1. STRUCTURAL CUES TO CLAUSE BOUNDARIES

The recognition of a clause is partly dependent on the recognition of the functional elements. We may not be sure whether we are reading a clause when we encounter the first noun phrase, but when we get to a main verb, we can be certain that we should be constructing a clause; similarly, if the verb is unequivocally transitive, or otherwise complement-taking, we will not close off the clause until we locate the Object or other complement—though once we have located the obligatory elements, we will close it off as quickly as we can. When a functionally complete (or apparently complete) clause aligns with a line boundary, the temptation to close the clause prematurely is even greater. Consider, for instance, these first lines from Shakespeare’s Sonnets 15 and 33 cited by Stephen Booth:²

(1) Full many a glorious morning have I seen
Flatter the mountain tops with sovereign eye

[Shakespeare, Son. 33]

(2) When I consider everything that grows
Holds in perfection but a little moment

[Shakespeare, Son. 15]

The first line of (1) appears to be functionally complete when we get to the verb at the end of the line, which allows recognition of the initial noun phrase as a fronted Object, but the verb in the second clause forces reanalysis of it as the Subject of a clause, the entire contents of which are the Object of seen. Similarly in (2), lacking the complementizer that, we perceive everything that grows as the complete Object of consider and close off the clause, only to have to reanalyze it as the Subject of holds. To be sure, the garden path is very short in these cases and the perceptual effect a kind of double-take—I merely wish to show that apparent functional completeness combined with a line ending results in misperception.

It is not always obvious, however, that the verb is transitive, so the decision to treat it as such can depend on spotting a noun phrase likely to be its Object. Perhaps for this reason, insertion of adverbs or parenthetic material between the verb and its Object is generally felt to be awkward. Howsoever that may be, insertion in that position with a verb that is possibly intransitive will impede perception by promoting premature closure in just the way line ends do:

(3) Each time she turned in again, each time, in her impatience, she gave him up, it was to sound to a deeper depth, while she tasted the faint, flat emanation of things, the failure of fortune and of honour.

[James, WOD, I. 3-4]

Sound could be intransitive (either “plunge” or “measure”), and it is not until we recognize that the failure of fortune and of honour is not really appositive to the faint flat emanation of things, but is in fact the Object of sound, that we are able to class sound as transitive. Further, a normally transitive verb can often occur with the Object ellipsed if it can be inferred from context, so that the absence of an Object does not automatically rule out the possibility that the verb is to be understood transitively. There is a rather unusual ellipsis of an Object in the next example:

(4) He had already relinquished, of his will, because of his need, in humility and peace and without regret, yet apparently that had not been enough, the leaving of the gun was not enough.

[Faulkner, p. 243]

I experience an almost intolerable suspension following relinquished in searching for an Object which never comes. One further consideration is that certain orders of elements seem to be canonical for recognition. For example, if a verb takes an Object and an Adverbial, we expect the Object before the Adverbial, and will be inclined to close the clause after the Adverbial even if we have yet no Object (it might have been ellipsed). Here are two examples (omitting the well-known Throw Mama from the train a kiss.):

(5) the planters with their gangs of slaves and then of hired laborers had wrested from the impenetrable jungle of water-standing cane and Cyprus, gum and holly and oak and ash, cotton patches which, as the years passed, became fields and then plantations.

[Faulkner, pp. 710-11]

(6) Shall she not find in comforts of the sun,
In pungent fruit and bright, green wings, or else
In any balm or beauty of the earth,
Things to be cherished like the thought of heaven?

[Stevens, CP, 67]

In reading these, one tends to forget about the missing Object of wrest and find, reading the last noun phrase as appositive to the preceding. This misreading serves Stevens' purpose nicely (that is, we find them to be identical), and we might note that to unravel Faulkner’s sentence (actually only the last part of a sentence) we must wrest the cotton patches from the preceding phrases. Similarly in the next three examples, once we complete the Adverbial and optional modifiers, we have a tendency to close the clause and be a little surprised by the Object (or Complement, in the case of [8]):

(7)           thir summons call'd
From every Band and squared Regiment
By place or choice the worthiest;

[Milton, PL., I. 757-59]

The Regiment is not the worthiest—rather, the worthiest are called from every band and regiment.

(8) [we are doomed] to remaine
In strictest bondage, though thus far remov'd,
Under th' inevitable curb, reserv'd
His captive multitude:

[Milton, PL., II. 320-23]

It is possible to read his captive multitude as appositive to we, but the sense is sharper as “remain.... reserved, his captive multitude.”

(9) his cousin McCaslin brought him for the first time to the camp, the big woods, to earn for himself from the wilderness the name and state of hunter

[Faulkner, p. 228]

With earn, as with remaine, one cannot be sure that an Object or Complement is to follow, and as the Adverbials accumulate, one begins to suspect not—and is slightly surprised when one does. It is important to note that interpolation of material between the verb and its Object does not necessarily produce perceptual difficulty. Such interpolations are characteristic of Joyce’s style (or styles), for example, but, as Anthony Burgess notes in Joysprick (André Deutsch, 1973, p. 70), the interpolated matter is usually only an Adverb or a short phrase. As long as one can easily look across the material to check for an Object, interpolation of material between Verb and Object requires only a slight adjustment by the reader.

