University of Virginia Library

The Suppressed and Altered Passages in Blake's Jerusalem
by
David V. Erdman

The following investigation began as an attempt to restore the textually significant deletions in William Blake's Jerusalem, notably the conspicuous deletions on Plate 3. It has gradually extended, in the rather surprising absence of any thorough collation of all copies of the work, into a critique of every discernible or alleged deletion or alteration that might be of value in determining the chronology of copies or the order of plates — or that might contribute significantly to our still scanty knowledge of those aspects of Blake's etching process which limited his control over his text. More or less valuable restorations of text and more or less technical analyses of mended plates fall side by side in the following plate-by-plate report, but this seems the most useful arrangement.[1]

Appended are a corrected List of Copies and some notes on Catchwords, on Plate Sizes, on Blake's Numbering of the Plates,[2] and on the Dating of Blake's Script.

Before Blake published his maturest Prophetic Book in "Illuminated Printing," Jerusalem: The Emanation of The Giant Albion, at least before he printed any of the copies now available for inspection (none earlier than 1818-1820 according to dated watermarks), he had erased from the copper plates from which he printed them a number of words and whole lines, most conspicuously in the prose and verse of his prefatory third plate addressed "To the Public". His method of erasure was to run a sharp tool across the raised surfaces of the copper, the printing surfaces of his relief etching. A vigorous gouging could level these surfaces beyond recovery, as it did on Plate 84. But in many instances Blake did not carry his negating beyond a few strokes, leaving a stubble of metal that would print broken outlines of letters and ghosts of words when the plates were inked and pressed.

On Plate 3 most of the single-word deletions remain fairly legible, and Blake's modern editor has been able to indicate with italics his nearly certain reconstructions of such passages as: "Therefore, dear Reader, forgive what you do not approve, & love me for this energetic exertion of my talent."[3] Yet a large erasure of nearly four lines in the center of the first paragraph has remained undeciphered, and the absolute blanks that show up in the excellent facsimiles of the Rinder and Mellon copies recently published by the William Blake Trust confirm the impression that this large deletion must forever defy reconstruction. Whether as Blake's "Public" we think of ourselves as "Sheep" or "goats" — words he added to Plate 3 with the same tool, perhaps with the same impulse, that deleted his expressions of love and friend-ship — it would seem we must resign ourselves to a less than perfect text.

Looking at the Morgan Library Jerusalem a few years ago, however, I noticed that there were no complete lacunae on this plate but visible crumbs of letters in every part of the erasure, from which a full restoration was, theoretically, possible. The curators allowed me to examine the page under their low-power microscope and under ultraviolet radiation. Photostatic and photographic copies of this and other pages, positive and negative, on high contrast and low contrast papers, enlarged by two and by three and four diameters, were "bruized and knocked about without mercy, to try all experiments," as Blake said of some of his "experiment Pictures." Through the owner's courtesy I was able to borrow the Rosenbloom copy for several weeks, collating it directly with the Morgan copy and keeping it near at hand for

photographic comparison with other copies. This posthumously printed copy, heavily inked and firmly impressed, is clearer in some plates than the other copies and, like the other posthumous copies, lacks the obscuring vines and washes that Blake applied to many of the deletions in the copies he printed himself.

The upshot is the following report, which for a time I hoped to be able to entitle "Jerusalem Restored." The major passage on Plate 3 has yielded to ocular and photographic attack, and lesser but significant and sometimes astonishing passages have been recovered on other pages. There remain, however, several places where I have been unable to go beyond conjecture, some where even conjecture must be silent.

Method of Restoration

The technique that finally yielded results was a fairly simple one, though infinitely time-consuming. Fortunately it is scientific in the sense that it can be tested and repeated, and persons who have followed a demonstration of the processes involved have been able to shed the doubts that are naturally inspired by a claim to have deciphered the supposedly indecipherable.

The method used relies on the demonstrable fact that Blake's script, on the pages in question, simulates printing so precisely that the shapes of letters, the shapes of words, and the spacing between particular letters or strokes on any given page are extremely uniform. Photostatic enlargement brings the words or fragments of words up to a size convenient to use under a sheet of transparent acetate (the kind called "workable") upon which a tracing can be made with modern india ink.[4]

Jerusalem 3 is reproduced below (Illustration I) in a slight reduction, but even in this size the restoration of some readings can be demonstrated. To test, for example, the hypothesis that "forgive" is the correct reading of the deleted word in the center of line 19 (two lines above the rhymed stanza) a tracing was made from the word "forgiveness" five lines above it and superimposed upon the deletion. Confirmation was immediate in this case, because the crumbs of "forgive" are sufficient for each letter, including the dot of the i and the ascender of the f, though not the descender of the g.

If no counterpart of the hypothetical "forgive" had been present on the page, a model could have been constructed from "for" in the next line, the "rg" from "energetic" in the same line, the spacing of "gi" from the "gh" in "righteous" (line 15), the "ive" from "live" (line 30). The labor consists, essentially, in extending this process by applying an ink-traced hypothetical word to each group of dots in the deleted area that appears to be a discrete word, until the word attempted fits the dots. Good guesses at this stage are of utmost importance. Actually the single words that have long since been recovered are not in themselves easier to read than most of the words in the large "impossible" area of lines 7-10. The difficulty in reading this larger deletion arises chiefly from the lack of good contextual clues to supply hypothetical readings to be tested against the fragments. But the difficulty has been enhanced, for Plate 3, by some wrong guesses embedded in the Nonesuch Blake — the editor having given the word "engraving" where he should have read "arroganc[e]" and "accurs'd" where the right reading is "absorb'd" and "ideas" where it is "Gods". These were fairly close guesses, for shape and length, though the g plainly visible in the center of the first word should have ruled out the first guess. Enlargement brought out the three ascenders of "absorb'd" and ruled out the second guess.

The precision of the present method makes possible a confident restoration of the deletions in the first rhymed couplet on this page (here given in italics):

but two famous guesses would long since have been dismissed as untenable if anyone had thought to measure them against the space available. (See Illustration II a.) Swinburne offered the reading "things" for the word in the second line now reconstructed as "books", and Professor Damon, looking more closely at the crumbs visible in the Morgan copy but not thinking of measurement, proposed "hymns".[5] The other words in the couplet are fairly legible. The s and the ascenders of b and k in "books" are there to be made out; but the ovals of "boo" are cut up beyond recognition. Only when "books" is traced from the line above and superimposed on these remnants does the exact fit demonstrate this to be the correct reading. Simple measurement of Swinburne's guess, however, or a tracing made of "thing" (to be found alone and in "nothing" just below) with an added s, would show that as Blake wrote "things" it would be a whole letter too long for this space; and a Blakean "hymns" would be almost exactly as long as the word "heaven" in the line just above. (In Illustration II a, "books" and a composite "hymns" traced from parts of nearby words have been placed below the deletion to show their respective fit and lack of fit.) Here context was perhaps too helpful; it will support any of the three readings somewhat, though once the right reading has been found it can be seen to fit most snugly of all, not merely into the broken outline on the plate but into the meaning of the whole passage: see the ensuing references to "Sinais awful cave", "the wondrous art of writing", and "my types". (Incidentally we may now abandon the suspicion that Swinburne had access to a more legible copy of this page, now lost.)

With so many crumbs of letters remaining in this portion of Plate 3, although the labor of arriving at correct solutions was long, the testing and retesting of these solutions took very little time. But a further complication should be mentioned. At times Blake laced his text with interlinear drawings, such as the small flourishes in the last paragraph on this plate: "**But * I *soon found . . . syllables. ***Every *word . . ." (to represent each letter-width of drawing by an asterisk). Very early in my scrutiny of the erased passage at the top of the plate, I suspected that a little trefoil of dots in the second line (between "entrusted" and "their love") and a sort of loop-and-a-half of serpent in the third line (between "I have" and "who Acknowledge") might be remnants not of words but of ornamental drawings. I tried stubbornly to think of words that could fit these dots, but as my deciphering of words gradually closed in on these spots, it became impossible to hold out for their representing tiny words. The snake would have had to be a two-letter word, and it was not any of those I could think of.

The same or similar methods,[6] with the varying degrees of success reported below, were applied to deletions on Plates 1, 2, 4, 15, 35, 36, 47, 73, 77, 82, 91, and 96. Deletions too thorough to be recoverable by any means are noted on Plates 24, 83, and (possibly) 93. Also dealt

with are various mendings, deletions, and additions, most of them involving no difficulties of reading, on Plates 7, 8, 11, 22, 24, 28, 29, 37, 45, 51, and 61.

Beneath the large picture on Plate 95 were discovered traces of the text now on Plate 94 and some pictorial evidence that the matter begun on a single plate had been extended to occupy two plates. Plates 96 and 99 were found to contain similar palimpsest traces of earlier use — some engraved writing on 96 and extensive pictorial matter on both plates.

The Nature of Blake's Alterations

We know from Blake's Notebook and other manuscripts that he was an extensive reviser and rearranger of his verses before he considered them ready for publication. But once he had made a fair copy in acid-resistant varnish or ground, to be applied in reverse to the copper plate, he was able to make only minor changes.[7] Some examples of mending in the varnish are the redrawings of lower-case letters as capitals on Plates 10, 21, 24, 61, 83, and 85. The correction of "furnaes" to "furnaces" and of "garents" to "garments" on Plates 53 and 61 must have been made after transfer to copper and also after some etching had been done: the inserted letters and carets are very faint when Blake forgets to retouch them after printing; if they had been added at once, they should have been as dark as the original lettering.

Before etching, or after only shallow etching, Blake could easily remove whole lines and replace them with birds or flames or tendrils drawn in varnish directly on the copper. We can hardly recover such deletions, but we can suspect their having existed, because Blake's normal paragraph spacing does not leave the exact height of a line as do some of the bird-filled hiatuses, and because there is usually some slight indention of the first line of a genuine paragraph. By the same token, we are able to detect, from signs of crowding, the late insertion

of a line within a stanza break (see Plates 39 and 53). But such insertions could only be made before the acid had been applied for any length of time.

When Blake called his etched or etched and engraved plates "types" or "stereotypes" he was expressing the fact that once finished his copper plates were monolithic. The illustrations could be modified by further etching (see Plate 95) or engraving, but the text was almost unalterable. Once in a while, when there was room, he could add a line above or below the block of text, by engraving. And we have seen that he could scratch out words or lines whenever he felt the need to. But it was almost impossible for him to put new words in the place of old. The complicated mending process involved in transforming one fourletter adjective into another on Plate 37 is a formidable illustration of the limits imposed on the blacksmith-poet by the nature of his medium.

His Probable Reasons

Blake's reasons for making those deletions in Jerusalem which can be recovered for discussion range from the obvious to the complex to the inscrutable. Changes of plan, aesthetic and moral considerations, changed thinking, increased secretiveness, the effects of depression irreversible whether or not temporary — all appear to have operated, singly and in combination, to produce the deletions and alterations noted below as well as numerous rearrangements and replacements of whole plates, about which the evidence is highly confusing in our present state of knowledge about such things as Blake's graphics.

An incalculably drastic change of plan is hinted at by the deleted phrase "In XXVIII Chapters" etched in the title page (Plate 2) of a work which in its extant form is geometrically arranged in a Fourfold scheme of four chapters.[7a] The deleted "End of the 1st Chap:" on Plate 14 does not tell us how long the chapters were to be, in any earlier scheme, because we cannot tell where this plate stood in any earlier sequence. Nor can we easily tell what to make of the date "1804" on the title page, undeleted, although it is well to recognize that the date is incised, not etched, and can have been added to the plate at any time (both in the sense that the plate may have been old or new when it was dated and in the sense that "1804" can have been added in that or, commemoratively, any subsequent year.)[8]

Other deletions related to changes of plan occur on Plates 1, 29, 35 (perhaps), 37, 41, 91, and 94-95.

Aesthetic as well as schematic considerations may have influenced

the decision to delete all text from Plate 1, to make it a silent and sombre frontispiece to the series continuing in 26, 51, 76, 100. That Blake may not have been dissatisfied with the text of Plate 1 in itself, however, seems indicated by the fact that he made a separate "proof" in color and outlined the letters carefully in dark ink. Moral and aesthetic considerations may have combined to cause the elaborate re-engraving required to uncouple the male and female figures on Plate 28 — though here too it may be significant that a print of the first state has survived. On Plate 37 the mending of "pale . . . feet" to "blue . . . feet" may be doctrinal as well as aesthetic. Rhetorical improvements were made on Plate 4, with deletion of the double exclamation mark, and on Plate 82, with the reduction of duplication in lines 47-48 and the removal of the "redning skeleton in howling woe" from line 69. The amplification of one plate to two near the end of the work (Plates 94 and 95) may be due simply to recalculation as the 100th plate approached — or it may have involved rejection of another plate, for reasons we cannot discover, though possibly from the overriding desire to increase the "illumination" on 94-95.

Changes of idea may have influenced the deletions on Plates 1 (perhaps), 4, 7, 35, 36, 47, 73, 77, 84. But the self-destructive deletions of Plate 3, withdrawing the affectionate terms addressed to the oncedear Reader, effacing and yet not quite thoroughly effacing the poet's confessions of faith and enthusiasm: these are of a different order. Blake had done something of this kind in a less irreversible way when he had added a stanza of harp-shattering despair to the Preludium of his America: at the bottom of a page, so that when his depression lifted he could cover the stanza and print the page free of gloom. The momentary effects of dismay upon the Preface of Jerusalem remain even in the most brightly colored copy, remained perhaps in his own will — for he could, after all, have made a new plate.

