I couldn't get the honours, to say nothing of the emoluments, of a great painter of portraits out of my head. My “illustrations” were my pot-boilers; I looked to a different branch of art (far and away the most interesting it had always seemed to me) to perpetuate my fame.
[James, Tales, VIII. 231]
We must look outside the parenthesis to the right to find out what the aspiration refers to. As Dwight Bolinger observes in Aspects of Language (2nd edition, Harcourt, Brace, Jovanovich, 1975, pp. 607-08), setting off material with parentheses usually signals that it is relatively backgrounded (i.e., of lesser importance in context)— in contrast to the dash—and is of course assumed to be a self-contained unit, but here it is not self-contained, nor is the question of incidental importance—in fact, it appears to be three questions collapsed into one: “was it because of the outcome that he confesses it?”, “did the aspiration lead to either thing?”, “if so, which?” Notice that if the reader delays processing the material enclosed in parentheses until he has processed the main sentence (a possibility mentioned at the end of Chapter One), the antecedent will be available (i.e., in mind) when the parenthetical material is processed. Thus the temporal order of processing would reverse the serial order of presentation. I am indebted to Guy Carden for this observation. There is, by the way, a difficulty of the opposite sort with a different branch of art, which, being indefinite, appears to be introducing a new thing but does in fact refer to 'portraiture'— the sentence is, as it were, restating rather than advancing the previous sentence and hence reintroduces portraiture.
The largest and most ill-defined constraint on perception of reference is the imagination of the scene itself. Some of the disagreement among the readers of I. A. Richards' Poem 13 in Practical Criticism turns on what is in the referential scene. The poem, given without title, begins
(8) In the village churchyard she lies,
Dust is in her beautiful eyes,
No more she breathes, nor feels, nor stirs,
At her feet and at her head
Lies a slave to attend the dead,
But their dust is white as hers.6
Richards' students objected that either she has gone to dust or she has not: if she has, she does not have eyes anymore. Just possibly her beautiful eyes could be taken as referring to the eyes that we knew before she died, but the beginning of the third stanza raises the problem again:
Who shall tell us? No one speaks;
No colour shoots into those cheeks. . . .
which again seems an overwhelmingly trivial observation if she does not now have cheeks. Richards suggests that we could imagine a monument with her carven image in the churchyard, which certainly removes the objections, but the process by which one arrives at that scene, or does not, is very obscure. Some of my students not only have not seen, but refuse on demonstration to agree, that what is being described at the beginning of The Sound and The Fury is golf. Delaying information necessary for composing a scene becomes a kind of game with the reader for a writer like Pynchon (e.g., the beginning of Gravity’s Rainbow). But to say what principles are involved here would require a separate, large scale study. In all of the discussion to follow, I will assume that the reader has got the scene right when he reads into the passage.
The traditional stylistic precept—that the first eligible noun phrase to the left of the pronoun should be the antecedent— implies a process of leftward search where one scans back to the first eligible (i.e., masculine, singular, or whatever) noun phrase to the left of the pronoun and takes it as the antecedent. Possibly this precept originates in the notion that the immediately preceding noun phrase is surest to be 'up' in the reader’s mind. While there are doubtless times when one does engage in such a structurally based search, two other strategies may render it superfluous if conditions favor them: context may narrow the range of possible referents down to one, and the predicate may help pick out the individual of which it is predicated. In the following discourse, for example, one experiences little difficulty or even hesitation in identifying the referents of the pronouns:
(9) The messenger approching to him spake,
But his wast wordes returnd to him in vaine:
So sound he slept, that nought mought him awake.
Then rudely he him thrust, and pusht with paine,
Whereat he gan to stretch: but he againe
Shooke him so hard, that forced him to speake.
[Spenser, FQ, I.i. 42]
We construct an imagined scene with two individuals, one of whom is in a deep sleep, so that only one individual (the messenger) is likely as the Speaker, Thruster, Pusher, and Shaker: we do not have to engage in a leftward search, which in this case would be very difficult to carry out. Here then, even with pronouns, the referentially based strategy is the more efficient.