Certain forms, in general, indicate the beginning of a clause: nothing to the left of these forms is part of the clause initiated by the word. These forms include the subordinating conjunctions if and when and relative pronouns. John Robert Ross claims that a relative clause with an element shifted to the left of the clause is ungrammatical.³ (Ross notes that we may regard the non-extractability of elements from adverbial clauses as coming under this constraint if we regard adverbial clauses as underlying relative clauses: when = at the time at which....) However that may be, all of the following examples, culled from the early cantos of The Faerie Queene, are hard to perceive: we do not expect the italicized element to be part of the clause which follows and look in vain for a main verb after the subordinate clause to match with it:

(10) . . . the labyrinth about;
Which when by tract they hunted had throughout,
At length it brought them to a hollow caue,...

[Spenser, FQ, I.i. 11]

(11) Deuoure their dam; on whom while so he gazd,
Hauing all satisfide their bloudy thurst,
Their bellies swolne he saw with fulnesse burst,
And bowels gushing forth:

[Spenser, FQ, I.i. 26]

(12) of whose most innocent death
When tidings came to mee, vnhappy maid,
O how great sorrow my sad soule assaid.

[Spenser, FQ, l.ii. 24]

(13) Like a young Squire, in loues and lusty-hed
His wanton dayes that euer loosely led,
Without regard of armes and dreaded fight:

[Spenser, FQ, I.ii. 3]

Note that the main source of difficulty is not that an Object or other argument is shifted to the left, since that occurs in Topicalized sentences—it is rather that our normal perceptual strategies are set not to integrate material outside a relative or adverbial clause into it or to seek missing phrases outside (i.e., to the left of) the relative pronoun or subordinating conjunction. There is an extra source of difficulty with these examples, however, which is that the subordinate clause is not obviously missing anything. Hence, should it occur to us that the relative pronoun does in fact belong in the clause, we cannot quickly see a place for it. This is most strikingly the problem in (12), where one must scratch one’s head a bit before deciding that the fronted prepositional phrase is the complement of tidings, but it is also a problem in the others, where gaze, hunt, and led are not so obviously transitive as to attract the relative pronoun to them (actually, the which in [10] is probably the complement of throughout: “when by tract they had hunted throughout the labyrinth”). There are other examples at I.iii. 13 and 26. Contrast the following, where the verbs are obviously transitive or otherwise complement-seeking:

(14) That none did others safety despize,
Nor aid enuy to him, in need that stands,
But friendly each did others prayse deuize,...

[Spenser, FQ, I.ix. 1]

(15) The fift had charge sicke persons to attend,
And comfort those, in point of death which lay;

[Spenser, FQ, I.x. 41]

(16) Which when the valiant Elfe perceiu'd, be lept
As Lyon fierce vpon the flying pray,

[Spenser, FQ, I.i. 17]

(17) His forces faile, ne can no longer fight.
Whose cor age when the feend perceiu'd to shrinke,
She poured forth out of her hellish sinke
Her fruitfull cursed spawne of serpents small,

[Spenser, FQ, I.i. 22]

(18) Which when he saw, he burnt with gealous fire

[Spenser, FQ, I.ii. 5]

Stands in (14) and lay in (15) attract the preposed Adverbial to them, and perceiue and see attract the preposed Objects in (16-18). In (19) there is a special difficulty: attaine is obligatorily transitive, but it is incongruous to attain a person:

(19) To seeke her strayed Champion, if she might attaine.

[Spenser, FQ, I.iii. 8]

I have spent so much time on this particular construction because of its theoretical interest: it is probably ungrammatical and certainly rare in most reader’s experience, so that readers presumably have no existing strategy for coping with it and a bias against it, but it is so common in Spenser that one must work out some strategy for it. If the reader does in fact find (14-18) easier than (10-13), he probably is adapting the strategy for identifying fronted elements outlined in the last chapter.

Spenser, by the way, so frequently inverts to + verb with its Object or an Adverbial that the usual function of to to signal the beginning of an infinitival Verb Phrase is virtually nullified:

(20) At last resoluing forward still to fare

[Spenser, FQ, I.i. 11]

(21) Therewith enrag'd she loudly gan to bray,
And turning fierce, her speckled taile aduanst,
Threatning her angry sting, him to dismay:

[Spenser, FQ, I.i. 17]

This case differs somewhat from the previous one in that the reader may well possess some perceptual principle for inserting material that appears before the to into the phrase following it, since one way not to ’split' an infinitive is to place the adverb before the to. In Spenser, we have only to add Objects to the list of things that can precede the to. The strain on perception is further mitigated by the fact that the material shifted out of the infinitival Verb Phrase is always shifted just to the immediate left of the to, or in the case of (20), one word over.⁴

Most of the last ten examples are noticeably harder to process than the initial set of ten. This is so, I think, because the first examples have mainly to do with premature closure of a clause rather than a radically false parsing involving misidentification of functions. One gets what appears to be a complete, well-formed clause, and correction of the error usually involves adding material to the preliminary structure rather than recasting it. The latter examples from Spenser involve a perceptual difficulty of an entirely different nature—namely, one cannot see how to construe the pieces into a clause at all. The experience is one of bafflement, not of committing a mistake that requires slight correction. We will examine the two other types and grades of difficulty in the next sections: the following section will examine uses of the semicolon that mislead by signaling that a clause is complete—the problem for perception is realizing that an apparent main clause is not independent of one that it precedes or follows but is dependent on it. This realization involves no restructuring of the basic clauses and is thus like the premature closure examples of this section. In the third section, however, we consider constructions where noun phrases appear to be part of one clause but are actually part of another: correction of such misperceptions does involve a major restructuring, and the examples do seem to be more difficult.