JERUSALEM: All the Deleted or Altered Passages

Here are given the texts of all deleted passages in Blake's Jerusalem, with restoration where possible; also all textual additions or emendations made during or after etching or after printing. In addition the list includes deletions and revisions of pictorial matter which are (or which have been thought to be) the result of changes of state, that is, changes made in the copper plate. I have not attempted a general collation of variations in the coloring or retouching of pictorial elements in the different copies, nor of the retouching of the lettering when no alteration or deletion is involved.

(Deletions or changes of catchwords and of plate numbers are dealt with in the separate addenda on these matters.)

Angle brackets < > enclose deleted letters, words, or passages; brackets within brackets signify small deletions within larger ones. Words printed within such brackets, and unqueried, constitute certain restorations.

Square brackets [ ] enclose editorially supplied letters or somewhat doubtful words (always preceded by question marks) or illegible deletions (indicated by blank space). Where the deleted text is blank or quite illegible, no conjectural words have been supplied, though possibilities are sometimes presented in the discussion following the text.

Each asterisk (*) represents a segment of ornamental drawing the width of an m or less. In the deleted passages it represents fragmentary remains of what I take to have been drawing rather than text.

All complete copies of Jerusalem known to exist have been examined (though the Rinder copy only in facsimile) as well as the few separate proofs or prints of single plates, as cited. Particular copies are cited (see appended list) only when there are significant variations among them.

Plate 1 [Frontispiece] (Illustration III a)

• [Top of page and dividing over archway:]
• <There is a Void outside of Existence, which if enterd into Englobes itself & becomes a Womb, such was Albions Couch A pleasant Shadow of Repose calld Albions lovely Land

•  

  His Sublime & Pathos become His Reason his Spectrous Jerusalem his Emanation <O A[ ] * [?behold]

  Two Rocks fixd in the Earth Power. covers them above is a Stone laying beneath [?Pitying]> behold the [7] Vision of Albion>

• [On right side of archway:]
• <Half Friendship * is the bitterest Enmity said Los
  As he enterd the Door of Death for Albions sake Inspired
  The long sufferings of God are not for ever there is a Judgment>
• [On left side of archway, in reversed writing:]
• <Every Thing
  has its
  Vermin O Spectre of the Sleeping Dead!>

All the writing on this page, designed not as a part of the text of the poem but as inscriptions on a large illustration, was applied not in etched masses but in line engraving which was so thin that Blake could easily delete it if he wished by an over-all inking of the plate. He did so in all copies of Jerusalem, but Sir Geoffrey Keynes has been able to recover the text (except for the erasures in line 7) from a proof copy in which the letters have been made visible by outlining "in black with a fine pen"

against the light brown printing of the plate.[9] (Text as punctuated above derives from this copy.)

Even in that proof, line 7 contains an erasure made by a gouging of the plate that left only the tops and bottoms of what I conjecture to be the b, h, l, d of "behold" and the y and g of "Pitying" (but the whole P is visible). These elements are fairly discernible in enlargements of the Rosenbloom copy and may be faintly seen in the brown and black collotype of Sir Geoffrey's proof (published with the Rinder facsimile).

From the posthumous copies it is apparent that Blake considered the whole text canceled, however, and not to be brought back by careful inking; at least all copies are from the same state of the plate, slightly but significantly modified from that used for the proof. Except for the early erasure in line 7 (perhaps a correction before canceling) Blake did not scratch out any words, but he incised lines of texture for the stone of the arch and lines marking the bricks of the wall — right across his lettering. Inking could do the rest.

Joseph Wicksteed suggested that the obliteration of the text from Plate 1 "was done intentionally to fit the design to function as a Frontispiece, which in Blake's Prophetic Books . . . never have any text"[10] — a rather circular argument, presenting Blake as a self-conformist who caught himself in an inconsistency. A more important consistency is that which he achieved by the tone and placing and silence of Plates 1, 26, 51, 76, and 100 — with the minor inconsistency that the lettering that remains on Plate 26 had been too broadly chiseled to be inked over. The names of "Albion" and "Jesus" on 76 could be hidden easily, and usually were. Was the original plan, after four chapters were decided upon, to have a spaced series of inscribed frontispieces? Did the idea of opacity only gradually gain ground, too late for 26? (See also the comment on Plates 51 and 76, below.)

Clues to possible reasons for deletion may be found in the content of the inscriptions. The text condemning "Half Friendship" implies a plea for whole friendship and may have been jettisoned along with the words of "love and friendship" on Plate 3. The lines on the Englobing Womb are a variant of similar lines in Blake's Milton which are among the few deleted passages in that work.[11] And a drawing of the phallic altar stones

alluded to here is deleted in Jerusalem 69, though only in the Rinder copy. (Noted by Wicksteed, p. 216, n.3.)

The lines in reversed writing were, of course, secretive from the beginning: a direct rebuke to any verminous fiend who deciphered them in his mirror.

Plate 2 [Title Page] (Illustration II b)

Jerusalem / The / Emanation of / The Giant / Albion / <In / XXVIII/ Chapters> / 1804 / Printed by W. Blake Sth Molton St.

The title was etched, the deletion subsequently made by vigorous scratching but not gouging; the date and imprint were added by engraving.

The deleted phrase, directly below the "on" of "Albion", can be made out even from the Rinder facsimile now that its location and shape are precisely known. I was able to decipher it with certainty, however, only from an enlarged photostat of the Rosenbloom copy. That Blake intended the four-chaptered Jerusalem as a work in 28 Chapters, even as late as the etching of his elaborate title page, comes as something of a shock, though it may be recalled that his Milton, in Two Books, is announced on its title page as "in 12 Books" — another of the monumental plans of 1804.[12]

Plate 3 [Preface] (Illustration I)

Sheep goats

To the Public

After my three years slumber on the banks of the Ocean, I again display my Giant forms to the Public: My former Giants & Fairies having reciev'd the highest reward possible: the <love> and
<friendship> [5]
of those with whom to be connected, is to be <blessed>: I cannot doubt that this more consolidated & extended Work, will be as kindly recieved **************** The Enthusiasm of the following Poem, the Author hopes <*no** Rea*der will think presumptuousness or
arroganc[e] when he
is reminded that the Ancients entrusted*** their love to their
Writing, to [10]
the full as Enthusiastically as I have ** who Acknowledge mine for my Saviour and Lord, for they were wholly absorb'd in their Gods.> I also hope the Reader will be with me, wholly One in Jesus our Lord, who

is the God <of * Fire> and Lord <of * Love *> to whom the Ancients
look'd
and saw his day afar off with trembling & amazement. [15]

The Spirit of Jesus is continual forgiveness of Sin: he who waits to be righteous before he enters into the Saviours kingdom, the Divine Body; will never enter there. I am perhaps the most sinful of men! I pretend not to holiness! yet I pretend to love, to see, to converse with daily, as man with man, & the more to have an interest in the Friend [20]
of Sinners. Therefore <Dear> Reader, <forgive> what you do not
approve, &
<love> me for this energetic exertion of my talent.

We who dwell on Earth can do nothing of ourselves, every thing is conducted by Spirits, no less than Digestion or Sleep. <to[ ]> ** <the best words of Jesus, [ ]> ****When this Verse was first dictated to me I consider'd a Monotonous Cadence like that used by Milton & Shakspeare & all writers of *English *Blank Verse, derived*** [40] from the modern bondage of Rhyming; to be a necessary and indispensible part of Verse. . . .

The "Sheep" and "goats" at the top, presumably divisions of "the Public", are words not etched but engraved. They may have been added when Blake was scratching out the terms of love addressed enthusiastically to the Public in his first paragraphs.[13]

All copies seem to have been printed from the same state of the plate, the relatively greater legibility of the deletions in the Morgan copy and (at

some points) in posthumous copies being the result merely of differences in inking. The single deletions in lines 5, 6, 14, 21, 22, 23 can be read in most copies; they are correctly transcribed in Keynes except for the capital D in "Dear Reader" (line 21). In the Harvard copy the large deletion of lines 9-12 is partly covered by drawings of vines, as also the deletion in line 24, but the same impression marks are discernible beneath. So are they in the Mellon copy, beneath washes and lines of color. Some of the other deletions in the Morgan copy are hidden with vines, but not these; for whatever reason the plate seems to have been inked and impressed with particular care.

Details of restoration are given in my introductory remarks. Only the final e of "arrogance" leaves no trace in any copy. The deletions at the end of line 36 and in the second half of line 37 remain baffling, although crumbs of some specific letters are visible and we are obviously faced with deleted text, not ornament. In line 36 the phrase may be "to Hear" or "to Note". In 37 the lower edge of a g seems attached to the first d of "dictated" in the next line; another g or a y is evident above the word "me". Certain ascending strokes are also to be made out.

It should be observed that the period given after "Sleep" in line 36 might be a comma. Blake used a dot-like mark that must be read sometimes as comma, sometimes as period.

Plate 4 ("Chap: 1")

. . . .
Beaming forth with her daughters into the Divine bosom.
<Where ! !> [15]
Where hast thou hidden thy Emanation lovely Jerusalem
. . . .
But the perturbed Man away turns down the valleys dark;
<Saying, We are not One: we are Many, thou most simulative> [23]
Phantom of the over heated brain! shadow of immortality!
. . . .

Editors have observed the deleted "Where" in line 15 but have failed to note the exclamation marks: these are deleted not by scratching but by treatment as part of the marginal network (and by blue wash in the Mellon copy). Line 23, read from enlargements of the Morgan and Rosenbloom copies, is almost invisible beneath vines or washes in the others.

With removal from line 15 of the redundant and exclamatory "Where ! !" the text is considerably reduced in histrionics (there are still seven exclamation marks in the first ten lines of the page). But the deletion also removes, perhaps inadvertently, the required seventh foot of the line.

The deletion of line 23 is more seriously disruptive, since it removes an indicator of the dialogue ("Saying") and half the epithet meant for the

one addressed ("simulative Phantom"), and because it obscures an important element in the poem's definition. This plate opens Chapter 1, and readers would have been helped by so explicit a definition of perturbed Man's error as the now hidden assertion, "We are not One: we are Many."[14]

Plate 7

. . . .
O holy Generation <Image> of regeneration! [65]
. . . .

The word "Image" was scratched through for deletion, but not replaced; it is illegible or nearly so in every copy except the Cunliffe and the Harvard, in both of which it has been retouched into legibility. Can this deletion have signified some temporary doctrinal change — or does Blake's failure to repair it in other copies represent negligence?

Plate 10

In line 47 the word "Righteous" is mended, before etching, from "righteous".

Plate 11

Between the text and the bottom illustration there is a shadow that might be the top of a deleted line, but careful measurement dispels the idea.

Plate 14

<End of the 1st Chap:>

This colophon, in catchword position, is severely deleted; clearly visible only in the Rosenbloom copy.[15] (The plate number, probably 14, is not clearly visible in any copy, but the text is tightly linked with Plate 13, which is visibly numbered "13".)

Plate 19

According to the Census the small recumbent figure below the large recumbent figure at the bottom of the plate is absent in the Cunliffe copy. Its absence is evidently a matter of coloring and probably does not indicate a different state of the plate. (Not personally examined.)

Plate 20

This plate is one of the few that occurs in two states. In the first state,

found in a separate leaf in the Rosenwald collection (with undeleted "16" incised in top right corner) and in the British Museum and Rinder copies, there are three wave-like points of flame or gown in the top right corner, the central point about 1 cm. from the corner of the plate. In the second state, alike in every other particular, the central tip has been extended to within 3 mm. of the corner and accompanied by a thin line on each side of it. This extension cuts into the numeral "16" but can hardly have been made simply to delete the number. This state appears in the Harvard, Mellon, Morgan, and posthumous copies, though in the Mellon a further change is made with black wash over the waves, practically eliminating them. (In the Cunliffe copy no number is visible, and the coloring of the illustration makes it difficult to determine the state of the plate.)

The variant does establish a chronological separation between the first two complete copies and the later ones.

Plate 21

In line 37 the C of "Childrens" is mended from a lower-case c, before etching.

Plate 22

In line 10 the first letter of "Hatred" is mended, somewhat inadequately. It was first written as an R, probably in anticipation of the word "Right" in the next line.

In line 13, "coutenance" is mended to "countenance" before etching.

Plate 24

. . . .
Dost thou forgive me? thou who wast Dead & art Alive? [58]
. . . .
I die! I die in thy arms tho Hope is banishd from me. [60]
<[ ]>
Thundring the Veil rushes from his hand Vegetating Knot
by Knot. . . .

The D and A of "Dead" and "Alive" in line 58 were mended from lower-case letters, before etching. After line 60 a full line was deleted, perhaps after shallow etching, for a few faint shadows of parts of letters remain. A line related to line 60, including "Hope is banish'd from me" (Albion's last words), is deleted and varied on Plate 47 (which see).

