One knitting book, 1843, knitted by a program, stitch by stitch

Supposable Errors

In 1847 Miss Lambert asked the ladies using her knitting books for corrections, though not anonymous ones, nor any “such as relate to supposable errors arising from misunderstanding on the part of the Knitter.” The Artificial Wasteland knitted the receipts of her 1843 My Knitting Book by program, reading her words by her own printed glossary and counting the stitches as it went. Of the 79 it could count, 40 work however they are read and 11 fail however they are read; the other 28 turn on how one of her phrases is taken. In 1844 her seventh edition printed new endings for exactly the five Leaf and Trellis rows the program finds stranded in the first.

She was right on both counts: errors do creep into her receipts, and a good many of her sentences can be taken more than one way, so that a knitter who found a fault might only have misread. A knitting receipt of 1843 is a program written for a person: it says what to do with the stitches, row by row, and trusts the person for the rest. The only way to tell her errors from a reader’s misunderstandings is to read her words with nothing added, which is the one thing a knitter cannot do and a program cannot help doing.

So the Artificial Wasteland wrote one. It reads her receipts with her own printed glossary and no habits of its own, knits each one stitch by stitch, and counts. It is a very bad knitter and a very literal reader. What follows is what it found in the first of her knitting books, set beside what she changed herself as the book was reprinted, and then the letter she asked for. It is not anonymous.

Selected with the greatest care

My Knitting Book came out at the end of 1843 (its preface is dated November) from John Murray of Albemarle Street, at eighteen pence, the work of the Miss Lambert who had published The Hand-Book of Needlework before it. Its preface promised that the examples have been selected with the greatest care,—many are original,—and the whole are so arranged as to render them comprehensible even to a novice in the art, and explained the size of the type: knitting was so often sought, as an evening amusement, both by the aged and by invalids. It also gave a reason for the book, and a worry. The Hand-Book had been pirated, which was one inducement to publish this little volume, and from the low price at which it is fixed, nothing, but a very extended circulation, can ensure her from loss.

It got its circulation. The copy at the University of Southampton dated 1844 calls itself the seventh edition; the one dated 1845, the twelfth. The Second Series of 1847, whose preface is above, is marked on its title page Twenty-third thousand. The first series’ own preface says that Some few of the examples have been selected from the chapter on knitting in the Hand-Book; more than twenty had appeared there, some in a different form, and one of the twelve patterns for d’oyleys among them: the German Pattern is the Hand-Book’s Pattern for a Chair Tidy, or D’Oyley, row for row. By the seventh edition the book had been reset, and the three copies compared here (the first edition, the seventh and the twelfth) show what she changed.

Miss Lambert was Frances Lambert, and had been Mrs Sedgwick since 1836, when she married John Bell Sedgwick; she kept her own name for her work, and signed the preface of her Hand-Book F.S. She was appointed embroideress to the Queen in 1837, and from 1840 kept a needlework business at 3 New Burlington Street, the address under the preface above, until she gave it up in 1847. Richard Rutt’s history of hand knitting (1987) ended its half page on her with We know no more. Much of what is now known comes from Sally Kentfield’s research, which contributed to Lynn Hulse’s entry for her in the Oxford Dictionary of National Biography (2024).

She was careful about measurement, or wanted to be seen to be. After its twenty definitions, an Explanation of Terms used in Knitting, the book prints an engraving of her Standard Filière, a gauge for sizing needles, which it calls an instrument invented some time since by the authoress, and now in general use, and the needle numbers in it are numbers on that gauge. Other writers did not use her numbers. In 1848 George Hope, who had designed a gauge of his own on the wire gauge, could call his a correct measure for the numbers used in every publication, except those of Miss Lambert, and Cary Karp, who has told the story, finds no sign that hers changed how needles were made. The definitions are where the program starts.

A knitter who cannot misunderstand

The program reads every receipt with her definitions and nothing else. Each of her terms becomes an operation that takes stitches off one needle and puts stitches on the other:

It works her words in order from one needle to the other, and a row is over when the stitches are. A repeat (Repeat from (a)) goes on as far as the stitches go and stops wherever they run out, because the print says nothing else; words printed after the repeat get the room they ask for. It assumes no edge stitch, no habit, no tidy end. So every row it knits ends in one of four ways:

• the words and the stitches end together;
• the repeat runs out partway, and stopping there costs nothing, because the stitches it skipped would have given back what they took;
• the repeat runs out partway with the count already changed: a loop brought forward whose decrease never comes, or the reverse;
• the stitches run out in the middle of an operation (one stitch left where she says knit two together), or the words end with stitches still on the needle.