2. PUNCTUATION AS A CLUE TO CLAUSE BOUNDARIES

In the most general terms, marks of punctuation are used in writing to indicate groupings and closures of various types. Typically, different marks can be used to indicate the same syntactic boundary; for example, complete main clauses may be separated by a period, a dash, a colon, a semicolon, or a comma (plus a conjunction). Choice of mark in such cases conveys 'degree' of separation or integration of the clauses as units of comprehension: using a semicolon instead of a period, for example, suggests greater integration or cohesion between the material in the two clauses; use of a colon indicates an even tighter and more specific kind of integration (the second clause is a specification of, or [in my usage] a consequence of the previous clause); the dash indicates a discontinuity in the line of thought—possibly interrupting the main clause—though one that is prompted by the preceding clause. The degree of syntactic integration is the same in all of these cases—each clause must be a complete, main (i.e., not subordinate) clause—but the degree of cohesion differs in ways that would be signaled in speech by intonation (fall or half-fall or level) and pause. This function of punctuation, sometimes called 'rhetorical,' is not opposed to its function of marking syntactic boundaries and is in a sense based on it: a mark of a relatively light syntactic boundary (such as a comma) would never indicate a rhetorical separation greater than a mark of a heavier syntactic boundary (such as a semicolon) where both are possible. Thus the notion put forth by various critics that punctuation is either rhetorical or 'logical' (syntactic) is over-simplified from its outset. This dichotomy would hold only if there were some use of punctuation where each mark indicated a certain syntactic boundary uniquely: that would be a truly syntactic system. Marks in such a system would be very good clues to syntactic boundaries. In regard to real punctuation, we can say that a mark is useful as the basis of a perceptual strategy insofar as it regularly marks the same syntactic boundary.

Few marks of punctuation are as reliable a mark of a main clause boundary as the semicolon: most modern grammar books give the very simple and straightforward advice to use it to separate complete main clauses and only main clauses. However, reading the following passage by T. S. Eliot with this function of the semicolon in mind produces problems:

(22) Donne, I suppose, was such another,
Who found no substitute for sense;
To seize and clutch and penetrate,
Expert beyond experience,
 
He knew the anguish of the marrow
The ague of the skeleton;
No torments possible to flesh
Allayed the fever of the bone.

["Whispers of Immortality" cited
in Empson, Ambiguity, p. 79.]

The semicolon at the end of the second line signals the end of a complete clause, forcing us to find a place for the infinitive of purpose in the third line in the clause that follows it. Empson fits it in as the complement of expert (i.e., “expert at seizing, etc.”), or it could be fitted in as the complement of anguish (i.e., “the anguish to seize and clutch”—which is rather nice, except for the ague). Neither parsing yields a perfectly idiomatic sentence, and Empson notes that the third line could go semantically with the second— the semicolon is the barrier. Interestingly, in the Collected Poems, 1909-62 (Harcourt, Brace, and World, 1963), the semicolon does appear after the third line rather than the second, thus obviating the need for Mr. Empson’s ingenuity and my own. Essentially, the semicolon in modern texts indicates that one main clause has been completed and another is about to commence. Hence what follows the semicolon cannot be subordinate to what precedes (or vice-versa) or appositive to it—rather, it must be construed as part of a new clause. The next examples suggest that in reading James, we must lower this expectation slightly:

(23) It was at this point, however, that she remained; changing her place, moving from the shabby sofa to the armchair upholstered in a glazed cloth that gave at once—she had tried it—the sense of the slippery and of the sticky.

[James, WOD, I, 3]

(24) He pulled himself then at last together for his own progress back; not with the feeling that he had taken his walk in vain.

[James, AMB, 67]

The modification we must make when reading James is that what follows the semicolon may be some sort of loosely attached participial or supplemental clause or phrase, not another main clause, but we can at least rely on the material preceding the semicolon being a complete main clause. (Note by the way in [23] the parenthetical “—she had tried it—” splitting a verb and its Object: the sentence is easy to misread, since gave can be taken here as intransitive ["She tried the armchair and it gave at once"]. The optional Adverbial following gave also helps to induce this premature closure.) What makes these examples vexing, particularly (23), is the amount of time the reader can pursue the garden path assumption that a Subject and verb are yet to come. The next three examples are even worse, since what follows the semicolon is a noun phrase which apparently is the Subject of the next main clause—we search and search for the main verb:

(25) He bad awake blacke Plutoes griesly Dame,
And cursed heauen, and spake reprochfull shame
Of highest God, the Lord of life and light;
A bold bad man, that dar'd to call by name
Great Gorgon, Prince of darknesse and dead night,
At which Cocytus quakes, and Styx is put to flight.

[Spenser, FQ, I.i. 37]

(26) He scarce had ceas't when the superiour Fiend
Was moving toward the shoar; his ponderous shield
Ethereal temper, massy, large and round,
Behind him cast; the broad circumference
Hung on his shoulders like the Moon....

[Milton, PL, I. 283-87]

This last one is frustrating because we can almost wrestle the material following the first semicolon into a clause, especially if we assume an ellipsed he ("he cast his shield behind him"), but there is no conjunction linking the two 'clauses', and it is therefore more regular grammatically to take cast as a participle, not a main verb ("his shield cast behind him").

(27) At first they had come in wagons: the guns, the bedding, the dogs, the food, the whiskey, the keen heart-lifting anticipation of hunting; the young men who could drive all night and all the following day in the cold rain and pitch a camp in the rain and sleep in the wet blankets and rise at daylight the next morning and hunt.

[Faulkner, p. 706]

After the fact, we can recognize the semicolon here as obeying the 'rule of weight'—the young men ... is to be taken as appositive to the entire preceding series, and since that series is punctuated with commas, the next heavier mark must be used. But as we read, the semicolon looks like a clause boundary. Consider as a final example (28):

(28) She was handsome, but the degree of it was not sustained by items and aids; a circumstance moreover playing its part at almost any time in the impression she produced.