For a second, more complicated example where referential ’scene' is our best clue, consider the following passage cited by Sugden as typical of Spenser’s “careless, confused, and ambiguous manner” of handling his pronouns (p. 31). In the previous stanza Una and her lion have just left the hut of the blind Corceca and her deaf and dumb daughter Abessa:
(10) Whom ouertaking, they gan loudly bray,
With hollow howling, and lamenting cry,
Shamefully at her rayling all the way,
And her accusing of dishonesty,
That was the flowre of faith and chastity;
And still amidst her rayling, she did pray,
That plagues, and mischiefs, and long misery
Might fall on her, and follow all the way,
And that in endlesse error she might euer stray.
[Spenser, FQ, I.iii. 23]
Clearly the scene gives us one individual pursued and vilified and two pursuers—the problem is the shift from they to she when we get to the praying. But then we may remember that Abessa is incapable of speech (though she is presumably able to bray, etc.—this is quite realistic). S. K. Heninger helpfully supplies this observation in a note in his Selections from the Poetical Works of Edmund Spenser.7 Another problem is that we do not expect the Rayler to be the same as the Prayer—that is, we suspect a shift in reference from her to she in the sixth line only to read in the next line that it is a wicked prayer. Thus Spenser is not really “careless” here, but he does demand that very close track be kept of the capacities of the individuals in the scene.
Often what is predicated of a pronoun will help pick out its referent. Garvey, Caramazza, and Yates give a simple example:
Rosemary trusted the secretary because she was efficient.8
We tend to identify the secretary as the referent of she because efficiency is commonly predicated of secretaries. (This example is a little more complicated than they suggest: the because clause gives an explanation of why one was either the Subject or Object of trust: efficiency is a reason that someone would be the Object of trust, but not the Subject; hence the Object of trust is the one who is efficient. I am indebted to Jeanette Clausen for pointing this out.) Thomas Bever has constructed more complicated examples of the way predicates specify the referents of pronouns: (a) is said to be easier to comprehend than (b) though they are structurally identical:
(a) The box it rolled out of scratched the ball it contained.
(b) The shovel it was below broke the rake it fell on.9
Given only the two noun phrases as possible antecedents and no context, it is easier to identify ball as the Subject of roll and box as the Subject of contain than it is to identify rake as the Subject of was below and shovel as the Subject of fell on because a ball is the likely Roller and a box the likely Container, while in (b) the predicates do not help to choose between shovel and rake as their Subjects. It appears that searching for antecedents involves a kind of 'plugging in' of candidates in place of the pronoun to see whether the resulting clause is congruent and likely. The problem with (b) is that even the plugging in is not decisive.
Often several phrases must be orchestrated into the likeliest reading of the whole. Comprehension of the following passage requires the reader to decide whether the warning refers to Chad’s answer or to Strether’s note:
(11) He had announced himself—six months before; had written out at
least that Chad wasn't to be surprised should he see him some day
turn up. Chad had thereupon, in a few words of rather carefully
colourless answer, offered him a general welcome; and Strether,
ruefully reflecting that he might have understood the warning as a hint to
hospitality, a bid for an invitation, had fallen back upon silence as the
corrective most to his own taste.
[James, AMB, 68]
The question of what the warning refers to is related to the reference of the italicized he. If James obeys the rule of first eligible antecedent to the left, the antecedent of he would have to be Strether, and the warning would have to refer to Chad’s answer—which indeed from the description might be the sort of thing Strether might take as a warning. But the thing referred to as warning and hint to hospitality must also be the thing referred to as a bid for an invitation (since it is appositive), and Chad’s answer could hardly be a bid for an invitation, since Chad is the resident, Strether the ambassador. The alternative analysis, where he refers to Chad and the warning to Strether’s note, makes a better reading, since his note could be taken as a bid for an invitation. This example is particularly difficult because the reference of two noun phrases is uncertain: if the warning were replaced by the term announcement, the passage would be easier to perceive, since the only he who would be understanding the announcement would be Chad. The difficulty is certainly compounded by the first available antecedent to the left not being the correct one.
The way that particular decisions about referents are controlled by the need to compose the entire passage into a scene is further illustrated by the following two passages from Faulkner:
(12) It should have been later than it was; it should have been late, yet the yellow slashes of mote-palpitant sunlight were latticed no higher up the impalpable wall of gloom which separated them; the sun seemed hardly to have moved.