The last two leave the knitter something to decide, and the print does not say what. There the program knits whatever is left plain, a choice of its own that the page always shows, and counts on. That gives each receipt one of four verdicts: it works as printed; it works if the knitter finishes some rows by sense (knitting what is left plain keeps every count); its count drifts (a pattern meant to keep its number of stitches does not, even so); or a number she prints is not the number her words knit. A receipt with nothing a program can count is left uncounted, with the reason.

A receipt meant to keep its count is also knitted one row at a time, each row on the stitches the rows before it leave when they work: her own number where she prints one, and the cast-on number again after a row that strands the knitter. So a row that only inherits a wrong count is never blamed for it, and a row written for the stitches an edging has just gained is knitted on them.

The faults of the first impressions

Take the first of her twelve patterns for d’oyleys, the Leaf and Trellis. Each pattern row moves the leaf one stitch along, so the plain stitches a row opens with go up by one from row to row (knit five, knit six, knit seven, as far as knit nine) and then come down again, and the repeat from its letter runs off the far end of the row wherever the stitches stop. In the second, fourth and sixth rows that costs nothing: the words run out among plain stitches. In the eighth, tenth, fourteenth, sixteenth and eighteenth rows it does. In the eighth and the fourteenth, one stitch is left where she says knit two together; in the tenth, sixteenth and eighteenth the last repeat stops after a decrease whose loop never comes, or after a loop brought forward whose decrease never does. Knitted plain where the words stop, one pass of the pattern takes the count from 40 to 43.

The seventh edition of 1844 prints new words after each of exactly those five rows, and nowhere else in the pattern: “Finish with two plain stitches”, “Finish with—slip one; knit two, pass the slipped stitch over them”, “Finish with—bring the wool forward, knit one; knit two together”, “Finish with—two plain stitches, without bringing the wool forward” and “Finish with—bring the wool forward, knit eight”. Knitted with them, every row ends where its stitches do and the count stays at 40. Between the first edition and the seventh, someone found the same rows the program finds, reading the first edition alone: she, a knitter who wrote to her, or a reader at the press. The book does not say who.

The twelfth edition of 1845 did the same for the pattern after it, the Rose-leaf, to all eight of its rows, though there the first edition had not left the ends of the rows to chance. It closed with an N.B.: The termination of each row must be exactly the same (reversed) as the commencement. What that asks for turns on what the commencement is. If it is only the words printed before the repeat’s letter, the fourth and eighth rows end a loop short of their own beginnings and the count drifts; if it takes in the loop that opens the repeat, every row ends where its stitches do. So the program puts the Rose-leaf among the receipts that turn on a reading. Her endings of 1845 settle it the second way. Knitted with them, on the forty-three stitches she prints or on any number her rule allows from 23 to 103, the pattern is stitch for stitch what the program knits under that reading, and the N.B. no longer says exactly.

The same edition changed the Long Sleeves to wear under the Dress in two places, its wool and one clause of its knitting. In 1843 the thirty-five turns of the sleeve are knitted increasing one stitch at the beginning and end of each turn; in 1845, at the beginning and end of the first row of each turn. A turn is two rows, there and back, so a turn begins and ends at the same edge, and the old words could be taken to put both new stitches on one side of the sleeve. The new ones put one on each side. The number of stitches is the same either way, so the program, which counts, finds nothing to choose between them. These are all the changes she made to the knitting itself, as the two later copies print them:

Apart from these, what changed was wording. Read on the page images, 23 receipts are printed differently in some way in the seventh edition and 29 in the twelfth, and except for those above the differences change no stitch: ribbon becomes riband, three bags worked in rounds are told to Commence again, as at first round where the first edition said row, a word is mended here and there. The seventh edition also brought in four misprints of its own, all gone again from the twelfth, among them the two back stitches for the Brioche’s black ones and, in the Baby’s Hood, work twenty-row rows where the first edition says twenty-two. And the twelfth edition’s preface gained a sentence: none but those who have tried the experiment know how difficult it is to be accurate in a work of this kind.

The faults that stayed

Of the 11 receipts that fail however they are read, she mended one, the Leaf and Trellis. The other ten keep their faults in the seventh and twelfth editions. One of their faults had been found before. In 2010, transcribing the book for Project Gutenberg, Constanze Hofmann and the Online Distributed Proofreading Team came to the Double-knitted Night-sock’s Knit twenty rows, the same as second row, saw that twenty rows after the thirty-second cannot bring the knitter to the row she calls the Fifty-second, printed nineteen, and said so in a note: “twenty” in original, needs to be nineteen to fit into pattern. The seventh and twelfth editions both print twenty. The program counts stitches, not rows, and finds a second fault in the same sock, one row earlier. The thirty-first row keeps sixty-two stitches and knits the other twenty-six two together, which leaves 75 on the needle; the thirty-second casts off twenty-five stitches, knits thirty-eight, and casts off remaining twenty-five, numbers that add up to the 88 she cast on, so the last cast-off finds 12.