[James, WOD, I. 5]

It might be more obvious that the material following the semicolon depends upon the clause preceding it rather than initiating a new clause if it were clearer what a circumstance referred to.

It would appear, then, that we should not conclusively close off the first clause and open a new clause when we reach a semicolon, for the material following the semicolon may have to be integrated back into the first clause. Either we might check the material following the semicolon for completeness as a clause before clearing the first clause, or we might check the material for possible attachment to the first clause as we read (this would work well in [23] and [24], a little less well in [28]).

Whatever procedure we adopt, we will have to use it more when we read texts of Spenser and Milton with original punctuation, for they frequently use semicolons between main and subordinate clauses. As long as the beginning of the subordinate clause indicates that it is subordinate, however, the semicolon preceding it will not be too misleading. Usually subordinators do begin conditional clauses (if), comparative and result clauses (as, so) purpose (for, to), temporal (when, till), and relative clauses (the relative pronoun). So (29) is not too hard to make out:

(29)   For Spirits when they please
Can either Sex assume, or both; so soft
And uncompounded is thir Essence pure....

[Milton, PL., I. 423-25]

When the subordinate clause comes first, however, and is set off from the main clause by a semicolon, no mark of subordination is present to warn the reader that the subordinate clause is to be integrated into the one to follow:

(30) As gentle Shepheard in sweete euen-tide,
When ruddy Phoebus gins to welke in west,
High on an hill, his flocke to vewen wide,
Markes which do byte their hasty supper best;
A cloud of combrous gnattes do him molest....

[Spenser, FQ, I.i. 23]

(31)   Or if Sion Hill
Delight thee more, and Siloa’s Brook that flow'd
Fast by the Oracle of God; I thence
Invoke thy aid to my advent'rous Song. . . .

[Milton, PL, I. 10-13]

Admittedly, for confusion to occur the subordinate clause must be long enough and complex enough for the reader to forget that it ought to be subordinated to something, but it is just when such a clause is long that a writer using Early Modern punctuation is inclined to set it off with a semicolon—and that is the source of the modern reader’s difficulty.

The reader of Spenser and Milton, then, cannot assume that a clause is complete when he comes to a semicolon, since the material on one side of it may have to be integrated into the clause on the other side as an adverbial or nominal modifier. In setting off subordinate clauses, Spenser and Milton have a choice between a comma and a semicolon much like the choice moderns have between a comma or no comma: degree of heaviness can indicate degree of interpretive separation or cohesion. In their discussion of Spenser’s revisions of punctuation in the 1596 edition of The Faerie Queene, the editors of the Variorum Edition of Spenser’s Works cite certain lines where a semicolon emphasizes the discreteness of a clause, giving as an example the change of a comma preceding till in the 1590 edition to a semicolon in this passage:⁵

(32) [ancient Kings and Queenes], that had of yore
Their scepters stretcht from East to Westerne shore,
And all the world in their subiection held;
Till that infernall feend w th foule vprore
Forwasted all their land, and them expeld:
Whom to auenge, she had this Knight from far compeld.

[Spenser, FQ, I.i. 5]

Spenser (and Milton) frequently do use a semicolon before till, and they may well mean to indicate a pause signifying duration. In any case, we must not close off the clause preceding a semicolon (or colon) until we have looked across it to see if the next material may possibly be a subordinate clause, and this should be easy to do if the next material begins with a subordinator.

We must look across the semicolon for another reason, which is that what follows may be a coordinated verb phrase:

(33) Halfe furious vnto his foe he came,
Resolv'd in minde all suddenly to win,
Or soone to lose, before he once would lin;
And strooke at her with more then manly force

[Spenser, FQ, I.i. 24]

(34)   he praisd his diuelish arts,
That had such might ouer true meaning harts;
Yet rests not so, but other meanes doth make,
How he may worke vnto her further smarts:

[Spenser, FQ, I.ii. 9]

(35)O Prince, O Chief of many Throned Powers,
That led th'imbatteld Seraphim to Warr
Under thy conduct, and in dreadful deeds
Fearless, endanger'd Heav'ns perpetual King;
And put to proof his high Supremacy,
Whether upheld by strength, or Chance, or Fate....

[Milton, PL, I. 128-33]

(36) He trusted to have equal'd the most High,
If he oppos'd; and with ambitious aim
Against the Throne and Monarchy of God
Rais'd impious War in Heav'n and Battel proud
With vain attempt.

[Milton, PL, I. 40-44]

A simple tactic for recognizing the coordinate verb phrase suggests itself in the first three examples: it is signaled by the sequence

...; CONJ — V...

The placement of the prepositional phrase with ambitious aim ... before the verb in (36) may hinder this strategy, but such a look across the semicolon is sufficient to avoid being misled most of the time. Example (37) may also pose the same sort of delay of needed information (i.e., the verb) that (36) does (this continues [29]):

(37)                       so soft
And uncompounded is thir Essence pure,
Not ti'd or manacl'd with joynt or limb,
Nor founded on the brittle strength of bones,
Like cumbrous flesh; but in what shape they choose
Dilated or condens't, bright or obscure,
Can execute thir aerie purposes,
And works of love or enmity fulfil.

[Milton, PL, I. 424-31]

We must scan across a line and a half from the time we see; but until we get to the verbs can execute. Interestingly, although Spenser and Milton will sometimes place the Object before the verb in a coordinate verb phrase (more on this below), they never place a semicolon before such an inverted verb phrase, for a sequence like

...; CONJ — NP — V...

would be taken as beginning a new main clause by the strategy just outlined.