[Faulkner, ABS, 22]
(13) [sentence on previous page: When the shadow of the sash appeared on the curtains it was between seven and eight oclock and then I was in time again, hearing the watch.]... I got up and went to the dresser and slid my hand along it and touched the watch and turned it facedown and went back to bed. But the shadow of the sash was still there and I had learned to tell almost to the minute, so I'd have to turn my back to it, feeling the eyes animals used to have in the back of their heads when it was on top, itching.
[Faulkner, SF, 96]
In (12), if one looks leftwards for the first available noun phrase as the antecedent for them, one gets slashes, but why the wall of gloom would be said to separate the slashes is unclear, so we go to the context which gives us two figures sitting in the room with a wall of gloom between 'them' (on which the sunlight played). In (13), one may find oneself caught in a visualizing dilemma: if the shadow of the sash is projected by the morning sun, then for him to turn away from it would seem to mean turning toward the sun. Furthermore, if one imagines an erect animal with eyes in the back of its head, whatever was “on top” would not be seen anyway. These difficulties are almost enough to make one try the watch as the antecedent of it, but the business about the eyes is no clearer, since back in bed he couldn't see the hands of the dial anyway. The scene does cohere, however, if we imagine the shadow of the sash to be on the top of the curtain: then turning away is burying your head in the pillow, and eyes in the back of your head would still see the shadow when it was on top. Perhaps there are readers who visualize this scene correctly on first reading, but for the rest of us some careful weighing of alternatives is necessary: one sets up a possible interpretation and sees if it 'clicks'.
The process involved in these cases seems to be a kind of shuttling back and forth, reading new information and integrating it into the scene which may be only tentatively sketched in the mind. The final example of this sort of shuttling does not involve a literal spatial scene but rather a set of predications and inferences that allow one, finally, to select the best antecedent (which is sentential) for the italicized which, but only very slowly and as it were after the fact—that is, I find this passage impossible to process by any usual strategy and must tackle it with an analytic problem-solving routine:
(14) That even, I my self, the Author of these momentous Truths, am a Person, whose Imaginations are hard-mouth'd, and exceedingly disposed to run away with his Reason, which I have observed from long Experience, to be a very light Rider, and easily shook off; upon which Account, my Friends will never trust me alone, without a solemn Promise, to vent my Speculations in this, or the like manner, for the universal Benefit of Human kind; which, perhaps, the gentle, courteous, and candid Reader, brimful of that Modern Charity and Tenderness, usually annexed to his Office, will be very hardly persuaded to believe.
[Swift, TOT, Quintana Ed. 347]
The noun phrase immediately to the left of which is the universal Benefit of Human kind. This cannot be certainly rejected until we get to believe when it becomes obvious that it cannot be the Object of believe. 'That his friends make him promise to write' is next to the left, and is a proposition that could be the Object of believe (i.e., find it hard to believe), but in that case, the reader would be skeptical, not charitable: the proposition we are looking for is one that the author would consider damaging to himself, and the likeliest candidate is 'that he is a person whose imaginations run away with him' (the antecedent of which Account also)—they certainly ran away with him in this sentence!
These examples provide several instances where leftward search can lead us down a garden path if there is an eligible antecedent nearer the pronoun on the left than the correct one. I assume that the 'plugging in' or congruence testing is a kind of preliminary check and deals in such large categories as animate/inanimate, singular/plural, etc. Only if a noun phrase passes this initial screening does it qualify as a bothersome garden path. Further, one 'plugs in' to a structure that is not completely processed itself, and, as (14) shows, the information needed to force a reanalysis is often delayed. Finally, as Eugene Charniak points out, a reanalysis may never be 'forced' by an out and out incongruity: we must be ready to reject the nearest candidate to the left even if it is plausible, provided there is a more plausible candidate around (see the references of note 1). The general effect is that of working a puzzle where a piece seems to fit in a certain place and only appears not to after several other pieces are drawn in. The longer the mistake persists unnoticed, the greater will be the amount of puzzle torn up.