Plate 26

Stigmata, or rather the raised heads of spikes, are visible in the palms of "Hand" in all copies but the British Museum and the Mellon, where they are hidden by drawing but not truly deleted. In the Mellon copy bright spots of red are added to Hand's ankles. (Compare note on Plate 35.)

Plate 27

In line 15 "Willans" is not very legible except in the Mellon copy, where it has been retouched. The word may have been corrected on the copper from "Williams".

Plate 28 ("Chap: 2.") (Illustrations IV a, b)

This plate is described in such a way in the Keynes-Wolf Census (p. 106) as to suggest that there were four states of the copper. I find only two states, one represented by the proof on 1802 paper which is tipped into the Morgan copy, the other represented by the adjacent page in the Morgan copy itself and by all other copies including the posthumous ones, in which the exact condition of the copper plate may be seen. The "states" noted in the Census are variants of coloring and inking; they establish no sequence of issue.

Unhappily the Census descriptions of each "state" are erroneous in major details. A correction of these is of importance both bibliographically — this being the only plate in Jerusalem thought to contain evidence establishing priorities among more than two groups of copies — and, in a sense, semantically or textually: for although the actual text is unchanged, the picture which fills the top half of the plate and announces a new chapter is radically changed in its second state. Most of the implications of this change do not fall within the present discussion, but one of them is potentially chronological. To put it simply, the primary change Blake made in this plate, picturing female and male figures embracing in the center of a "lilly of Havilah", involved turning the legs of the two nude figures from a position in which they can be assumed to be copulating to one in which they cannot. The change was made by extensive additional engraving of the etched and engraved plate. The secondary changes, both in the engraving and in the retouching in the British Museum, Mellon, and Morgan copies, are largely consequent upon it.

Joseph Wicksteed in his Commentary (pp. 159-160 and 204) is extravagantly impressionistic in his interpretation and in his details but appears to have grasped the primary change correctly. The Census, while reporting the degrees of clearness with which certain (to me invisible) "genitals" appear in various "states" of the plate, assigns to the male and female figures a position in the first state which would be as essentially chaste as their position in the later.

Confusion of left and right may cause some of the difficulty, and confusion of references to the "woman" (presumably the smooth-faced figure on the left) and the "man" (the short-haired figure on the right, on whom Wicksteed sees a beard, and whose back muscles are knotted in the first state). In the 1802 proof, according to the compilers of the Census, "the woman's right thigh [is] visible parallel to and underneath her left; the man's left leg encircles her body; the moulding of the man's back is clear; a caterpillar is below the figures on the leaf."

Correction: Only in the second state does the woman have two visible thighs, parallel. In the first state only her right thigh is visible — not underneath but above and parallel to the left thigh of the man. His leg does not encircle her but is bent back at the knee so that his foot is, quite visibly, beneath his left buttock. It is his right arm that encircles her waist, for we see what must be his right hand resting on her right hip. No right leg for the man or left leg for the woman is in the picture. The caterpillar, on the petal ("leaf") below the man, is not attached to him, though grossly phallic if you will.

Census: "In the second state . . . the woman's right thigh and the man's left leg have been removed; only ineffective cross-hatching remains on the man's back; the caterpillar has been removed; the man's genitals are defined."

Correction: In the second (and final) state of the plate the woman has two visible thighs, for the first time, a left thigh having been created for her out of what was the man's left thigh, so that she now appears to sit side-saddle on the petal. The man's left leg below the knee has been removed by an upward spread of the petals he rests upon; he has been given a right thigh, and his buttocks have been (ineffectively) moved round to the left, so that he is now sitting side-saddle and turned the other way from the woman. His back muscles are flattened (and his spine surely broken). The caterpillar has been (almost) removed by additional engraving of all the petals. In neither state can any genitals be seen on either figure; curiously enough even the compilers of the Census see that the man has his back to the viewer in either state.

They cite "copy D" (the Harvard) for this "second state" (skipping the Rinder) and find a "third state . . . in copy F" (the Morgan), with "the genitals less clearly shown; the leaf below Chap. 2 . . . lightened, and the white lines increased on the lower part of the leaf on which the woman's leg extends." Laborious comparison of photographic copies with each other and with the originals in turn has failed to disclose the slightest difference in any part of the picture except for ink lines added in a few places in the British Museum and the Morgan copies and more elaborate drawing in the Mellon copy. The posthumous copies make it clear that no lines were added to the plate or subtracted from it. (Patchy inking in the Harvard copy does give a vaguely phallic appearance to a portion of the female's right leg, but comparison with other copies makes evident the accidental nature of this effect.)

Keynes and Wolf discern, nevertheless, a "final state" in the colored (Mellon) copy "and all the posthumous ones": "all indications of the genitals removed; the man's right hand . . . eliminated." Here theory supplants observation, for the right hand is painted over in the Mellon copy — but reappears in the posthumous plates.

Wicksteed's Commentary makes out a tremendous difference between the Rinder and Stirling (Mellon) copies, defining the male and female as "almost completely mingled" in the first state and as "distinct individuals,

whose bond is expressed by the embrace of their arms and not by amalgamation" in the later: a distinction implying that intercourse destroys individuality. Graphically, from the waist up, the two figures are just as much "distinct individuals" in one state as in another, even if we include "states" made by supplementary coloring.[16]

As for the bibliographical deduction that seems possible the drastic change implies, I should think, a revised attitude toward copulation (or increased sensitivity to moral disapproval from dear Readers) which in turn may imply the passage of years rather than days. Though admitting of more than one explanation or interpretation — as does the early date, "Edmead & Pine 1802", of the paper of the proofs of Plates 28, 45, and 56 — the point is worth bearing in mind. Some day enough biographical, stylistic, and graphic evidence may accumulate to make it useful.

Plate 29[43][17] (Illustration II c)

. . . .
All is confusion: all is tumult, & we alone are escaped <!>[82]
So spoke the fugitives; they joind the Divine Family. <?Albion
?slept]> [mended to] trembling

A laborious mending at the end of line 83, the last line on Plate 29, replaced an earlier word or words with "trembling" and cut into the punctuation mark in line 82, a colon or exclamation point. My reconstruction of the earlier reading, "Albion slept", fits the space but remains somewhat conjectural.[18] The "tr" of "trembling" was very crudely hacked, the "bl" given imperfect ascenders, and the "g" not effectively built up at all. Blake always had to retouch the b, l, and g to make them clear. The appearance of the plate when retouching was neglected can be seen in the Rinder copy and in posthumous ones. Such imperfect mending, even at the corner of the plate, suggests that Blake was not free to hammer the back — or else that the chief purpose of the mending was to efface the earlier reading and that he counted on retouching to perfect the page.

Plate 35[31] (Illustration III b)

. . . .
So spoke the voice from the Furnaces, descending into Non-Entity [17]
<To Govern the Evil by Good: and [?States] abolish [?Systems].>

The last line is not fully legible in any copy, but enlargements of the Morgan, Fitzwilliam, and Rosenbloom copies, collated, made clear all but the words probably, but conjecturally, to be read as "States" and "Systems." "States" is a void with several crumbs of letters in it. The y of "Systems" is certain and the S probable and that word does fit the space.

The deleted line would have ended the page and may have led to further discussion of Evil and Good, if not States and Systems, in some earlier position; no surviving plate opens on these themes. The preceding and following pages are in a different style of lettering, and one could plausibly argue from contextual looseness of fit that this is a plate of late insertion.[19] Yet the deleted line may have been removed because of its aphoristic simplicity. Compare Plate 4.

The stigmata in the hands and feet of the soaring figure at the top of the plate derive in all copies from a single state of the plate, though in the Mellon copy only those in the feet are brought into prominence.

In the Harvard copy, in the one visible ankle of the lower figure, there is a deep erasure (or possibly a defect in the paper) retouched in grey to create a stigma where none was in the etching. (Compare Plate 26 in the Mellon copy.)

Plate 36[32]

. . . .
And the Four Zoa's who are the Four Eternal Senses of Man
Became Four Elements separating from the limbs of Albion
These are their names in the Vegetative Generation [33]
<[?West][ ] dividing Generation [ ]>
And Accident & Chance were found hidden in Length Bredth &
Highth

The Fitzwilliam copy is clearest in the first part of line 34, the Morgan copy in the latter part, the Mellon copy perhaps all through (and not vine-covered or washed); yet deletion by scratching was too thorough for a complete or confident restoration. Except for the final word, I can propose a conjectural reading which seems to fit the few fragments of letters discernible and is at least not contradicted by any fragments or by the context: "West Weigh[ing] East & North dividing Generation South [ ]g". The colon which Keynes (p. 663) supplies at the end of line 33 may be misleading.

"These" may refer only backward to "Senses" and "Elements" in the preceding lines and not to anything in the deleted line. Or is Blake intentionally removing a key to his symbols?

"Zoa's", the first word in line 44, is described as "indistinct in Blake's autotype" by Sloss and Wallis (1, 511); yet I find it clear in all copies examined. These editors seem occasionally to have misread their own notes and to report difficulties where none exist.

Plate 37[33] (Illustration V a)

And One stood forth from the Divine Family & said [1]
. . . .
So Los spoke: But when he saw <pale> [mended to] blue death
in Albions feet, [10]
. . . .

The Census describes the "last four words of the first line" of this plate as "incised instead of stereotyped"[20] but fails to note that the first six words of the line were also incised or engraved, though in relief instead of intaglio. In short the whole first line was added to the plate after etching. Wicksteed (p. 166) builds a precarious structure of speculation upon the deduction that "Blake made at least one attempt to obliterate his first line . . . and another attempt to restore it" (Wicksteed's emphasis). The attempt to obliterate, I suppose, is the fact that the line prints badly; the attempt to restore, the fact that in the Mellon copy the whole line is retouched (being especially pale in light brown ink). What Blake really seems to have attempted (with rough success) was to cut the first six words in relief, a process that required applying the letters in reverse on the smooth copper at the base of the illustration and then cutting around their outlines and hatching the background. With the seventh word Blake reverted to the traditional copperplate engraver's hand (incised in reverse line) with evident ease, though at a faulty angle.[21] One might conjecture that Blake made this switch with the intention of giving a sort of italic distinction to the concluding words ("Divine Family & said"), but the technical explanation — that the first process was too laborious — should suffice. There are no grounds at all for Wicksteed's supposition that Blake had "some perplexity in his mind" about what he wanted to write.

The mending of "pale" to "blue" in line 10 was perhaps a continuation of the impulse that produced the six raised words in line 1. But here instead of beginning with a smooth raised surface, Blake had only the raised surfaces

of the word "pale" to work with, plus whatever he could build up on the plate beside them. He could leave the final e unchanged, but to mend "p" to "b" and "al" to "lu" he had to cut away the descending stroke of the p and the ascending stroke of the l (which he managed incompletely, leaving visible crumbs) and then somehow to build up the metal for the ascenders of "b" and "l" (as he did with visible gaps). The remaining segments of the old letters could be made into the new by judicious excisions. The result demonstrates again (compare Plate 29) how limited the possibilities were for changing etched lettering. Simpler to make lacunae, to fill with tendrils.

(Nevertheless it must not be supposed that Blake had never learned how to delete and re-engrave or re-etch a small area or large by punching the copper from behind and burnishing the raised area for a fresh start. See the clusters of punch marks that scar the backs of the Gough plates, now in the Bodleian, which were made during Blake's apprentice days in James Basire's shop, some of them by Blake himself. The presumption must be that Blake was using the backs as well as the fronts of his plates and that what was already etched on the back of Plate 37 was not something slated for cancellation. See appendix on Sizes of Plates.)

A proof of this plate in a state before any tampering with the word "pale" is in Sir Geoffrey Keynes's collection and may be seen in collotype reproduction in the Census (facing p. 106). The compilers are wrong to describe the British Museum copy as "in the earlier state" (p. 113) if this means that the page was printed from the plate before the mending. The reading of this copy is "pale" rather than "blue", but it was effected by a retouching, in somewhat darker ink, of a printing from the mended plate.[22] All four copies of Jerusalem, in short, derive from the same state of the plate, with its imperfect mending of "pale" to "blue", but in the British Museum copy Blake restored the original word—a wavering not repeated in the other copies. (The Rinder facsimile shows "blue" with no trace of "pale", but the evidence of the posthumous copies proves this effect to have been the result of inking—or of the making of the facsimile. The state of the unretouched plate is clearly exhibited in the Mellon copy and in its facsimile.)

Both this revision and the addition of line 1 suggest an early plate being adapted to changed context. There is no catchword on the preceding plate; there is a catchword on this plate (37) fitting the top of Plate 38 — and it may be noted that Plate 38 was once numbered "6". The textual

reason for the change from "pale" to "blue" was probably Druidic: further on (J 65:4-5) Albion is "staind . . . with poisonous blue . . . To die a death of Six thousand years . . . ."[23]

Plate 39[35]

Line 6, "For Human beauty knows it not; nor can Mercy find it! But", has been crowded into a stanza break, after the initial inscription of the page but before etching. The indention of the first word should be taken as still marking a new paragraph.

Line 10, "And here begins the System of Moral Virtue, named Rahab.", is similarly crowded in before etching but not indented. I should be inclined still to treat the following line as a new paragraph, but Blake's final intention is not plain.