These passages give us further illustrations of the principle of compensation. After we recognize that we must not let a semicolon trigger clause closure and removal from immediate processing, we must develop some strategies that will give us a relatively quick indication of what to do with the material that follows the semicolon. Checking to see whether the material might attach back, or might itself constitute a clause, will help, as will noticing a subordinator which might link the clause across the semicolon. Further, anything such as length or distance of the verb that impedes this sizing up of the material will make the sentence hard to process. It is interesting that Spenser and Milton for once assist us, in that material looking like a Subject—verb sequence following a semicolon may be taken as such. Note that their avoidance of an O V sequence following a semicolon and conjunction is especially helpful just because the semicolon leads us to expect a new clause. We will see in the next section that they are not so careful about the same situation with commas.

3. COORDINATE CLAUSE BOUNDARIES

The comma is of course less conclusively a clause boundary marker, since it may occur, for example, between coordinate or appositive phrases. Suppose we have an apparently complete clause ending with a noun phrase, followed by a comma, possibly a conjunction, and another noun phrase: should the second noun phrase (italicized in the examples) be taken as the Subject of another clause coming up?

(38)At once as far as Angels kenn he views
The dismal Situation waste and wilde,
A Dungeon horrible, on all sides round
As one great furnace flam'd,...

[Milton, PL, I. 59-62]

In this example, the decisive information is the verb flam'd, which is spread quite far from its Subject, a Dungeon horrible. This phrase, furthermore, is so semantically similar to dismal Situation waste and wilde as to be a likely appositive to Situation. In other words, it is easy to read it as an appositive, and the information that we need to decide that it is not is withheld. Hence it is easily misread. In the next example, the line end may induce the reader to misperceive the Thunder as coordinate to fiery surge, since the decisive verb (which is boxed) is again delayed for over a line:

(39)           the Sulphurous Hail
Shot after us in storm, oreblown hath laid
The fiery Surge, that from the Precipice
Of Heav'n receiv'd us falling, and the Thunder,
Wing'd with red Lightning and impetuous rage,
Perhaps hath spent his shafts, and ceases now
To bellow through the vast and boundless Deep.

[Milton, PL, I. 171-77]

The same effect of uncertainty produced by a delayed verb occurs in the next two examples:

(40)the house, to his restless sense, was in the high homely style of an elder day, and the ancient Paris that he was looking for—sometimes intensely felt, sometimes more acutely missed— was in the immemorial polish of the wide waxed staircase...

[James, AMB, 145]

(41) all access was throng'd, the Gates
And Porches wide, but chief the spacious Hall
(Though like a cover'd field, where Champions bold
Wont ride in arm'd, and at the Soldans chair
Defi'd the best of Paynim chivalry
To mortal combat or career with Lance)
Thick swarm'd, both on the ground and in the air,
Brusht with the hiss of rustling wings.

[Milton, PL, I. 761-68]

In this last example we do apparently have an appositive (the Gates and Porches wide) to all access, and we are tempted by the line end and parenthesis to group the spacious Hall with them—Milton makes us wait four lines before we can see that this is wrong. Semantic clues here are not too helpful because Hall could, on the face of it, be grouped with accesses, though in fact here it refers to the principal chamber of Pandemonium. Note that there is parallel structuring here which leaps out if the parenthesis is removed: access... throng'd, Hall. . . swarm'd.

Empson cites two instances where a noun phrase hangs perceptually between two clauses, appearing to be appositive or coordinate to the Object of the first clause, and his examples seem almost contrived to force this misperception by interposing material between the noun phrase and its governing verb:

(42) Sometimes 'tis grateful to the rich, to try
A short vicissitude, and fit of Poverty:
A savoury dish, a homely treat,
Where all is plain, where all is neat,
Without the stately spacious Room,
The Persian Carpet, or the Tyrian Loom,
Clear up the cloudy foreheads of the Great.

[Dryden, in Empson, Ambiguity, p. 75]

(43)   Thou mak’st a Taper see
What the sunne never saw, and what the Arke
(Which was of Soules, and beasts, the cage, and park)
Did not containe, one bed containes, through thee,
Two Phoenixes, whose joyned breasts....

[Donne, in Empson, p. 51]

In the first, we read the colon as setting off an enumerating appositive ("fit of poverty, viz. a savoury dish, a homely treat") but four lines later find a verb (clear up) without a Subject and must reanalyze. In the second example, we tend to read what the Arke... did not containe as coordinate to what the sunne never saw as Objects of see, but when we finally get to one bed containes, we realize that the and in the second line really marks the beginning of a coordinate clause ("and through thee one bed contains what the Arke ... did not contain, two Phoenixes..."). The two phoenixes, that is, is not directly the Object of containes, as it at first appears to be, but is appositive to the Topicalized Object, what the Arke. . . did not contain. The effort involved in reading across the parenthesis to put what the Arke. . . did not containe together distracts one from the falsity of the initial hypothesis. Milton achieves this effect at the beginning of his Sonnet 18:

(44) Avenge, O Lord, thy slaughter'd Saints,
whose bones Lie scattered on the Alpine mountains cold,
Ev'n them who kept thy truth so pure of old
When all our Fathers worship't Stocks and Stones,
Forget not:

The ev'n them appears appositive to thy slaughtered Saints, and again we must wait until the fifth line for the Objectless transitive verb (forget...) to trigger reanalysis. To be sure, if we think about it, the recently slaughtered Protestants cannot be the same as the ancient saints, but the point is that perceptual mechanisms of the type we have been studying operate prior to such higher-level processing; hence it does not save us from misperception here. If we do not think about it, we will 'forget' exactly who the sonnet is about, which is obviously the effect Milton is seeking.