While writers may on occasion set up a garden path for their own purposes, the commoner case is for them to risk one now and then and hope the reader will not take it. Situations arise, for example, when one wishes to delay a relative clause so that it no longer immediately follows its antecedent. Milton, for one, does this quite a lot:
(15) All these and more came flocking; but with looks
Down cast and damp, yet such wherein appear'd
Obscure some glimps of joy, to have found thir chief
Not in despair, to have found themselves not lost
In loss it self; which on his count'nance cast
Like doubtful hue:
[Milton, PL, I. 522-27]
Loss it self is possible, if vague, as the antecedent of which, but the basic imagery is of light playing on Satan’s face, and glimps of joy appears to give a better reading, given the usual association of joy with radiance. Again in the next example:
(16) Our labour must be to pervert that end,
And out of good still to find means of evil;
Which oft-times may succeed, so as perhaps
Shall grieve him, . . .
[Milton, PL, I. 164-67]
Means of evil is marginally possible as the antecedent—if we are content to have means succeeding—but a better antecedent is labour. We might save ourselves the misperception if we take Milton’s semicolon preceding the relative clause as a warning to the reader not to take the previous noun phrase as the antecedent (see also PL, I. 435). Faulkner is another heavy user of relative clauses who takes certain chances:
(17) And the librarian knew why he was laughing, who had not called him anything but Mr Compson for thirty-two years now....
[Faulkner, p. 747]
In context, this is not very confusing: the librarian is a woman, and so by the time we get to called him we can see that he cannot be the Subject of called. The next example is also confusing only momentarily or for just a bit longer until we process the his gifts in the relative clause:
(18) For he whose daies in wilfull woe are worne,
The grace of his Creator doth despise,
That will not vse his gifts to thanklesse nigardise.
[Spenser, FQ, IV.viii. 15]
The one using “his gifts” must be the creature, not the Creator. The next example from Faulkner is a little more confusing out of context:
(19) the Negro woman, his sworn enemy since his birth and his mortal one since that day in 1911 when she too divined by simple clairvoyance that he was somehow using his infant niece’s illegitimacy to blackmail its mother, who cooked the food he ate.10
The Negro woman, not the mother, cooks his food, though this may not be totally certain until the next page when reference is made to the Negro cook. However, the attentive reader may have been able to identify “the mother” of the infant niece as Jason’s sister Caddy, who, we have been told three pages earlier, had abandoned her infant daughter to Jason’s keeping never to return. Hence she is not “in the scene” as a possible referent of who. (The repetition of infant in the two noun phrases may help the reader to perceive the coreference.) Interestingly, in The Portable Faulkner the troublesome relative clause appears immediately following the Negro woman, where of course it occasions no difficulty, the garden path being avoided (p. 750).
Thus far a leftward search has always produced an antecedent or at least a referent. It would seem that such a process would encounter most difficulty in cases of 'backward pronominalization'— where the pronoun precedes its 'antecedent,' which is on the right:
The impression that one must search to the right in such cases may be erroneous, however, for Susumo Kuno has argued that there is a general constraint on discourses that the pronoun may precede its 'antecedent' only if it is predictable from prior discourse—if, that is, the individual the pronoun refers to has already been introduced and is 'on stage' or given in the situation of utterance.11 Hence, Kuno notes, the pronoun can never have an indefinite 'postcedent':
Since an indefinite noun phrase is used to refer to a new individual, Kuno’s constraint explains why it cannot be a postcedent: the individual must already have been introduced. Thus Bever’s examples concerning the ball and the box, rake and shovel, are artificial insofar as no previous context is given: the reader must therefore really make a rightward search for the postcedent of it. Equally, Kuno’s principle renders unnecessary Bever’s perceptual principle devised to account for acceptability judgments in backwards cases ("Cognitive Basis," p. 319).
Kuno’s principle appears to hold over the great majority of cases of apparent backward pronominalization in our texts, even in the following passage where the referent of they is represented to the left as well as to the right, though perhaps not obviously so, since the first occurrence of afternoon and evening does not set them up as 'things'—rather, the phrase tends to be read in terms of “spent time.” Instead, the immediate and the sensible appears the likely antecedent, hence we have a garden path:
(20) he had . . . given his afternoon and evening to the immediate and the sensible. They formed a qualified draught of Europe, an afternoon and an evening on the banks of the Mersey, but such as it was he took his potion at least undiluted.