Plate 40[36]

. . . .
Gloucester and Exeter and Salisbury and Bristol; and benevolent
Bath [61]

The word "Bath" at the end of Plate 40 is repeated as the first word of Plate 41. It stands where it might seem to be a catchword yet can equally well represent the final word of line 61, crowded below the line for lack of space. Sloss and Wallis pass over it as a true catchword; Keynes ("rightly," says Wicksteed) treats it as the true ending of the line.

Keynes and Wicksteed are right, but the evidence is not easy to evaluate. "Bath" is not separated from the rest of the text, by ornament or spacing, as are the true catchwords in Jerusalem (this is, I think, the most decisive point); yet Blake himself appears to have mistaken it for a catchword when he was inking or retouching this plate. In the British Museum copy "Bath" is legible enough; in the Rinder copy it is deleted, apparently by underinking. The posthumous copies show that it was not deleted on the plate, but then a catchword that still worked would not have been deleted there, either. In the Morgan copy "Bath" is retouched, in the same greyish ink used to mend other words in this copy, including some nearby catchwords.[24] In the Harvard copy the word is retouched, but carelessly, the B written as a D, perhaps from confusion of this plate with Plate 44, where the word

"Death" is in the same position — and is not a catchword. In the Mellon copy one can see that the lower parts of the "th" did not print clearly and have not been retouched; also a line of border has been drawn so close as to scathe these letters — yet "Bath" has not been drawn over with loops, as have some true catchwords in that copy.

If we reject "Bath" as a catchword, however, we cannot continue to argue that one of the plates beginning with that word had to follow.

Plate 45[40]

. . . .
And these the names of the Eighteen combining with those Ten [40]

Line 40 is an engraved addition to the plate, after etching, made in what was originally the bottom border. It is present in all copies except the proof page, unnumbered and on paper watermarked "Edmead & Pine 1802", tipped into the Morgan copy. This might suggest an early date for the original plate, but Keynes and Wolf (Census, p. 112) think it likely that this and the two similar proofs were made on paper that just happened to be old, and Wicksteed (p. 178) finds reason in the pictorial and verbal symbolism to regard this plate as "one of the latest and maturest" in the book. The symbolism is certainly not especially early, and the script is approximately that of the majority of the plates. The added line, however, does show that it was made before the present arrangement of plates, that it was first made for one position and later modified to fit another, in short that this plate cannot be very strictly described as "one of the latest".

The historical allusions in the text do not call for a date later than 1808 — if my conjectures are valid that the voice of "Bath" is that of the pacificist Richard Warner and (more certainly) that the cutting of the chains of "Africa" signifies the abolition act of 1807.[25]

Plate 47

<When Albion utterd his last words Hope is banishd from me> [1]
. . . .
Therefore I write Albions last words, Hope is banish'd from me. [18]

The deleted first line can be read in the Harvard, Morgan, and Mellon copies, but it is sufficiently scratched out to be inconspicuous on the page. The reason for the deletion is more likely to have been a rearrangement of plates (note the closeness in text to the line above the early deletion in Plate 24) than a decision to remove the redundancy of lines 1 and 18; in

the positions of first and last lines on the plate, these could be read as a characteristic Blakean refrain.

Plate 51

[Inscribed beneath pictures, in separate prints only:]
Vala Hyle Skofield
[Incised in lower left corner, and visible in British Museum, Morgan, and posthumous copies and separate prints:]
W B [in monogram] inv & s

Blake evidently put his monogram on this pictorial plate for issuing as a separate print (several examples of which are alluded to in Keynes's Blake Studies, p. 115; Census p. 113). He let the ink cover it when printing the Rinder, Harvard and Mellon copies; yet it can be detected in the Harvard.

The Census reports an impression now in the collection of Sir Geoffrey Keynes with the names "Vala Hyle Skofield" "in white line" and the signature, quoted as "W B inv." The posthumous copies prove that the names were never on the plate; they do not appear in any copies of Jerusalem; "white line" must mean white ink, not incision; the Census statement that "this lettering is present in some copies, though mostly obliterated by heavy inking", ought perhaps to read: "present in some (or one) separate print(s) but not visible elsewhere."

Plate 53

Line 8, "Their Giant forms condensing into Nations & Peoples & Tongues", appears to have been added within a stanza break, after inscription of the page but before etching.

In line 20 "Furnaes" is corrected to "Furnaces" by insertion of a c above the line and a caret below, apparently in late stages of etching. To be clearly visible the correction had to be retouched, as it is in the Mellon and Morgan copies. It is faintly visible in the Rinder and in the posthumous Rosenbloom copies but not to be seen in the British Museum, the Harvard, or the Rosenwald.

Line 24, "The Mystic Union of the Emanation in the Lord; Because", appears to have been added within a stanza break, before etching. This knits two stanzas into one, whereas the line 8 insertion still leaves line 9 as, in effect, the beginning of a new stanza.

Plate 56

Subservient to the clods of the furrow! the cattle and even
The emmet and earth-worm(s) are his superiors & his lords. [37]

Lines 36-37 of this plate were etched right across the plate-maker's stamp (on the back of the plate) and hence printed badly in all copies. In the

Morgan copy Blake retouched several spots with pen and ink — and then added an s to "earth-worm". Was this done from aesthetic choice, as it might well have been, to match worms with cattle instead of emmet? Or from a hasty assumption that a plural noun was required before "are"? The correction does not appear in any other copy.

"The singular character of the engraved script" of this plate (matched, we may note, in Plates 61 and 88) led Sloss and Wallis (I, 548) to infer a late addition. From the same singularity Wicksteed (p. 208) infers that Plates 56 and 61 were "probably . . . the two earliest Plates to be etched for the Work." Possible evidence of an early date is the existence of a proof of Plate 56 (tipped into the Morgan copy) on paper with an 1802 water-mark — though I am inclined to accept the Census hypothesis that the three proof pages of this sort were made much later than the date of their paper. Wicksteed further calls attention to the narrowness of the plate, but narrowness in itself does not afford any evidence of earliness or lateness (see addendum on Sizes of Plates).

There may, however, be an indication of earliness in the fact that Plate 56 shares its odd size with Plate 95. If these two pages were made from the opposite sides of the same piece of copper, they may have been made at about the same date. Plate 95 does seem to have been originally etched rather early, since before final use Blake drastically remade the plate; yet even so, Plate 56 would have been made after 95 because it uses the back of the copper.

Perhaps the most datable feature of these pages will prove to be their style of lettering. But their most striking difference from adjacent (later or earlier pages), which nobody seems to have remarked upon, is surely their style of poetry:

and so on.[26]

Plate 61

In line 40 the first two letters of "Chaldean" are mended from "ch", probably in the varnish. In line 51 "Garents" is corrected to "Garments" by insertion of an m above the line and a caret below (compare Plate 53). The Census reports this correction in copies D and E, meaning either the Mellon and Morgan or the Harvard and Mellon. Actually it is visible in all copies except the Rinder. It was made in a late stage of etching and is faint unless retouched, as it is in the British Museum and Harvard copies.

Plate 63

In line 40 the last word, "Danube", is perfect in all copies except the Harvard, where the drawing of the marginal border covers the e and the bowl of the b. A purely accidental deletion.

Plate 69

In line 1 the fifth word was etched "combined" but mended to "conjoined", the m being cut and given a tail to make "nj" and the b cut to "o". In the Mellon copy, however, the word is erased from the paper and the original "combined" is restored by pen and ink.

Plate 73 (Illustration V b)

. . . .
Permanently Creating to be in Time Reveald & Demolishd
Satan Cain Tubal Nimrod Pharoh Priam Bladud Belin
Arthur Alfred the Norman Conqueror Richard John
<Edward Henry Elizabeth James Charles William George> [37]
And all the Kings & Nobles of the Earth & all their Glories
These are Created by Rahab & Tirzah in Ulro; but around
These, to preserve them from Eternal Death Los Creates
Adam Noah Abraham Moses Samuel David Ezekiel
<Pythagoras Socrates Euripedes Virgil Dante Milton> [42]
Dissipating the rocky forms of Death, by his thunderous Hammer

Sloss and Wallis (I, 588) and Keynes (713, 714) note the two lines "erased from the plate" but attempt no readings. The Rinder facsimile shows few traces of the deleted kings in line 37 and only enough of line 42 to reveal the name of Virgil among the deleted prophets. In the Mellon copy Blake drew green vines and in the Morgan looping tendrils over the erased names. But the Harvard copy shows more and the Rosenbloom copy, enlarged, reveals enough to leave nothing in my reconstruction conjectural.

The unexpected names among the preservers, to me, were Pythagoras and Euripedes (sic). Only one queen is put among the Satanic but glorious Kings of England. In 1954, I speculated that Blake had scratched away the names after Richard and John "as a precaution against Watch-fiends" (Blake: Prophet, p. 385). I did not try to guess why he also scratched away the countervailing names after David and Ezekiel. Presumably he did so to keep the balance in Eternity. Now that we see what the names all were, other reasons suggest themselves. The Blake who wrote scathingly in the 1820's of "Homer, Virgil & Ovid" and found "Tyrannical Purposes" in Dante's Comedia[27] may have had direct reasons for reducing his list of

prophets; yet Milton at one end, and some of the kings at the other, may have had to go simply to make whole-line deletions.

A similar, yet impersonal, deletion occurs, however, in both of the late copies of Milton (in both the New York Public Library and the Rosenwald copies, despite Keynes's note of erasure in only "one copy"):

Plate 76

[Incised below figures in the picture:]
Albion Jesus

From posthumous copies we can see that the names "Albion" and "Jesus" below the figures in this full-page engraving are cut into the plate in about equal strength. But in some copies Blake, by selective inking, deleted the name "Jesus", in some both names. Both are clear in the British Museum copy. "Jesus" is deleted in the Rinder and Morgan copies but "Albion" is clear, being touched with white in the latter copy. Both names are hidden in the Harvard and Mellon copies.

Wicksteed (p. 220) explains: "The name ALBION appears in most prints . . . below the youthful figure who might otherwise be mistaken for LOS. The name JESUS is less necessary, but is also found in many copies." Compare, however, the deletion of allusions to Jesus in Plate 3. Wicksteed's "many" and the Census generalization (see next footnote) obscure the

fact that "Jesus" is deleted in all copies made by Blake except the British Museum copy. (The posthumous should not count.)

The apples on the tree, four on each side, go through a different evolution: barely visible in the British Museum copy, and hidden in the Rinder and Harvard copies, they are retouched in white in the Morgan copy, in black and gold (and increased to six on the left side) in the Mellon.

It took careful inking, or retouching, to bring out these fine-line elements. All copies were printed from the same state of the plate.[28] (The nail heads are outlined in black in the Mellon copy, as they are not on Plate 51.) (In the Harvard copy the apples are outlined in black on black, with ambiguous effect: probably to cover white lines left after printing, but possibly to assert the apples' muted outlines.)

Plate 77 (Illustration VI a)

[Incised in bottom corners:]
<The Real Selfhood in the
[ ]thin P[ ] T[ ]m>

Extensive photographic attack has not recovered more of this deleted inscription than Sir Geoffrey Keynes was able to read in the Fitzwilliam copy (Complete Writings, p. 919), except for a trace or two of the doubly deleted second line. The word below "Selfhood" may be "within". To indicate the lengths of the other words, a purely speculative reading can be invented: "The Real Selfhood / is our Spectre within / in the / Poisonous Term".

Whatever the message, covered with thick foliage in the Mellon copy, it was perhaps first written as a sort of marginal motto. Then deep horizontal lines of shading were gouged across the letters, and, outside this pattern, the first and last words in the bottom line were directly erased by scratches.

Plate 82

. . . .
I have naild his hands on Beth Rabbim & his hands on Heshbons Wall
. . . . [43]
So saying: She drew aside her Veil from Mam-Tor to Dovedale
Discovering her own perfect beauty to the Daughters of Albion
And Hyle a winding Worm <be>neath <her Loom upon the scales.

Hyle has become a winding Worm:> & not a weeping Infant. [48]

. . . .
To feed the afflicted in the Furnaces: She minded not
The raging flames, tho she returnd <consumd day after day, [67]
A redning skeleton in howling woe:> instead of beauty
Defo[r]mity: she gave her beauty to another: . . .

One of the words "hands" in line 43, probably the second, is a mistake for "feet". Other signs of extreme carelessness on this plate are the misspellings: "Defo[r]mity" (line 69), "siste[r]s" (72), "soften[in]g" (77), and "Furna[c]es" (78, 79). Mended letters include the first w in line 5 and the second b in 21.

The deletions in lines 47-48 and 67-68 are inked over with vines in the Mellon, Harvard, and Morgan copies but are almost legible in the Rinder copy and can be made out completely from the posthumous ones, by enlargement.[29] The excision in lines 47-48 removes a redundancy (but the cut into "be" of "beneath" seems an accidental slip of the tool).[30] The excision in lines 67-68 if it does not improve at least does not make any worse the syntactical context of the dangling phrase, "instead of beauty Deformity".[31]

Everything about this plate suggests that it was prepared in a condition of unusual haste, or inattention, or illness, and that the text was not freshly composed but transcribed from a fair copy of much earlier vintage — consider particularly the "redning skeleton", inscribed and etched but then rejected. From the style of lettering (and from the crowding of the plate) I would conjecture a late date of etching.

Plate 83

In line 30 the A of "Affection" is mended from lower case a in the varnish, before etching.