Two aspects of Milton’s syntax make commas more problematic as clause boundary markers than they are in other writers. First, Milton does not observe the constraint on split Subjects that, in Modern English, requires that if a second conjoined sentence is reduced down to just the Subject, the Subject must be inserted next to the Subject of the first sentence. Thus (a) can be reduced to (b) or (c), but (d), if reduced to (e), must then become (f):

a)  John has thrown rocks, and Mary will throw rocks.

b)  John has thrown rocks, and Mary will do so.

c)  John has thrown rocks, and Mary will.

d)  John cast stones, and Mary cast stones.

e)*John cast stones, and Mary.

f)  John and Mary cast stones.

Note that (e) would be more ambiguous in speech than in writing: the comma should discourage one from reading stones and Mary as coordinate. However, commas may be present for several reasons if a noun phrase is complex or lengthy, so that a supplementary semantically based strategy is probably useful. As discussed in the first section of Chapter One, we can check the noun phrase following the and for semantic similarity to the preceding Object noun phrase. Thus stones and Mary are fairly unlike (inanimate objects versus the name of a human), and so the temptation to group them as coordinate is negligible. In practice, one might engage in a comparative feature matching with the Object and Subject of the preceding clause. Consider for examples:

(45)   Or if Sion Hill
Delight thee more, and Siloa’s Brook that flow'd
Fast by the Oracle of God; I thence
Invoke thy aid to my adventrous Song. . . .

[Milton, PL, I. 10-13]

(46) Say first, for Heav'n hides nothing from thy view
Nor the deep Tract of Hell, say first....

[Milton, PL, I. 27-28]

Both examples can be misperceived by taking the trailing noun phrase (italicized) as the Subject of a new sentence (though this misperception is quickly recognized, since the search for a verb to go with it is quickly shut off—the semicolon does it in [45], the verb say instead of says in [46]). Additionally, (46) can be misperceived by grouping Nor the deep Tract of Hell with nothing as a split coordinate Object (if the reader reads nor as “not even”: “for Heaven hides nothing, not even the deep tract of Hell, from thy view”). Again, the feature-matching strategy will help, provided the right features are noted (i.e., in the second, Heaven/Hell, which, as opposites, are semantically closely related), and it will quickly rule out any attempt to group Siloa’s Brook... with thee. Matching semantic features can require a lot of attention to the meanings of the words, as the next example shows:

(47)  temperat vapors bland, which th'only sound
Of leaves and fuming rills, Aurora’s fan,
Lightly dispers'd, and the shrill Matin Song
Of Birds on every bough;

[Milton, PL, V. 5-8]

Here again the semicolon rules out taking the shrill Matin Song ... as the Subject of a new clause, and it can be matched back to sound as coordinate with it once we realize that the shrill Matin Song of birds is a sound. If one is relatively in touch with the text, the next example is not too difficult:

(48)Regions of sorrow, doleful shades, where peace
And rest can never dwell, hope never comes
That comes to all; but torture without end
Still urges, and a fiery Deluge, fed
With ever-burning Sulphur unconsum'd:

[Milton, PL, I. 65-69]

In Hell, the fiery Deluge is a torture, so it links up with torture as a split coordinate Subject of urges. (We may thank Milton, by the way, for the comma after Deluge—otherwise we would take fed as a main verb and look in vain for its Object.) Example (49) is our last illustration:

(49)   and that strife
Was not inglorious, though th' event was dire,
As this place testifies, and this dire change
Hateful to utter:

[Milton, PL, I. 624-27]

Here the second part of the split Subject (this dire change) is attracted to this place by repetition of this rather than by semantic similarity, though on reflection the place is the most obvious effect of this dire change.

To review briefly at this point, we may say that when we encounter a noun phrase following an apparently complete clause, we may look forward for a verb of which it could be the Subject while also considering whether it might belong back in the clause just processed. If for some reason the look forward does not give quick results, and if the noun phrase could be tucked back into the preceding clause, we tend to take this option. One reason that (43) is so difficult is that it is hard ever to find a place in the following clause for what the Arke did not containe. Notice that this uncertainty does not arise (45-47), where the look forward is cut off by punctuating or a non-agreeing verb: it is clear that the noun phrase must be a part of the preceding clause, unusual as the correct analysis as a split Subject is.

Matters are even a bit more complex in Milton (and Spenser) because of a second irregularity in syntax: they often invert verbs and their Objects in coordinate verb phrases:

(50) To winne him worship, and her grace to haue,
Which of all earthly things he most did craue;
And euer as he rode, his hart did earne
To proue his puissance in battell braue
Vpon his foe, and his new force to learne;

[Spenser, FQ, I.i. 3]

(51) Yet she her weary limbes would neuer rest,
But euery hill and dale, each wood and plaine
Did search, sore grieued in her gentle brest. ...

[Spenser, FQ, Lii. 8]

(52) 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. 638-43]

(53) Such place Eternal Justice had prepar'd
For those rebellious, here thir Prison ordain'd
In utter darkness, and thir portion set
As far remov'd from God and light of Heav'n
As from the Center thrice to th' utmost Pole.

[Milton, PL, I. 70-74]

Example (52) suggests what is difficult about such an inversion: the Object preceding the verb may be taken as its Subject ("his Regal State put forth . . ."). Presumably one of the first decisions a reader must make as he encounters a coordinating conjunction following an apparently completed verb phrase is whether the next material is conjoined as a clause or as a verb phrase, and the immediate occurrence of a verb is the cue that the Subject of the first sentence is to be carried over as the Subject of what follows. A noun phrase between the conjunction and the verb, however, will be identified as the Subject of the verb. We can schematize this as follows:

If:   NP V (NP), CONJ NP V...

and:S       (O)

then:                                  S

Placing the Object before the Verb baffles this strategy and would seem to invite misperception by setting up a garden path. There are again several tactics available to help the reader perceive particular cases correctly. Sometimes pronominal clues are present (i.e., the inverted Object is in the oblique or accusative case):

(54) Those two he tooke, and in a secret bed,
Couered with darknesse and misdeeming night,
Them both together laid, to ioy in vaine delight.