[James, AMB, 18]
Compare the following passage where there is no garden path to distract the reader:
(21) Such had at any rate markedly been the case for the precipitation of a special series of impressions. They had proved, successively, these impressions—all of Musette and Francine, but Musette and Francine vulgarized by the larger evolution of the type—irresistibly sharp:
[James, AMB, 67]
Notice by the way that (20) seems to involve a violation of Kuno’s constraint, in that the specifying noun phrase is indefinite (an afternoon....). Clearly, however, the indefinite here is a special use meaning roughly “any such”—substituting that for an would insist too much on the specific afternoon and evening. Example (21) illustrates Kantor’s point about demonstratives: James concludes that the reference of the pronoun they is not sufficiently clear (it refers to a subpart of the preceding noun phrase the precipitation...) and signals the extra effort by the demonstrative these (impressions). So also with (23) and (24) below. (On this construction, see Dwight Bolinger, “Pronouns and Repeated Nouns,” pp. 25-26.)
Two factors make the following passage genuinely hard: it does violate Kuno’s constraint, and it sets up a garden path: the leftward these ("shapes") is to be contrasted with the new these (things beneath the shadow of a shape):
(22) The bed, the books, the chair, the moving nuns,
The candle as it evades the sight, these are
The sources of happiness in the shape of Rome,
A shape within the ancient circles of shapes,
And these beneath the shadow of a shape
In a confusion on bed and books, a portent
On the chair, a moving transparence on the nuns,
A light on the candle tearing against the wick
To join a hovering excellence, to escape
From fire and be part of that of which
Fire is the symbol: the celestial possible.
[Stevens, CP, 508-€9]
The new these, that is, are the portent, transparence, and light. As before, the passage is confusing not just because the 'antecedent' can be found only to the right, but also because there appears to be a very likely antecedent to the left, which, according to the general strategy of leftward search, gets tried first.
We can explain the difficulty of these passages by assuming that leftward search has precedence over rightward search—if indeed one ever does make a rightward search. In general, I think one does not, but there are exceptions. The first is the pronoun it when sentence-initial (this is different from Bever’s examples, where it is not initial). It does point to the right in both the cleft construction (It was Oswald who killed Kennedy) and in extraposed sentences (It is surprising that you forgot his name). James does use it rather loosely related to its context with a genuine specification following:
(23) [He] had presently strolled back to the Boulevard with a sense of injury that he felt himself taking for as good a start as any other. It would serve, this spur to his spirit, he reflected, as, pausing at the top of the street, he looked up and down the great foreign avenue, it would serve to begin business with.
[James, AMB, 57]
(24) She abounded in news of the situation at home, proved to him how perfectly she was arranging for his absence, told him who would take up this and who take up that exactly where he had left it, gave him in fact chapter and verse for the moral that nothing would suffer. It filled for him, this tone of hers, all the air;
[James, AMB, 60]
(25) His greatest uneasiness seemed to peep at him out of the imminent impression that almost any acceptance of Paris might give one’s authority away. It hung before him this morning, the vast bright Babylon, like some huge iridescent object....
[James, AMB, 64]
and this use of it, sometimes even without any preparation in context, is typical of Stevens also:
(26) It can never be satisfied, the mind, never.
[Stevens, CP, 247]
(27) It is a child that sings itself to sleep,
The mind, among the creatures that it makes.
[Stevens, CP, 436]
(and perhaps also the sentence cited in [5]: It is a wheel, the rays/Around the sun.). The effect in all of these is that of the meditating mind making little leaps forward, and specification is never long delayed. The point is that the reader is prepared to look to the right for further specification when he sees a sentence-initial it, so these constructions are not hard to process.