Plate 84 (Illustration VI b)

The Corner of Broad Street weeps: Poland Street languishes
To Great Queen Street & Lincolns Inn all is distress & woe, [16]
<[ ]>
<[ ]>
<[ ]>

The night falls thick Hand comes from Albion in his strength

Some crumbs of letters remain here and there in these three gouged-out lines, but too few for any hope of reconstruction — a great pity, for Blake's neighborhoods were apparently the subject matter. Perhaps he was erasing as the place names reached toward his present life.[32]

Tendrils are added in the Harvard and Mellon copies.

Plate 85

In line 11 the M of "Myriads" is mended from lower case m in the varnish before etching.

Plate 90

In line 58 the A of "And" is mended from lower case a before etching.

Plate 91

<Forgiveness of Enemies ?can [ ] only [ ] God [ ]> [1]
It is easier to forgive an Enemy than to forgive a Friend:
. . . .

We seem very close to a restoration of this deleted first line, but the deeper gouges baffle: the deletion is not made by hatching but by a scooping out of the centers of words. Redundancy (of lines 1 and 2) or a changed position in the sequence of plates might account for the deletion — but until we know what exactly was erased we cannot safely speculate about the motive. Tendrils are added only in the Harvard copy.

Plate 93

Enitharmon heard. She raisd her head like the mild Moon [1]
<[ ]>

In the blank space following line 1 are slight traces of a deletion, irrecoverable and probably made in the varnish. Can it have consisted of an extension of the epic, but brief, simile "like the mild Moon"? That would be an early manner.

Plate 94

. . . .
O all ye Nations of the Earth behold ye the Jealous Wife [26]
The Eagle & the Wolf & Monkey & Owl & the King & Priest were there

The concluding line, 27, may be a late addition. From the evidence on Plate 95 (see comment) three conclusions can be drawn: that the text of Plate 94 was transferred without substantial change (beyond the addition of line 27, perhaps) from its earlier position on what is now Plate 95; that

the picture at the top of the page is new; and that the picture at the bottom of the page lacked, in its earlier version which is still partly discernible beneath a later picture on 95, the rays of light and the dolmen on the horizon. The Census (p. 110) describes the two pictures on Plate 94 thus: "Above, a threefold man supine. Below, a woman lying face downwards on the body of an old man; Druid stones and a sunset [or sunrise?] glow in the background."

The only error of retranscription appears to be "Chastitity" for "Chastity" in line 23.

See comment on Plate 56 for the inference that this text is of early origin.

Plate 95

[Note: In this exceptional instance, the words or letters given within square brackets are not in the usual sense conjectural: they are not visible on Plate 95 at all but are borrowed from the identical text on Plate 94 in order to indicate just where the words or letters that are here legible occur on the page. Explanation follows.]

• <[Albion cold lays on his Rock storms & snows beat round] him
• [Beneath the Furnaces & the starry Wheels & the Im]m[ortal] To[mb]
• [How]ling winds cover him [roaring seas dash fur]ious [ag]ai[nst him]
• [In the] d[e]ep dar[kness] broad li[ghtnings glar]e l[on]g th[unders] r[oll]
• [The we]ed[s] of [Death inwrap his] han[ds & fe]et [bl]o[wn inc]ess[an]t [5]
• [And wa]shd [in]cessant [by the f]or-ever rest[less] se[a-waves foaming abroad]
• [U]pon t[he] white Rock of [En]gland a Female [S]hado[w as d]ea[dly] dam[ps]
• [Of] the M[in]es of [Cornwall &] Derbyshire la[ys upon his bosom h]ea[vy]
• [M]ov[ed by the] wind [in volumes of] thick cloud [returning fold]ing [roun]d
• [His loins] & bos[om unremovable] by [swelling storms & l]ou[d rending] [10]
• [Of enr]ag[ed thunders. Around t]hem [the St]arr[y Wheels of their Giant Son]s
• [Revolve] & over the[m the Furnaces of L]os [& the Immortal Tomb around]
• [Erin sitting] in the [Tomb, to watch them unceas]ing [night and day]

---

34




• [And t]he body of [Alb]io[n was closed apart from] all [Nations.]
• [Over them the] fa[mishd Eagle screams on boney Wings and around] [15]
• Them ho[w]ls [the] Wolf [of fam]ine de[ep heaves the] Oce[an black thundering]
• [Arou]nd [the wor]my [Garments of Albion then pau]sing in [deathlike silence]
• [Ti]me [was Fini]shed [The Bre]ath [Divine Br]eath[ed over Albion]
• [Be]ne[ath t]he Fur[naces & Starry Wheels and in the Immortal Tomb]
• [And England who is Brittannia] awo[ke from Death on Al]bions [bosom] [20]
• [She aw]oke [pale &] col[d she] fainted [seven times on the Body of Albion]
• O pitious [Slee]p [O] pitious Dream [O God O] God aw[ake I] ha[ve] sl[ain]
• [In Dreams of] Chasti[ty & Moral Law I have Murdered Albion Ah]
• [I]n S[tone-h]en[ge] & on [Lond]on Sto[ne & in the Oak Groves of M]ald[en]
• [I hav]e slain h[im] in my [Sleep with the Knife of the Druid O E]ngl[and] [25]
• [O all] ye [Na]tio[ns] o[f the] E[arth] b[eh]old [ye the J]ealo[us Wife]
• [ ? ]>

A few traces of the text transcribed above can be detected in any copy, closely examined. But none of the usual techniques of magnification or photography yielded anything more than a half word or so (in the dark lines below the arm of the large drawing of Albion that fills this area), until I noticed that the open spaces that were white in the photostatic positive (enlarged) were not solidly dark in the enlarged negative but filled with shadowy images of words, quite legible in the right light. "Derbyshire" was a word that quickly led to conjecture that the text was related to the Cornwall & Derbyshire passage on Plate 94. Comparison of the two pages, line by line, yielded the evidence presented above. The legible words or letters are in relatively the same positions on Plate 95 as on 94, though not an exact retracing. For example, on 95 the first s of "restless" is above the final e of "Female" in the next line and the final e of "Derbyshire" in the next; on 94 the s is above the m of "Female" though still above the final e of "Derbyshire". In Plate 94 a paragraphic space is inserted between lines 4 and 5 which was not present in Plate 95; and so on.

In all the fairly well recovered lines it is evident that the slight variations

in position are accounted for by variations in spacing, not changes in text. The fit is so nearly exact throughout that the probability of unfound variants in the larger unrecovered portions is almost completely ruled out, except for the possibility that line 27 is new on Plate 94. There is room for it on 95 but no visible trace. In meaning it is parenthetical; it may well have been written when the passage was moved.

Below the canceled text is a canceled illustration which is practically identical to the lower two-thirds of the bottom picture on Plate 94. The new picture, of young Albion rising in anger with wrath "bright flaming on all sides around", is etched on top of the old, his flames and left leg and right foot being drawn on top of the original lines representing "Brittannia" lying upon the bosom of aged Albion, asleep or dead. Only traces of Brittannia remain, remnants of her legs above the man's; but the legs of aged Albion were only slightly cut into when Blake cut the row of flames on the right side. On the left side he scraped away most of the man's head but left three fragments, including part of an eye, now serving as crevices in a pile of rocks (consolidated into one in the colored copy). From the condition of the plate, as revealed in the posthumous copies, it is easy to suppose that Blake did not expect his earlier picture to show at all. In the British Museum copy grey wash disguises legs and feet. In the Harvard copy he added lines to increase the rock-pile effect and efface the head; in none of the uncolored copies are there positive indications that the remnant calves of Brittannia and calves and feet of old Albion are meant to be taken as part of the picture. Yet perhaps (though his text does not directly suggest it) Blake did mean to have us see what the compilers of the Census discern: "flames, which appear to be consuming a body of which only the legs and feet are seen on the right and the head, vaguely, on the left" (p. 110). For in the colored (Mellon) copy, where he might have covered all, Blake has resurrected the aged corpse (while covering the remnants of Brittannia) and given it a good head and legs: young Albion firmly elevates himself with one foot on the thigh of the old. Furthermore Blake has painted out the heavy lines on the plate which separated the young man's buttocks from the flames between his legs and has changed the flames into a sort of triple tail or threefold fleshy garment extending downward as part of the young man's back.

Also in the colored copy is added, with india ink and spots of white, what the Census describes as "a worm at the lower margin." But there is only one state of the plate.

If, as seems likely, Blake was dealing with an already finished plate (possibly the right side of the plate from the wrong side of which Plate 56 was printed: see comment above), he must have had to apply a ground to the whole plate, letting it fully cover the lower half (which he wished to retain) while he cut and etched his new picture in the top half, on top of the old picture and the text he was removing. It must then be supposed that he etched the whites so deeply that he could almost completely avoid

inking the remnants of text in the hollows of his picture — or, it may be that the shadowy letters brought out by photography represent impressions made in the paper by the acid-reduced and un-inked letters which still retained relative elevation. This explanation would account for the appearance of the lines of the body of old Albion and of some letters in the thick lines of flame or body not touched by the second etching.[33]

Plate 96

The peculiar arrangement of this page, the only one in Jerusalem in which Blake's fourteeners are doubled back on themselves to make room for a picture, can be accounted for by an earlier use of the plate. Elsewhere the pictures accommodate themselves to the shape of the text, but here the shape of the picture, belling out oddly at its base to no apparent purpose, has been determined by the shape of earlier engraving on the plate, I find. There is a good deal of submerged cross-hatching in the area below the center of the picture and along its left side near the text, the only distinguishable form being a perspective drawing of what looks like a small Grecian temple. From the top to the center of the picture some 7 irregularly spaced lines of cursive italic writing are fragmentarily visible as white loops across the thick outlines of Blake's drawing. In the Rosenbloom copy the first words are decipherable as "The greatest".

These prove to be the first words, in the same cursive engraver's lettering, of a commercial manifesto in the center of a large poster etched by Blake for "Moore & Co's Manufactory & Warehouse of Carpeting and Hosiery, Chiswell Street, Moor-fields" (reproduced as Plate 10 in Keynes's book of William Blake's Engravings: The Separate Plates, 1956).[34] The Grecian temple turns out to be the roof and walls of a "Common Carpet Loom"; the belling out of the Jerusalem picture is necessitated by the shape of the base of one of the large pillars that flank the "Carpeting and

Hosiery" advertisement. Jerusalem 96 is etched on a piece of copper cut from the lower left quarter (roughly speaking) of the plate.[35]

The visible seven lines are the beginnings of seven of the lines in the announcement of carpets and stockings for sale to private families and to merchants:

The original design, with sun-lit royal crests above, carpet-hung pillars, and "innocent" sketches of sons and daughters of Albion spinning, rolling carpets, and laboring at three sorts of looms — Common, Persia & Turkey Carpet, and Stocking Frame — may be thought of as the commercial contrary to Blake's painting of Hervey's Meditations, or the innocent contrary to "Nelson guiding Leviathan." How appropriate to etch upon it the picture of Albion regenerated and "England who is Brittannia" rejoicing, the climactic page of Jerusalem in which "all the Cities of Albion" rise from their slumbers. Too bad it was necessary to obliterate at the bottom the adaptation from Martial which Blake had inscribed for Moore & Co., if he had not suggested it to them:

— which is to say, the world of London makes you this gift: now the loomfork of Britain has conquered Babylon's needle.

Plate 98

• . . . .
• Glorious incompreh[en]sible by Mortal Men & each Chariot was Sexual <Two>fold [mended to] Threefold [11]
• . . . .
• . . . & the all tremendous unfathomable NonEns [33]
• Of death was seen in regen<ations> [mended to] regenerations . . .
• . . . .
• . . . And I heard Jehovah speak [40]
• Terrific from his Holy Place & saw the Words of the Mutual Covenant Divine
• . . . .
• . . . Humanize [44]
• In the Forgiveness of Sins according to thy Covenant Jehovah. . . .
• . . . .
• . . . where are all his Human Sacrifice<s> [48]

A wretchedly mended plate. "Two", imperfectly mended in the copper to "Three" in line 11, was doubtless a mistake of copying (note the dropped syllable earlier in the line).[36] There is an uncrossed t in line 13 (in "this"). The last words in line 33 are so crowded that "Non Ens" becomes almost a single word, "NonEns". In line 34 Blake laboriously mended "regenations" by changing "e" to "a" and "t" to "r" and eking out the parts of "ions" to make "ations". And a crowding occurred in line 48 that drove the final s of "Sacrifices" into the ornamental border; though it is clearly visible only in the Harvard copy, I take this to be a false deletion since the context demands the plural.

An extremely curious double revision occurs in line 45. What Blake originally etched on the plate was "thy Covenant Jehovah". But then he made "thy" into "the" by cutting away the long stroke (still very faintly visible in all copies) and mended the word "Covenant" into "Coventof" to achieve the reading "the Coven[an]t of Jehovah". The mending is far from perfect, possibly from technical rather than textual inattention; yet in the latest copy, the Morgan, Blake used india ink to restore the original reading, "thy Covenant Jehovah". Now either reading will work in the

passage, although "the Covenant of Jehovah" is what the context most obviously suggests. In so hasty a page, the original "thy" may have been a mistake, the mending of the plate a correction to the wording intended. Oddly, however, in no copies did Blake repair his imperfectly mended word "Covent" by inserting "an" or "na" above it and a caret below, as was his usual practice in such cases. In the only copy in which he attended to the imperfection, he restored instead the original though presumably accidental reading—possibly because he recognized that it would work textually. His not perfecting the other copies does rather argue that Blake was not strongly attracted to the reading "the Covenant of". Compare the case of Plate 69. In each instance, the only time he attends carefully to his partly mended text he restores a reading that preceded the mending. The case of the restoration on Plate 37 of the original word "blue", in the British Museum copy, is slightly different since the printing of later copies of this plate does not leave any ambiguity about the reading of the revision to "pale".