[Spenser, FQ, I.ii. 3]

Second, a running check on the rough semantic congruity of the noun phrase as the Subject of the verb should suffice to reject the misperception fairly quickly: in (53), for example, thir Prison can be quickly rejected as the Subject of ordained, and thir portion can be rejected as the Subject of set, on grounds of gross semantic incongruity. In (52), by contrast, Regal State and his strength just conceivably could refer to animate things suitable as the Subjects of put forth and conceal: semantic incongruity is not obvious enough here to reject the false analysis of the noun phrases as Subjects. (The situation here is roughly like that in examples [28-31] of Chapter One.) In (51), however, the semantic clues again work fairly well to reject every hill and dale, each wood and plaine as the Subject of search. Even if the semantic compatibility check fails to give a clear indication that the noun phrase is not the Subject of the verb, the perception of a transitive verb with a missing Object should trigger the re-analysis (i.e., when we get to haue, to learne in [50], or any of the verbs in [51-53]). Also, one might take advantage of the fondness of these writers for syntactic parallelism: Spenser and Milton often invert 'across the board', so that if the first clause is inverted, the second may also be (giving orders SOV + OVorOSV + OV). One might think of this tendency to parallel inversion as a warning to the reader, or as a warm up of the perceptual mechanisms needed to process the subsequent string. The pattern in (51), for example, is S O V + O V, in (52) it is S V A + O V + O V and in (53) it is O S V + O V + O V. Our writers also enjoy an occasional chiastic criss-cross, however, as in (50) (V O + O V; V O (A) (A) + O V). Note that (50) is not hard to process, since the Objects his puissance and his new force are so strongly similar that the parallel function is clear even though the order with respect to the verbs is reversed. In the next examples of chiastic order from Milton, the Objects again are linked together as contrasts:

(55) Consult how we may henceforth most offend
Our Enemy, our own loss how repair,
How overcome this dire Calamity....

[Milton, PL, I. 187-89]

(56) Henceforth his might we know, and know our own
So as not either to provoke,
or dread New Warr....

[Milton, PL, I. 643-45]

We should probably conclude that for Milton and Spenser the commitment to parallelism is not pronounced enough to base a perceptual strategy on, although it might provide confirming evidence for a certain analysis.

The next example illustrates the operation of these strategies in a more complex way:

(57) As when Alcides from Oechalia Crown'd
With conquest, felt th' envenom'd robe, and tore
Through pain up by the roots Thessalian Pines,
And Lichas from the top of Oeta threw
Into th' Euboic Sea.

[Milton, PL, I. 542-46]

Here the expectation of parallelism is violated (the structure is S V O + VAAO + OAVA). The only way to avoid misperceiving Lichas as the Subject of a new clause is to know the story as well as Milton does, since Lichas appears congruent enough as the Subject of threw, and we do not get to the transitive verb threw for several words. Again we brush up against Milton’s learning—knowing the story helps not only to perceive the aptness of the simile, but the simile itself. One reason (43) is so hard to perceive correctly is that Donne sets up a misleading parallelism what. . . what. . .—they are both Objects, but they are not coordinate.

The punctuation adds to the difficulty of (58), where some fairly close reasoning about the total meaning is needed:

(58)By falsities and lyes the greatest part
Of Mankind they corrupted to forsake
God thir Creator, and th' invisible
Glory of him that made them, to transform
Oft to the Image of a Brute, adorn'd
With gay Religions full of Pomp and Gold,
And Devils to adore for Deities:

[Milton, PL, I. 367-73]

Th'invisible glory is in a position where it could be coordinate with God, and the two noun phrases are semantically similar enough and congruent enough as the Object of forsake to support that grouping. There appears to be an Object missing after transform, however, though one may seek it in the following lines for some time before giving up the search as fruitless. The comma before to transform still stands in the way of viewing th'invisible glory as the Object of transform, and, as noted earlier, we tend not to expect the Object of an infinitival verb to precede the to. After some pondering, semantic clues seem decisive: the Image of a Brute suggests that th'invisible glory refers to God’s image in man—this he does not merely forsake, but actively perverts. Perhaps this can be viewed as a case of 'double syntax', yielding both senses: “forsake, and yet go further in their perversity.” In the next example, we do not even have an obviously missing Object to trigger reanalysis:

(59) But what if he our Conquerour...
...
Have left us this our spirit and strength intire
Strongly to suffer and support our pains,
That we may so suffice his vengeful ire,
Or do him mightier service as his thralls
By right of Warr, what e'er his business be
Here in the heart of Hell to work in Fire,
Or do his Errands in the gloomy Deep;

[Milton, PL, I. 143-52]

It takes some puzzling to identify whate'er his business be as the Object of work, partly because work is not obviously transitive (i.e., obviously missing its Object). In this elaborate set of (basically) coordinate purpose and result clauses, Beelzebub is trying to plumb the Divine Purpose, and if one does not get the Object matched to work, one misses the elegant twist that they will not merely work in fire in the heart of Hell, but work his business, whatever it be. It thus appears that the 'missing Object' tactic, which was the only clue in (41), is not always reliable—it fails for the same reason in the following:

(60)   our better part remains
To work in close design, by fraud or guile
What force effected not:

[Milton, PL, I. 645-47]

Again, work appears intransitive, and the delayed Object which appears in the next line is a slight surprise. Also, as noted at the beginning of the chapter, when we have a verb followed by optional Adverbials, we tend to close the clause after the Adverbial. Finally, of course, line division invites us to close the clause after guile. There is a conspiracy to make the reader misperceive here.