The second exception are uses of pronouns (and definite noun phrases) at the beginning of stories and elsewhere in tales told in the free indirect style. Kuno stresses the beginnings of stories as points where his constraint does not hold, though of course confusion will not arise, as leftward searching is not possible. Beginning with a definite form has a kind of in medias res effect: the reader must scramble to find his bearings in a world that is, as it were, going on without regard for him. For example, the they at the beginning of (27) in Chapter Two has not yet been specified—it is the third sentence of the story. Note that the reader does not have to choose among possible referents already in the scene—rather, he craves more information about those that are being introduced. In the free indirect style, material is presented from the character’s point of view, so that things are treated as 'in mind' by use of definites which have not been previously mentioned.12 It should be noted, however, that rightward search will not necessarily pay off either: what the reader must do is imagine himself into the character’s point of view as quickly as possible so that he will be able to share what the character is treating as 'given'.
The last exceptional situation, which is almost a trick rather than an exception, occurs in Faulkner when he presents the significance of a thing before naming it. Robert Zoellner cites one such sentence from Absalom, Absalom! where Sutpen “just walked on, erect, with the new hat cocked and carrying in his hand now that which must have seemed to them the final gratuitous insult....” There follow nine printed lines before the final participial identifies what he was carrying ("carrying his newspaper cornucopia of flowers").13 There is a similar sentence in (44) of the next chapter where lines and lines of ornate description precede the name of the thing described (the old bear).
None of these exceptions really requires the reader to innovate a strategy for rightward search. The reader already has a rightward strategy for initial it, and he simply will have to endure the suspension of beginnings and of Faulkner’s delayed 'revelations'. There is a general sense in which we read ahead hoping that things which are yet somewhat unclear will fall into place with further information—we will see some further illustrations of this in the next section—and this is probably the way we read into a story or through a passage of suspended identification.
It is possible to regard the processing of subjectless participials as either a problem of perception (since one must establish a modifier-to-head relation called 'attachment') or one of comprehension (because of the similarity of participials to relative clauses). Though there is no relative pronoun present, participials usually immediately follow the noun phrase they modify (though they can precede the Subject if they modify it), but, like relative clauses, they can be shifted to the end of the clause and attach not to the preceding noun phrase but to an earlier one, usually the Subject. The parallelism with relative clauses is captured in a transformational framework by deriving participials from relative clauses via deletion of the relative pronoun and the form of be:
A girl came in (who was) pursued by a bear.
It is not obvious, however, that comprehension involves restoring a relative pronoun and then searching for its antecedent, and so I treat the attachment problem as distinct from those involving coreference, although the parallel will be obvious.
It is quite crucial that we distinguish between reconstructions the reader must make to perceive a text from those he need not. Reconstruction (or identification) of a deleted Logical Subject in a passivized construction (i.e., the Logical Subject of the corresponding active sentence) has sometimes been treated as necessary, sometimes not—indeed, contradictory claims on this point provided Stanley Fish with some of his most telling arguments against transformational stylistics.14 In a model of perception that involves reconstructing a deep structure of passivized constructions like
(28) Infinite goodness, grace and mercy shewn
On Man by him seduc't, but on himself
Treble confusion, wrath and vengeance pour'd.
[Milton, PL, I. 218-20]
by 'undoing' passive to get (in part),
... PRO show infinite goodness, grace and mercy on Man...
... PRO pour treble confusion, wrath, and vengeance on Satan...
the indefinite pronoun appears as a restored Logical Subject and presumably would have to have an antecedent assigned to it, or at least a referent from among the individuals in the world of the work. Thus Seymour Chatman claims that at some level we specify God as the deleted Subject and, as it were, unwittingly acknowledge His power.15
An alternative view is possible, however, which treats passivization as a way of making transitive (i.e., two-place) predicates intransitive (i.e., one-place). The logical form of a passivized construction would be satisfied even if the Subject of the corresponding active were not specified—it might appear as an optional modifier (as it does in by him seduc't), but the process of perception would be complete without it, as is the construction. This is roughly the view of passives presented by Ronald Langacker and Patricia Munro in their article “Passives and Their Meaning.”16 This does not mean that the reader may not go on on his own to specify a deleted Subject—only that passives are not “hidden persuaders” or inherently tools of “brainwashing” as Chatman among others assumed.17 Certainly the traditional warning about the deviousness of passives is based on the assumption that we may not reconstruct the deleted Logical Subject.18
The question that we must ask in perceiving a participial, however, is to which noun phrase it is attached—which noun phrase, that is, would be the antecedent of the relative pronoun if it were present. When a participial is delayed and does not immediately follow the noun phrase it attaches to, other noun phrases may intervene, and the possibility of a garden path again arises. Consider, for example, the following two passages from Paradise Lost:
(29) While smooth Adonis from his native Rock
Ran purple to the Sea, suppos'd with blood
Of Thammuz yearly wounded:
[Milton, PL, I. 450-52]
(30) [scene: Uriel observing Satan] his gestures fierce
He mark'd and mad demeanour, then alone,
As he suppos'd, all unobserv'd, unseen.