Plate 99

Beneath the picture that occupies most of this plate may be seen relics of some earlier use. Turned the other way around, the plate contains large architectural elements in the foreground and what may be a title, perhaps unfinished but thick with engraved swirls of wings or bows or large looped capitals — or all these together. It is not the pattern of the present title page of Jerusalem but seems somewhat in its spirit.

There is only one state of this plate, though variations are made in the colored and retouched copies. In the Harvard, Mellon, and Morgan copies the hands of Jehovah — or of "that God from whom all books are given", to quote the restored Plate 3 — are redrawn, concealing the small oblong that looks something like a small bible or psalm-book held between the thumb and second finger of his left hand, with his forefinger inserted between the pages. A black wash covers it in the British Museum copy. It is fairly visible in the Rinder facsimile. But this is perhaps a wrong construction.

Plate 100

That a thick thread, passing from the distaff to the woman's right hand, is "seen only in" the Mellon copy (Census, p. 110) does not signify a variant state of the plate. In the color facsimile the orange thread appears as an integral part of the plate. But in the original we can see that the whole area was printed an even color and that the thread was created, on this copy, by addition of a black wash (making a corridor of the light brown print, for the thread) and reddish orange streaks.

Conclusions as to the Sequence of Copies

Although some of the apparent evidence for the relative order of copies of Jerusalem put forward in the Keynes and Wolf Census of

1953 does not stand up under scrutiny, that order itself is, in the main, confirmed by the bits of evidence that emerge.

A shuffling of folio numbers in Chapter 3 of the British Museum copy (numbers 54-66 being written over erasures) is not repeated in other (hence presumably later) copies. If priority in a single plate were sufficient evidence for priority of an entire copy, the state of the mending on Plate 20 would prove the British Museum and Rinder copies to be earlier than the other three complete copies. But the watermark evidence, in conjunction with the Linnell receipt of a completed Chapter 2 in December, 1819 (see note 8), indicates that Chapters 1 and 2 of both these copies, and the Cunliffe copy of Chapter 1 — all on 1818 paper — were printed more or less simultaneously and before any copies of Chapters 3 and 4. These final chapters of the British Museum copy (at least) were apparently printed before the Harvard and Mellon copies, however, since they contain paper of 1818 and 1819 as well as 1820, while the latter contain only 1820 paper.

Nothing indicates the relative positions of the Harvard and Mellon copies; the Census designation of these as copies D and E becomes an arbitrary matter.

The watermarked paper of 1824 and 1826 in the Morgan copy remains the only specific evidence that it is the latest of the five.

To return to the question of the order of plates in Chapter 2, the evidence is fairly conclusive that the supposedly "standard" order is neither the latest nor the earliest order established by Blake. It is found in the middle pair of copies, the Harvard and the Mellon, but in his latest copy, the Morgan, Blake returned to the order of his first two copies. The strongest argument for considering the variant order of the 1820 pair as Blake's final preference rests on his statement in a letter of 12 April 1827 about Jerusalem: "One I have Finished." The usual assumption is probably a sound one, that he was here referring to the one completely colored copy, the Mellon. It does not follow that he had not finished it some years earlier; he had it on hand because he could not "get a Customer for it." Moreover, concentrating on the illumination of that copy he may not have concerned himself greatly about the textual sequence — else why return to his original order for his most recently printed copy, the Morgan? Perhaps the soundest conclusion is that Blake found both sequences attractive but considered neither definitive.

Evidence as to attention paid to the text itself, after printing, is perhaps too slight to be of much value in this connection. The unique restoration in the Mellon copy, of "conjoined" to "combined" (Plate 69), is more impressive than the Morgan copy revision of "worm" to

"worms" (Plate 56). On the other hand the neglect of the imperfect mending on Plate 98 ("Coventof" for "Covenant of") except for the restoration of "thy Covenant" in the Morgan copy points the other way. Furthermore the retouching of rude mendings in other plates is carried through more conscientiously in the Morgan than in any other copy and rather less conscientiously in the Mellon than in any other. While it cannot be said that Blake seriously neglected the Mellon text even while devoting his greatest effort to the color and high finish of its illumination, I think that we are forced to conclude, from the treatment of weakly printing passages on Plates 22, 27, 29, 37, 40, 53, 61, 76, and 98, that Blake took more pains to make the text legible in the Morgan copy than in the Mellon. Finally, that he should take the pains to rearrange the plates or pages of Chapter 2 in his latest copy, to return to his earliest order after having tried an alternative arrangement in two copies, lends perhaps decisive weight to the conclusion that the text of the Morgan copy comes the closest to representing Blake's final preferences. (See addendum on Numbering of Plates)

Addenda List of Copies

The letter designations given in parentheses are those of the Keynes Bibliography of 1921 and the Keynes and Wolf Census of 1953, respectively. (See note 1 above.)

Copies Printed by Blake:

• British Museum copy (B in Bibliography, A in Census) 100 plates on 100 leaves foliated by Blake 1-100; arrangement (using Harvard and Mellon [Stirling] copies as standard, not to disrupt editorial tradition) 1-28, 33-41, 43-46, 42, 29-32, 47-100. Watermark J Whatman 1818 (20 leaves), 1819 (1 leaf), 1820 (6 leaves) (see note 8.)
• Cunliffe copy (A* in Blake Studies; B* in Census) 25 plates on 25 leaves, foliated by Blake 1-25. Watermark J Whatman 1818 in 6 leaves. Red-brown with watercolor washes. Collection of Lord Cunliffe. (The condition of Plates 1, 2, 3, 4, 7, 14, and 20 was kindly checked by Lord Cunliffe, that of Plate 20 also color photography; and all plates were searched for numbers.)
• Rinder copy (A, C) 100 plates on 100 leaves, foliated by Blake 1-100; arrangement as in British Museum copy. Watermark J Whatman 1818, 1819, or 1820. Black, some plates touched with sepia and chinese white; green tinting on Plate 16. Collection of the late Frank Rinder; facsimile in deep etched gravure (Blake Trust, 1952). (Examined only in this facsimile.) (See postscript.)
• Harvard copy (C, D) 100 plates on 100 leaves foliated by Blake 1-100; arrangement adopted as "standard" in Keynes's list and editions. Watermark

  ---

  42

  
  
  
  J Whatman 1820 in 21 leaves. Black, touched with india ink; some plates lightly washed with brown. Harvard College Library.
• Mellon copy (D, E) 100 plates on 100 leaves, foliated by Blake 1-100; arrangement adopted as "standard." Watermark J Whatman 1820 on 25 leaves. Orange, with elaborate watercolor washes, india ink, and gold. Collection of Paul Mellon (previously in the collection of William Stirling); facsimile in color, Blake Trust, 1951 (for bibliographical purposes this splendid facsimile is less reliable, because of the use of multiple stencils, than is the simpler photographic facsimile of the Rinder copy; facial expressions and other nuances not our present concern are also not always faithful to the original).
• Morgan copy (E, F) 100 plates on 100 leaves, foliated by Blake 1-100; arrangement as in British Museum and Rinder copies (though posthumously bound in "standard"). Watermark J Whatman 1824 (5 leaves), 1826 (19 leaves, including Plate 36). Black, with some gray washes and india ink and lighter ink. (Plate 36 ["standard" 40] labeled "from another copy.") Pierpont Morgan Library.
• Isman copy (F, G). The copy which "was believed to be in the possession of Mr. Felix Isman, New York, in 1921" (Blake Studies, p. 117) has never been described nor, recently, located. (It might turn out to be a posthumous copy.) Mr. Isman died in 1943. Today his widow has no recollection of having seen a copy of Jerusalem in their library; "Mr. Isman was in the habit of giving a first edition to a valued friend, as a Christmas present. . . . it was not in the library when he died."

Copies Printed Posthumously:

• Fitzwilliam copy (G, H) 100 plates on 100 leaves. Watermark J Whatman 1831 or 1832 in some leaves. Printed in red-brown. Fitzwilliam Museum, Cambridge.
• Rosenwald copy (H, I) 100 plates on 100 leaves. Watermark J Whatman 1831 or 1832 in some leaves. Red-brown. Library of Congress, Lessing J. Rosenwald Collection.
• Rosenbloom copy (I in Blake Studies, J* in Census) 100 plates on 100 leaves, but with a duplicate Plate 20 in place of Plate 16. Watermark J Whatman 1831 in some leaves. Red-brown. Collection of Charles J. Rosenbloom.

Single Plates (Designated "Proof" when unnumbered by Blake):

• Pl. 1 proof in brown & black. Collection of Geoffrey Keynes; collotype facsimile as frontispiece to 1952 Rinder facsimile.
• Pl. 9 (numbered 25 or 23 on plate) in green with green and grey-blue washes. Rosenwald collection.
• Pl. 19 proof in blue green. Rosenwald collection.
• Pl. 20 (number 16 incised in upper right corner) in black. Rosenwald collection.
• Pl. 24 proof (not seen; reproduced in Laurence Binyon, English Watercolours on loan to Japan, 1929).

---

43




• Pl. 25 proof, red-brown with watercolor washes; printed on back of Europe frontispiece; trimmed to exclude text. Collection of Kerrison Preston.
• Pl. 28 proof, black. Watermark Edmead & Pine 1802. Tipped into Morgan copy.
• Pl. 32[46] proof, red brown with pink and blue washes; on back of Europe title-page; trimmed to picture only. Preston collection.
• Pl. 37[33] proof, black with blue and grey washes. Keynes collection.
• Pl. 38[34] proof, black with grey wash. Rosenwald collection.
• Pl. 38[34] (number 6 incised in corner) green, with black additions (or smudges). Rosenwald collection.
• (These two 38's may be those listed as "28" in Blake Studies, p. 115.)
• Pl. 41[37] proof, red-brown with watercolor; on back of Europe frontispiece; trimmed but including text in scroll. Preston collection.
• Pl. 45[40] proof, black. Watermark Edmead & Pine 1802. Tipped into Morgan copy.
• Pl. 47 proof, red-brown with color washes; on back of Europe title-page; trimmed but including line of text beneath picture.
• Pl. 48 proof, grey. Rosenwald collection.
• Pl. 50 (trace of "19" in corner), black. Rosenwald collection.
• Pl. 56 proof, black. Watermark Edmead & Pine 1802. Tipped into Morgan copy (not noted in Census).
• Pl. 58 (trace of numeral in corner?) dark grey, with yellow watercolor and india ink. Watermark J Whatman 1818. Rosenwald collection.
• Pl. 78 proof, dark green and black. Rosenwald collection.
• (Plates 50, 51, 99 listed or mentioned in Blake Studies, p. 115, not examined.) (For prints of Pl. 5 & 53 back to back, see Postscript.)

Note

Mistaken report of the Morgan copy foliation (given earlier as "standard" but in Census as 1-28, 43-46, 29-37, 42, 38-41, 47-100, from a confusion of numerical equivalents) created a legend that there are three Blakean arrangements of plates, with the corollary that the "standard" order is that of "the majority of the copies" (Keynes, 1957, p. 981). To confuse matters further, a printer's error on the same page makes Keynes seem to declare that "the arrangement is constant except in one copy" (he meant to say "two"). Actually it is the variant order that is found in a majority of three of the available copies.

The possibility that some of the scattered single plates — the four printed in green or some of the seven in black — may be remnants of another complete copy does not seem great. For one thing, none of these single leaves has any trace of the foliation, outside the plate, which Blake inscribed on the leaves of all finished copies.

Table of Catchwords

Angle brackets signify deletion made in the copper; square brackets conjectural readings. No catchwords appear in the plates not listed here. In the Mellon copy the catchwords of Plates 5, 9, 12, 13, and 37 remain but in effect are deleted by loops, lines, or washes. That of Plate 13 is hidden also in the Morgan copy.