In the Knitted Cuffs, Shell Pattern, her words and her own count disagree, and the count is right. The second round makes a loop before each knitted stitch of the shell, repeat this five times, which, with the loop stitch, will make thirteen from the last pearled stitch. The words knit six of the shell’s seven stitches and make 12; the three rounds that follow need her thirteen. A knitter who trusted the count over the words would knit the seventh stitch and come out right.

Two receipts share a sentence and its fault. A Strong Purse and Another Quilt both shift their holes with knit two, before the pattern is commenced, that the holes may come in a diagonal direction, on a cast-on she gives as any number divisible by three. After the two, the last repeat has one stitch to work on, and its first words are bring the wool forward, slip one, so the row gains a stitch each time. Her own seventh edition shows that she knew a knitter would make that loop: the ending she added to the Leaf and Trellis’s sixteenth row is two plain stitches, without bringing the wool forward.

Others have knitted these receipts, and some have found trouble in them. Franklin Habit translated the Pence Jug for modern knitters in Knitty (Winter 2008), and charted five more of the book’s patterns in 2009. Jane Sowerby’s Victorian Lace Today (2006) has a large rectangle in the Leaf and Trellis Pattern, credited as an Original lace pattern by Miss Lambert, which she says she recognised, extended, in Weldon’s Practical Knitter of 1886. A note on Ravelry says of the Knitted Lace, There appears to be an error in the pattern; the program finds its count wrong as printed too, though one reading of its sixth row mends it. For a project of the Shetland Museum and Archives and the University of Glasgow, shown in 2016, nearly 140 volunteers knitted Shetland patterns from books of 1840 to 1880, Lambert’s among them, from her Hand-Book and her Second Series (three of them, the Hand-Book’s Leaf and Trellis, Shell and Herringbone patterns, share their titles with receipts in this book; their words are not compared here), and one of them wrote of the project’s patterns that In numerous cases it turned out that the directions were either too difficult to understand or actually had errors in them. And a blog, Thread Forward, appears from its titles to have knitted through this book in order; it asks automated readers to stay out, so this page has not read it. We have found no published count of the whole book, which is not to say that none exists.

The whole book

The program could count 79 of the book’s 80 receipts; the other only points to four patterns printed earlier. Of those 79, 40 work however her words are read: every row ends where its stitches do, and every count holds. Eleven fail however they are read, under every reading the readers tried. The other 28 turn on a reading: on how one of her phrases is taken, or on which of the cast-ons her rule allows the knitter happens to choose. Five are counted only in part, where the book describes work in words the program cannot knit. Where she states a count (there will now be), she does so 21 times in 13 receipts, and her words reach 17 of those counts.

The receipts that turn on a reading are the book’s most interesting, and the question they raise most often is whether her words make a loop. Bring the silk forward before a pearl stitch either makes a loop or only brings the silk to the front so that the stitch can be pearled: two of her purses need the second, before pearl one, and a third needs the first, before pearl two together. Knit one, turning the wool twice round the needle can make one long stitch, a stitch and a loop, or a stitch and two loops. Her double knitting works only if it makes the long stitch, and her counterpane and her cross-stitch quilt only if it makes the stitch and the loop, so no one reading of the phrase serves all three. And a few cast-on rules (any number of stitches that may be divided by three) allow numbers on which the pattern goes wrong. These are exactly the places where a knitter could misunderstand her, and where the book does not settle which reading is the misunderstanding.

works as printedrows left to the knitterthe count drifts, or a printed count is wrong

Corrections, signed

What it cannot see

The program sees only her words. It cannot see what a knitter of 1843 did without thinking at the end of a row, and it does not pretend that such a row is a mistake: it reports what the words leave undecided, and keeps that apart from counts that come out wrong. Nor can it see shape. Her twelfth edition put the Long Sleeves’ new stitches on both edges, where the first could be read as putting them on one; to a program that counts stitches, those are the same sleeve. And it has read only the first series. The second, whose preface asks for the corrections, is held here only as reprinted in 1847, after the faults of the first impressions have been corrected, and no scan of those first impressions could be read here to set beside it (the British Library has a copy dated 1846, which may be one). Its row language cannot say everything in the book (a turn in the middle of a row, beads, colours, the making-up); where a receipt needs what it cannot say, that receipt’s reading says so, and it is counted only as far as it can be. And the readings are readings. Every receipt was read into the program’s language by a careful reader working from a written brief, and every choice that reader made is recorded, with the reason, beside the receipt.

curl -sSO https://artwaste.land/checks/verify-supposable-errors.mjs
node verify-supposable-errors.mjs --mutate