[Milton, PL, IV. 128-30]
In (29) suppos'd... wounded could be attached to the Sea, since it is the immediately preceding noun phrase, but this reading does not make much sense, since there is no reason to think the sea is being viewed as having a body, so we must go back and try rock or Adonis, and of course Adonis is semantically congruent with wounded—the problem is really how it can also make sense as the name of a river, but one concludes that Milton is collapsing natural fact and its mythic explanation here. Example (30) is also momentarily difficult, since both he’s refer to animate beings. The context has Uriel observing Satan, and the best way to tell whether then alone. . . unseen is to be attached to he = Satan or he = Uriel is to note that Uriel would not suppose himself alone, since he is watching Satan. As long as the semantic incompatibility of the mistaken noun is fairly obvious, the mistake is quickly noticed:
(31) where stood
Her Temple on th'offensive Mountain, built
By that uxorious King....
[Milton, PL, I. 442-44]
The mountain cannot have been built—it must be the temple. Contrast the next example, where semantic clues to attachment are a bit more indirect:
(32) she loved him not only in spite of but because of the fact that he himself was incapable of love, accepting the fact that he must value above all not her but the virginity of which she was custodian and on which she placed no value whatever: the frail physical stricture which to her was no more than a hangnail would have been.
[Faulkner, p. 744]
Accepting the fact... could be attached to either he or her. But since love usually involves acceptance, the Lover (’she') is likely to be the Acceptor also. It is perhaps somewhat easier to find the noun to which possessing. . . and being. . . should be attached in the next example, since the alternatives truck and horses can be quickly ruled out:
(33) It was himself, though no horseman, no farmer, not even a countryman save by his distant birth and boyhood, who coaxed and soothed the two horses, drawing them by his own single frail hand until, backing, filling, trembling a little, they surged, halted, then sprung scrambling down from the truck, possessing no affinity for them as creatures, beasts, but being merely insulated by his years and time from the corruption of steel and oiled moving parts which tainted the others.
[Faulkner, pp. 712-13]
The sheer length of the participials appears to be one of the sources of difficulty in Faulkner. We could account for this if we supposed that we may wait to compose the entire participial before deciding what it is attached to, or, if we encounter uncertainty, to read forward into the participial hoping for further clues as to what was affected. The more we can find out about 'what happened', the easier it should be to infer whom or what it happened to. So in (34) (which repeats [51] of Chapter One) the final participial does not contain enough information to allow a certain choice among mother Earth, her, or even him, but the first line of the next stanza does:
(34) Her sire Typhoeus was, who mad through merth,
And drunke with bloud of men, slaine by his might,
Through incest, her of his owne mother Earth
Whilome begot, being but halfe twin of that berth.
48
For at that berth another Babe she bore....
[Spenser, FQ, III.vii. 47-48]
That berth, we conclude, is the consequence of the begetting just mentioned; therefore it must be her (Argante’s) birth, but we cannot see why she (Argante) is just a half twin until we get to the next stanza.
Empson cites a number of cases where the attachment of participials is uncertain, among them the following lines from “The Waste Land”
(35) Reflecting light upon the table as
The glitter of her jewels rose to meet it,
From satin cases poured in rich profusion;
[Empson, Ambiguity, p. 77]
He comments: “What is poured may be cases, jewels, glitter, or light, and profusion, enriching its modern meaning with its derivation, is shared, with a dazzling luxury, between them; so that while some of the jewels are pouring out light from their cases, others are poured about, as are their cases, on the dressing-table.” Empson nicely suggests the sort of shuttling back and forth the reader executes as he checks the compatibility of each of the likely candidates for attachment. This shuttling exactly parallels that described in the case of pronouns above.