                                                   

Plate	Catchword	Plates so beginning; * means plausible fit.
5	His	6* 17 31[45] 38[45]
7	<Con->	9*
8	Con-	9*
9	To	11* 65*
11	<[?One]>	14
12	And	13* 15 30[44] 37 57 59 70 71 72 75
13	One	14*
14		(Deleted colophon: <End of the / 1st Chap:>)
18	<His>	19* 17 31[45] 38[34]
19	<Jeru>	23* Jerusalem
30[44]	His	31[45]* 38[34] 17 6
37[33]	His	38[34]* 31[45]* 17 6
38[34]	By	39[35]* 67
40[36]		("Bath" not a catchword; see above.)
43[38]	With	44[39]*
47	These	48*
48	The	49* 50* 60 74 90 5
49	The	50* 60 74 90 5
50		(Colophon: End of Chap. 2d.)
53	<The>	50 49 60* 74* 90 5
65	I[n]	(half hid by border) 66* 58 54 10 (Into)
66	And	70* 71* 72* 57* 75* 37[33] 30[44] 15*
70	His	38[34] 31[45] 17 6
		(Mended in Rinder copy to And) 71* 72 75* . . . 15*
79	En-	80* (Encompass'd) 93 (Enitharmon)

From these meager data, the following deductions are further supported by graphic and textual evidence:

Plates 7 and 9 were originally together, and Plate 8 is an insertion. Plate 10 is an amplification of material on 9, which was once followed by 11. If "One" is a correct reading of the deleted word on 11, the plate originally following it must have been discarded. The deletion of "His" on 18, despite its present fit, signifies two moves — once to precede a plate that did not begin with this word, once to its present location. Thicker lettering also suggests that 18 was not etched in the same period as 17 and 19. But we also know that 19 was once differently numbered (see note on Numbering); and its deleted catchword shows that 19 was once followed by 23 (which itself once had a different number). Plates 20-22, however, do

appear to be insertions in the present position, and we know that 20 was once numbered "16".

From the evidence of catchwords and plate numbers, then, Chapter 1 shows most signs of rearrangement and supplementation after the first etching of plates. Little shifting seems to have occurred in Chapter 2 (but see note on Numbering) until after the printing of two copies.

Several signs of earlier shifting turn up in Chapter 3. Plate 53 may once have been followed by 60 or 74: each of the latter begins on the same theme and one may have been made to replace the other and then in turn been moved ahead. Plate 66 was not designed for its present position; 70 must have been followed by a plate now discarded.

Sizes of Plates

Measurements, in centimeters, were made from edge of impression to edge of impression at the widest point, of the plates in the Morgan copy of Jerusalem. Since some edges are uneven, another measuring might not give identical results; in other copies the actual impressions may vary by a millimeter or two either way.

As measured, the plates of Jerusalem vary in size from 19.9 x 13.7 (Plate 56) to 22.5 x 16.4 (Plates 24 and 75). But 89 of the 100 plates fall into three discrete sizes. Exactly fifty cluster with slight variations around a size of 22.2 x 16.2; another 28 plates cluster around the size of 21 x 14.8 (somewhat more than a centimeter shorter and narrower than the first size); and 11 plates are in a size as long as the first but as narrow as the second. The remaining 11 plates are scattered in what appear to be five discrete sizes, larger or smaller than these three.

Copper plates were poured into moulds and hardened and then hammered and beveled; for small sizes they were cut in half before beveling: the edges would vary from hammering, cutting, and beveling, but the discrete clusters of sizes found here seem to signify different moulds — or at least different batches of cuttings. The plates of America (early 1790's) run somewhat larger in both dimensions than those of Jerusalem, around 23.8 x 17 cm. The plates for Visions of the Daughters of Albion (of the same period) can be accounted for as halves of the same size. Halves of the Jerusalem size appear in the plates of Milton. If America and Jerusalem were productions that overlapped in the workshop, we would expect to find that some of the odd sizes in Jerusalem came from the batch used for America. But the sizes in Jerusalem (and Milton) do not lie close to any of the sizes found in America and other early works. If the inference of discrete batches is correct, this evidence supports the usual assumption that there was a lapse of some time between the earlier Illuminated books and the beginnings of Milton and Jerusalem. Jerusalem at least (it is harder to tell about the half-plates of Milton) does not appear to use any copper left over from the earlier works.

It may also be inferred that Blake's economical practice of using backs and fronts of plates (there are several visible platemaker's marks in

Europe) had not left him any unused backs among the plates from which he continued to print copies of his Lambeth books.

Of the 11 plates in five odd sizes, I find nothing very significant to say. Plates 71 and 77 are in the widest size (22 x 17, 21.9 x 16.8) — the latter plate for the good reason that it contains the widest matter in Jerusalem, pentameter lines in double column; the former for no obvious reason (can it be printed from the back of the same plate?). The narrowest plates, 56 (19.9 x 13.7) and 95 (20 x 13.6), may be early (on the basis of other evidence) but do not demonstrate earliness or lateness by their narrowness, for there is no early work by Blake with plates of these dimensions. Very probably the suspicion that plates of odd sizes may be intruders — early or late insertions — is mistaken. Plate 96, cut from the Moore & Co. plate, is not textually an intruder.)

More fruitful is the discovery that, when we abandon the grouping of roughly equivalent sizes and make a table of particular sizes, many are represented by two or four examples and many of these pairs or potential pairs are closely related in pagination and in appearance (style, lettering, thickness of varnish, or other qualities). Can anything be made of the hypothesis that two pages with plates of the same size may represent the two sides of a single plate? My measurements were too crude to be taken as proof of identical shapes even when identical — or to rule out such identity even when varying by a millimeter or so. Nor should we expect the two sides of a plate to have identical beveling and to make identical impressions.[37] Something other than chance distribution, however, appears to lie behind the following groupings.

Here is a list of all sizes represented by two or more plates of Jerusalem. In parentheses I indicate which potential pairs seem "likely," i.e. closely related in content, script, or some other indication of vintage. An asterisk (*) denotes presence of platemaker's stamp, a certain but not always present indication of the back of a plate.

• 20.2 x 14.4 (pages 64* & 96) (likely, and the pair of 96 would have to be the back of the plate; we would expect it to be used first)
• 20.8 x 14.8 (97 & 98) (likely)
• 20.8 x 15 (86 & 88) (unlikely, at least in lettering)
• 20.9 x 14.6 (54 & 55) (likely)
• 20.9 x 14.7 (58 & 81) (unlikely; 58 seems to pair more obviously with 57 [20.9 x 14.8])

---

47




• 20.9 x 14.9 (36[32] & 62 & 84 & 93) (possibly two pairs, though all four pages seem disparate: unlikely)
• 21 x 14.8 (69 & 85) (likely)
• 21 x 14.9 (34[30] & 67 & 94) (the latter pair likely, the odd third not pairing easily with either)
• 21.1 x 14.8 (60 & 91) (likely)
• 21.1 x 15 (15 & 17 & 22 & 30[44] & 44[39] & 49) (likely pairs are: 15 & 22, 30 & 44)
• 21.8 x 15.9 (79 & 80) (likely)
• 22 x 16 (8 & 29[43] & 43[38] & 70 & 82) (likely pair: 29 & 43)
• 22.2 x 14.6 (10 & 33[29]* & 73) (likely pair: 10 & 73)
• 22.2 x 16.1 (1 & 20 & 28 & 31[45] & 35[31] & 42 & 65 & 66 & 76) (likely pairs: 35 & 42; 65 & 66)
• 22.2 x 16.2 (4 & 7 & 26 & 48 & 53) (likely: 4 & 7)
• 22.3 x 16.1 (18 & 19 & 23 & 27 & 37[33] & 38[34] & 46[41]) (likely pairs: 18 & 19, 38 & 46)
• 22.3 x 16.2 (3 & 5 & 6 & 12 & 45[40]) (likely: 5 & 6, 12 & 45)
• 22.3 x 16.3 (14 & 74) (not very likely, yet possibly same vintage)
• 22.4 x 16.1 (13 & 40[36]) (likely)
• 22.4 x 16.2 (11 & 50 & 52) (any combination likely, but pagination favors 50 & 52)
• 22.5 x 16.4 (24 & 75) (likely)

This table accounts for 73 pages and holds 33 possible pairs or 21 likely pairs (saving 47 & 48 for the next list). It leaves 27 pages unaccounted for, each unique in its size, as now measured. But any of the nearly identical pages may be disguised pairs. The following table includes all such possible pairs (defined as within one millimeter of identity in either or both dimensions) which are also likely (but excluding the unlikely).

• 19.9 x 13.7 & 20 x 13.6 (56* & 95) (discussed above)
• 20 x 14.3 & 21.1 x 14.4 (89 & 92*) (likely)
• 20.9 x 14.8 & 20.9 x 14.7 (57 & 58) (likely)
• 20.8 x 15.9 & 20.9 x 16 (47 & 48) (likely)
• 22.2 x 14.7 & 22.2 x 14.8 (59 & 63*) (likely)

(Proof that all likely pairs may not be actual pairs lies in the example of 72* & 100*, which are very close in measurement — 22.3 x 14.7 & 22.3 x 14.6 — but are both backs of plates.)

Altogether we have found 26 likely (and possible) pairs, involving over half the pages in the book. How closely in sequence do the partners appear? Here is the list in sequent order:

Note that in the Rinder order the pairs involving the second chapter would run: 12&40 13&36 31&42 34&41 38&43 39&44 47&48 50&52. (I fail to see why — is it only chance? — Plates 38 and 39 in this order, which are paired respectively with 43 and 44, become 43 and 44 in the Standard order!)

In either order the plates of 8 of these pairs are adjacent; the plates of almost half (12 out of 26) are either adjacent or within three numbers of each other. Nor are most of the others distributed at random. The series 40 41 42 43 44 (in Rinder order) are (hypothetically) on the backs of another, more scattered series 12 31 34 38 39. (In the standard order the series 40 42 43 46 are on the backs of the series 13 29 34 38.) Further on there is another tight series 91 92 94 95 96 on the backs of a looser series 56 60 64 67 89. These patterns altogether account for 22 out of 26 pairs and seem to indicate something more than a chance distribution. The adjacent or straddling pairs (5&6 4&7, for instance) suggest a working from front to back at once or almost at once. The more scattered series suggest accumulations of one-sided plates all used on their backs at once (i.e. for plates 40 to 44 and 91 to 96).

If we grant that these patterns may approximate the methods actually followed, can we make any meaningful deductions at this point? Perhaps not in isolation. But I suggest that further study of the textual and technical continuities and discontinuities in the plates of Jerusalem might benefit from attention to these potential pairings. Such odd couplings as 10&73 and 24&75 may provide clues to an earlier arrangement. Indeed one clue of this sort enters into the discussion, above, of Plates 56 and 95.

Blake's Numbering of the Plates

In each copy of Jerusalem printed by Blake he numbered the leaves consecutively with pen and ink just above the top right edge of the printed plate. It is this foliation, after printing, that gives us the two different arrangements for Chapter 2 discussed above. Fragments of a more mysterious tale, however, are told by the numbers which Blake engraved in (or sometimes etched near) the top right corners of his plates. These were apparently made for his guidance (or a helper's guidance) before printing; they were engraved or etched so lightly that they were usually hidden by the inking of the plates; yet they show up, occasionally in a row of deleted and undeleted numerals, on nearly all the pages of one or more of the posthumously printed copies, being particularly legible in the Rosenbloom copy.

In the British Museum copy only nine numbers are visible — and only one of these corresponds to the position of the plate in this or any known copy. Obviously Blake did not bring his numbering up to date before printing; quite possible most of the plates lacked numbering of any kind when he printed the British Museum and Rinder copies. (In the following discussion, "J" is used to desgnate the number of the plate in the "standard" Keynes arrangement, i.e. that of the Mellon and Harvard copies.)

In the British Museum copy the title page, J2, is numbered "1" and the preface, J3, bears a deleted "2" (indicating a time when the frontispiece, J1, was not present or did not precede title and preface); J28, the first page of "Chap: 2", is numbered "1°" (possibly meaning "10", possibly from an incised "18" or "12" — in any case pointing to a time when the chapters were considerably shorter; J35 is numbered "31" (its actual position and folio number in this copy); J37 is "36" (a symptom of early rearrangement in this area, for its actual position and folio number are 33); J38[34] is numbered "6" (and superficially this plate looks as if it might originally have followed J5 — but see below); and J45[40] has two digits side by side, the first deleted, the second an undeleted "9" — probably one number replacing another, but just possibly a half-deleted two-digit number.

(On J23 there also seems to be a trace of some deleted number.)

Perhaps the "6" on J38 is worth discussion. This is one of a group of plates, J5, J38 ("6"), J7, J45 ("9"), and possibly J4 (first page of "Chap: 1"), which are very similar in design, in line, in general appearance, and are unlike other plates in the first half of Jerusalem. J6, lying between J5 and J7, is strikingly different in design and even in thickness of lettering — an obvious insertion. J38 ("6") looks as if it belonged here; it even meshes with the catchword on J5 ("His"). Yet its own catchword ("By") does not fit the top of J7; the fragmented sentences at the beginning and end of its text cannot possibly be spliced to the adjacent fragments of J5 and J7; moreover the text on J6 is so tightly related to the beginning of J7 that it must be a repetition or variant of some earlier canceled plate in this position. As for J45 ("9"), it does not link directly to any others in the group. In short, we lack sufficient evidence for a valid reconstruction even of a small segment of an early arrangement of Chapter 1. We can, however, see that the first etching of plates was followed by a good deal of shuffling and reorganization and substitution of new plates before Blake was ready to print the first extant copies.

Plate numbers also apparently deriving from early stages of production are found on four of the Jerusalem pages in the small bundle of miscellaneous pages and proofs in the Rosenwald Collection. Here is another J38 numbered "6"; here are a J9 numbered "25" (or just possibly "23", as someone has penciled on the mounting paper), a J20 numbered "16" (a number deleted but visible in posthumous copies, beside the "20" that must have been added later), and a J50 numbered "19" (a number still visible and not apparently deleted in posthumous copies). Oddly Blake never changed or supplemented the "19" on Plate 50, which is marked "End of Chap: 2d" (perhaps guide enough). It indicates that at one time the chapters were under ten pages in length. Perhaps the number of J28, first page of Chapter 2, was actually "10" (see above). At that rate the "XXVIII Chapters" would have made a book of about 250 plates.