Just as look-ahead was a recurrent explanatory concept in the first chapter (and in this also), so there is a general principle of perceptual difficulty that emerges from these examples, viz., that a particular passage is not likely to be difficult unless it baffles several basic strategies at once. The notion of simultaneous operation of semantic strategies and those based on linear sequence implies that the failure of one set of strategies would not mean a breakdown in the process of comprehension as long as the other sorts of clues remained helpful. There is fairly good experimental evidence (mentioned in the Introduction) that structurally complex sentences are easier to perceive if there are semantic clues to the correct analysis, and many of our examples provide confirming instances. Indeed, our discussion suggests that the set of strategies operating simultaneously may well include clues of punctuation and parallelism, since these appear to make some passages easier or harder than others. A reader might of course take one type of strategy and ride it to the bitter end, but he would be a naive reader, and, in some of these passages at least, the end would be quick in coming.

An unexpected placement of a phrase or mark of punctuation, then, does not always result in misperception, and hence one cannot always assume a perceptual motive for every peculiarity of syntax or punctuation. Sometimes the motive can be found in the requirements of verse form⁶ or of information structure. The tendency we have noted of Spenser and Milton to place the Object of an infinitival verb before the to can give rise to numerous misperceptions, one of which takes the Object to be the Object of the preceding verb. This may occur in (19), where her strayed champion appears to be the Object of seeke. However, such a motive is not present in every case of this inversion, since it occurs fairly often at the beginning of sentences, and of course the Object is not always congruent as the Object of the preceding verb.

Two further general points about the perceptual difficulty of garden paths can now be made. Almost any mistaken analysis can be called a garden path, including the premature closures discussed in the first section, but there is a marked difference in the felt complexity of various mistakes. Mistakes like the ones discussed in the third section can lead to considerable reanalysis, involving either the removal of a noun phrase from one clause and placement of it in another, as in (58) and in (38), (41)-(44), or the reassignment of the function of a noun phrase, as in (52), (57), and (59), where initially perceived Subjects are reassigned as Objects with the Subject of the preceding clause then being carried over as Subject. Such reanalysis gives rise to a greater sense of effort than the premature closures with which we began the chapter. So, for example, (59) involves more reanalysis than (60) since the Object of work in (59) can initially be misperceived as appositive to service in the preceding clause. The second point is a delicate one bordering on paradox. The notion of a garden path involves the recognition that it is a false path. We noted in the previous chapter that the longer it is pursued before it is discovered, the more serious the perceptual strain is. In the light (darkness?) of the examples in this chapter, we can add that the more obvious the final dead end is, the easier it is to escape, for the 'trigger' for reanalysis is often the clue to the right analysis. This is above all the case with a strongly transitive verb plainly missing an Object—it is the weakly transitive verbs like work that make the greatest difficulties. The paradoxical aspect is simply that if the clue to trigger reanalysis is weak, and the initial analysis is not awful, the garden path may never be perceived as such—in short, one experiences no perceptual difficulty at all! Literary critics and teachers often find themselves arguing that a certain reading is better than one initially arrived at: here it is no longer a matter of right and wrong, since the structure of the text does not necessitate reanalysis.

The garden paths described above are one type of what William Empson called double syntax. Empson found value where Fowler found only “obvious folly,” though a particular instance does not qualify as a significant ambiguity of the second type unless the two readings, correct and erroneous, are “resolved into one” in an “ordinary good reading” (Ambiguity, p. 48). In Empson’s practice, such resolution is often only that the aberrant reading also functions expressively. He clearly wants the reader to respond to double syntax by entertaining all possibilities rather than by excluding the less grammatical reading, and to be willing to find expressive function in the struggle with the text rather than to reject the by-ways with annoyance. Empson does discriminate between better and worse readings, giving some priority, at times grudgingly, to the 'main, grammatical' reading. He is right to hold fast to this principle, at least for the authors he considers, for one crucial reason—there is always a grammatical reading. A rather different attitude might be warranted for texts where a grammatical parsing is not always possible. If, for example, one had overlapping syntax where a single noun phrase played different functions in two clauses, a decision that it should go with one or the other would be misguided, since it would make the one clause grammatical at the expense of the other. Similarly, if one consistently had to choose between two somewhat ungrammatical parsings, one would gradually reduce the degree of priority given to the 'more grammatical' reading. Finally, if it frequently made little or no difference which clause an element went with, the reader might also begin to pay less attention to clause-boundaries as the basic unit of processing the text. The 'hovering' elements in so many of Empson’s examples, which can go with the preceding or following clause, do not trigger reanalysis—either reading is right—and an acute awareness of sentence structure will only weary the reader with inconclusive comparisons of alternative parsings. All three of these conditions are considerably more prevalent in Spenser’s works than in Milton’s or in the later writers' and suggest a line of approach to Paul Alpers' claim that Spenser’s syntax is 'permissive' and that attention to sentence structure never helps in reading The Faerie Queene. I will take up this matter in Chapter Five. If the great majority of a writer’s sentences can be parsed into grammatical structures, however, the reader can aim for such a parsing and assume that his comprehension is correct if the sentence is grammatical. Whatever misperceptions he has made on the road may be interesting, valuable, and expressive, but they are, finally, misperceptions.