The next passage again sets up this shuttling back and forth, which in this case is complicated by the length of the participial—one perhaps begins checking back before the participial is completely processed, or identifies the noun phrases to which the participial might be attached, hoping that by the time he finishes processing the participial he will be able to guess which among the candidates is right:
(36) For six years now he had heard the best of all talking. It was of the wilderness, the big woods, bigger and older than any recorded document—of white man fatuous enough to believe he had bought any fragment of it, of Indian ruthless enough to pretend that any fragment of it had been his to convey.... It was of the men, not white nor black nor red, but men, hunters, with the will and hardihood to endure and the humility and skill to survive, and the dogs and the bear and deer juxtaposed and reliefed against it, ordered and compelled by and within the wilderness in the ancient and unremitting contest according to the ancient and immitigable rules which voided all regrets and brooked no quarter....
[Faulkner, p. 227]
Obviously the problem begins with the slightly uncertain reference of the it immediately preceding the participial in question. There appear to be two leading candidates: the wilderness and the best of all talking. The two are close, of course, in that the tale is of the wilderness, but wilderness seems more congruent with juxtaposed against since it has a physical location. We can tentatively conclude then that the participial does not attach to it (= the wilderness) but attaches to the long coordinate noun phrase the men . . . and the dogs and the bear and the deer (and is then essentially appositive to the preceding participial juxtaposed . . .). All of the members of this long coordinate noun phrase are congruent as the Subject of ordered and compelled. . . . But one may also wonder whether the participial attaches to it (=the best of all talking)—it too is congruent if we think of the story as shaped and necessitated by its subject and heard by the boy each year on the hunt. Perhaps the material immediately following the dash will help:
the best game of all, the best of all breathing and for ever the best of all listening. ...
The best game of all appears to be an appositive coreferential to the ancient and unremitting contest, but what are the best of all breathing and the best of all listening coreferential to? The last would seem to be the best of all talking, with the breathing as perhaps a bridge linking the hunt with the tale of the hunt. One senses that Faulkner is deliberately fusing the primary experience of the hunt with the verbal telling of it through difficult pronominal reference, uncertain attachment, and an obscure sort of summarizing appositive. I do believe that there is a 'best reading' here along the lines I have indicated, but I also believe that the other possibilities (a veritable maze of garden paths) contribute to the effect Faulkner is aiming at. One can see here, albeit darkly, how apposition figures in the pursuit of coreference, and we will return to the fusional effect in Faulkner at the end of the next chapter.
In this chapter again we have seen ways that processing involves both semantic and syntactic strategies. Comprehension of definite noun phrases and pronouns takes place in relation to an imagined scene: the referent must be identified from among those present in the scene, or those which can be imagined, as the case requires. In many instances, the range of possible referents is limited or narrowed down to one by the scene, so that no further strategies are needed. There does seem to be a process of leftward search, however, that gets underway in problematic cases involving a kind of trial-and-error matching of potential antecedents to the pronoun, and head nouns to participial, which can reject or continue to try candidates as the clause is processed. Thus (17) and (18) are relatively easy to make out, (29) somewhat harder, and similarly (31) and (33) may be easier than (32). There seems to be a clear tendency for the nearest candidate on the left to be tried first—hence the possibility of garden paths—though we should note that this tendency is most marked with relative pronouns and participials, since these generally do immediately follow the noun phrases they modify (other pronouns typically do not immediately follow their antecedents). I have assumed that when the immediately prior noun phrase proves unsuitable as the antecedent of the relative pronoun, one searches left for the antecedent just as one does with a personal pronoun (rather than pursue an integrative strategy of moving the entire clause back to its antecedent to create a 'canonical form' for perception of relative clauses). This may be incorrect, but it seems right to me. In the next chapter, which is a continuation of this insofar as comprehension of appositives also involves coreference, we will have further opportunity to explore strategies for perceiving delayed or strayed coreferential elements and once again to consider the interaction of semantic and structural processes.
We use cookies to analyze our traffic. Please decide if you are willing to accept cookies from our website. You can change this setting anytime in Privacy Settings.