In the Cunliffe copy (Chapter 1) no numbers are visible.

In the Rinder copy the only number to be made out from the facsimile is this same "1°" on J28

The Harvard and Mellon copies tell a different story. Before their printing Blake evidently numbered or renumbered all or nearly all his plates to conform to his current plan. Visible in one or the other of these copies, or in both, are numbers corresponding to the actual positions of J1-13, 16, 19-22, 25, 26, 28-32, 36, 38, 40-43, 47, 49, 52, 54-56, 58, 59, 61, 63, 65, 67, 68, 70, 72-74, 77-81, 84-86, 88, 90-93, 98. Some of these numerals were etched, possibly when the corner positions had already been occupied by engraved numbers now deleted: J52 (near center of page), 46, 56, 65, 68 (at end of line 2), 72 (end of line 11), 74 (half inch from corner). From the posthumous copies we can further deduce that at this time numbers indicating their current positions were given to J14, 15, 17, 18, 23, 24, 27, 33-35, 37, 39, 44, 48, 51, 53, 60, 62, 64, 66, 69, 71, 75, 76, 83, 89, 94, 96, 97, 99, 100.

(In the Harvard copy the etched number on J75 looks very much like "77" and may possibly be another trace of earlier arrangement; yet the etched number in the same position that turns up in the Morgan and the Rosenbloom copies is "75".)

This tally leaves only six unaccounted for. On J45[40] the "9" (or the two-digit number ending in "9") was never (visibly) replaced. In the Harvard copy the "46" on J46[41] is very lightly etched; it reappears as "45" in the Morgan and posthumous copies. J50 with its engraved "19" remained unchanged. J57 was probably correctly numbered; the "5" is visible in the Morgan copy, in a two-digit number. The same may be said of J82, an "8" being visible in the Morgan copy. J95, however, which we know began as 94, has no visible number in any copy. Since Blake numbered its replacement, J94, we cannot argue that the change was made after the time of numbering; indeed it was made before the earliest known copies. By this recount, all but two of the 100 plates seem likely to have had correct numbers for the printing of the Mellon and Harvard copies; more likely than not they all had.

In the Mellon copy further indications of early shifting in Chapter 1 appear in traces of two earlier numbers. Beside the "18" on J18 there is a deleted number that looks like "12", though in the Morgan and Rosenbloom copies it looks more like "20". And on J26 there is a trace of a deleted number below the "26".

No further numbering appears to have been done after the printing of the Mellon and Harvard copies — yet when he came to print the Morgan copy Blake did not simply disregard his numbers. It might be supposed that he simply printed — or had printed — a complete set of 100 leaves and then rearranged the second chapter before adding folio numbers (inscribed decisively but in the extremely shaky hand of illness or of age). But there are indications that he thought about his sequence and rearranged

the copper plates before the printing. He did not renumber J29-41, 43-46, the plates that change position in Chapter 2. He did, however, scratch the "standard order" numbers on these plates, to delete them. The deletion marks are clearly visible on J31, 32, 33, and 41. The numbers on J30, 35, 40, 44 are too faintly printed to reveal scratch marks, but this faintness may be the result of scratching. And more thorough deletion would be a possible explanation for the invisibility of "standard" numbers on J45 and 46. All things considered, the final act of returning to his earlier order seems to have been as deliberate as the decision to depart from it had been.

There remain to be noted three further instances of abandoned early numbers which show up on the posthumously printed plates. On J19 is a row of three numbers, "19", possibly "16", and "19" (apparently indicating a double move). On J23, beside "23", is a deleted number possibly "24". And on J37[33] beside the "37" there is a deletion that may be the "36" visible in the British Museum copy, there undeleted.

Note: Blake's foliation of the Morgan copy is worth further remark. Little study has been made of the differences between what is called "forger's tremble" and the tremble of physical weakness. The reason the forger's hand trembles, however, is that his hand is not habituated to the style he is imitating; as he proceeds his trembling will decrease. The numbering of these pages shows a reverse symptom: the first numbers are inscribed with a bold confidence (for Blake's hand is indeed well trained in his own style), but a wave of weakness causes the "5" to waver and the "6" to tremble exceedingly; recovering, perhaps with a pause, the writer makes a steady 7, 8, 9, but trembles again on 10; again slightly on 13, greatly on 17, and so on. Waves of weakness seem to affect the writing, reaching subsequent peaks on 23, 29-30, 37-38, 47, 56-57, 63, 70, 76-77, 82, 87-88-89, 96-97.

In this context, the signs of erasure and mending may point only to mistakes made in weakness. The "82", a very shaky numeral, was vaguely mended during inscription; there is similar mending and possible erasure in "38". But something more curious occurs in the first pages of Chapter 2. The "29" (on J33[29]) is written on top of an erased number, possibly a "33". The "30" (on J34[30]) is over an erasure, and 30, 31, 32 are in black ink but mended in gray ink, though the mending seems only a retracing without changes. Did prophet and spectre have one final wrestling over these twice-shifted pages? If so, the earlier Blake firmly won the victory.

The Dating of Blake's Script

Blake's handwriting, of which we have large samples over a period of 43 years, exhibits some gradual modifications over the decades but would not be easy to arrange in a close chronological series from appearances alone. Milton scholars count themselves fortunate to have found one precisely datable change in their author's handwriting, his discarding the

Grecian for the Italian e in the course of the writing of Lycidas. No similar conscious or unconscious change of a letter-style has yet been found in the manuscripts of Blake.

Blake's public script, however, is another matter. The lettering of his Illuminated Books and of his etched or engraved inscriptions differs from his handwriting not only in appearance but in the fact that it is a product of deliberate art and comprises several deliberately chosen varieties and changes of style. For formal inscriptions Blake early learned both a flowing engraver's cursive script and a simple upright roman letter. In his first experiments in Illuminated Printing he employed both of these plus a more efficient italic script, print-like and only slightly cursive. This simple italic became the staple of his subsequent text style, varied more or less subtly on different occasions and sometimes combined with the simple roman.

In the period 1790-1795 he tried several interesting modifications of his standard alphabet. Several variant letter-forms, various flagged h's, an f with an extra serif at the top, and certain contrived variants in the descenders, can be seen to spring up, flourish, and then suffer gradual neglect. But nearly every one of these reappears from time to time as a graceful variant in later texts. Its presence or absence in an undated plate is extremely difficult to assess as evidence of period, although it may be useful to distinguish certain trends.

One striking exception, however, is found in a letter-form that Blake invented in about 1791, employed with unwavering consistency for fourteen years, and then as suddenly and with ruthless consistency discarded. This is an italic small g with its serif or topknot on the left side instead of the right. The conventional g may be seen throughout the Songs of Innocence (1789), alternating occasionally with a sort of manuscript g that has no proper serif; the left-turned g appears throughout the Songs of Experience (dated 1794 but etched, there is now reason to see, 1791-94), except for those songs originally etched among those of Innocence. The date of adoption can be narrowed by noting the overlap of Thel and The Marriage of Heaven and Hell. Thel, dated 1789 but containing (as others have observed) a Motto and a concluding plate that seem late additions, uses the conventional and manuscript g's — except in the Motto and last plate, where the new g takes over. The Marriage is dated 1790 by internal and marginal evidence on its third plate but was the product of several years. The old g rules up through Plate 5 and in Plates 11-13 and 21-24; the new g in 6-10, 14-20, and in the "Song of Liberty" (25-27). This overlap narrows the date of change to late 1790 or early 1791. Every serifed g thereafter points leftward, even in inscriptions made for Cumberland or Hayley, including the inscription "June 18, 1805 . . . Bridge Street" in the plates for Hayley's Ballads. After that date Blake not only abandoned the leftward g but corrected it in reissues! In the 1807 reissue of these Ballads the date is not changed, but the g in "Bridge Street" has a new rightward serif.

Blake's Gates of Paradise, etched "For Children" in 1793, with leftward g's, was reissued many years later with some further etching, during which the g's were given new, right-facing serifs, opposite visible remnants of the old. In late issues of Songs of Experience, for example in the Rosenwald and Fitzwilliam copies, which may be seen in facsimile and microfilm, when Blake retouched the lettering he added new serifs beside the old. Even the g in the large title of "The Tyger" has two large serifs like horns. Going over the text of "The Angel" in gold, he put two golden serifs on each g, this time respecting both serifs — perhaps absently, perhaps with some amusement at his own device.

Blake thus, for whatever reason, left us a distinctive chronological zone marked by two precise termini. After two years of searching I have found no exceptions in any of his work the dating of which is founded on more than conjecture. There are a few single plates with conventional g's which have traditionally — but more or less arbitrarily — been assigned to the middle period (1791-1805). Some of these are rather obviously mis-assigned. The etched "A Divine Image", a rejected Song of Experience, has been collected after the Songs as an "additional poem" and dated 1794 but purely as a convenience. It fits much more plausibly ahead of the published Songs of Experience, as a trial Song etched in 1790 but supplanted by the subtler "The Human Image" before the final collection was assembled. (See Robert F. Gleckner, "William Blake and the Human Abstract," PMLA, lxxvi (1961), 373-379, for a reasoned plea for an early date in the absence of "real evidence to the contrary.")

A more difficult exception is "To Tirzah" (with conventional g's) which has always been recognized as a late addition to the Songs of Experience but given a variety of conjectural dates. In his text Keynes puts it "Probably . . . about 1801" (p. 220); but in the Census (p. 55) four copies of Songs containing "To Tirzah" (F,I,J,K) are assigned to 1795-98 and four (L,M,N,O) to 1799-1801. Moving the dates of these eight copies after June 1805 will play hob with the Census table of copies A to CC, but it is time to realize that this is only a table of convenience resting largely on impressionistic evidence and, in its few points of apparent data, full of self-contradiction. Copy L is said to be inscribed with "the initials JS, dated 1799," but the initials (a kind of doodle at the bottom of the page) and the "date" (at the top) are in pencil; moreover the date is written "1799-", the hyphen indicating that this is an owner's or bookseller's suggested terminus ab quo. Copy O has an ink inscription "Mrs. Flaxman April 1817" yet is assigned to 1800 because the order of arrangement of the songs "suggests a date of about" that year — an argument resting upon circles within circles. Actually copies P (on "1802" paper) and Q (on paper dated 1802 and 1804) are the first "To Tirzah" copies with tangible termini ab quo: their dates could easily be after 1805.

I believe the only other exception is the inscription on "Albion rose". Sir Geoffrey may be right in supposing that the color-printed copies, in

which the inscription is not visible, were made in about 1794. But the conventional g in "Giving" plus the symbolism in "the dance of eternal death" (which belongs in the Milton period) now put the inscription after 1805, for the symbolism rules out 1790.

Certainly none of these apparent exceptions bears any weight of evidence against the rule of g. As for Milton and Jerusalem, unfortunately the title pages, dated "1804", contain no lower-case g's. The plates of text, however, employ exclusively the conventional post-1805 g and I think cannot safely be dated earlier. This at least rules out the whimsical hypothesis I once entertained that certain curiously different plates inserted in the two later copies of Milton might be relics of a very early version. At least they cannot have been inscribed very early.

Postscript: In the British Museum there are proofs of Plates 5 and 53 back to back on one leaf, both similarly colored in a variety of opaque and translucent colors including blue, yellow, and red. Just possibly these are not proofs but two pages of a lost copy. J53 is of course headed "Chap: 3", but J5 is rather close in subject and may once have been the second page of that chapter.

The facsimile of the Rinder copy proves faithful to the original, now examined through the courtesy of its owner. But the caret and insert on Pl. 61 are visible; and besides the "1°" on 28, two other incised numerals can be read—a "7" (of "37"?) on 39[35] and a "1" (of "100"?) on 100. Added blacking of borders prevents discovery of other numerals that may have been on the plates. Folio numbers show some erasure and change, this time in Chap. 1 (15, 16, 17 may have been 16, 17, 18; 19 is changed; 21 may have been 15 or 18; 11 was first in the 20's). Elsewhere only 97 is over erasure, probably 98.

A real surprise comes in the watermarks, for the distribution given above (note 8) proves mistaken. Blake could have delivered Chap. 2 to Linnell in 1819—its dated leaves are ff. 29, 40, 42, 47, 49 (1818) —but not the present Chap. 1, which has 1819 paper in f. 19 and 1820 paper in ff. 11 and 16, besides the expected 1818 in ff. 3, 7, 22, 24, 25. (Note that 11, 16, and 19 are among those with changed folios: all possibly inserted as replacements in a delivered copy if not in a copy held for better pages.) Dates in the last chapters are as expected, though later than those in the British Museum copy: Chap. 3—f. 59 (1819), ff. 51, 60, 66, 68, 72, 73 (1820); Chap. 4—ff. 84, 88, 91, 93 (1820).

I have now examined the separate print of Plate 51 which is in Sir Geoffrey Keynes' collection and can see that the names said to be printed in "white line" were added after the printing of the plate but not in white ink; they were cut into the thick ink on the paper with a sharp tool.

Notes