PROBERT, Susan. 31. For half a second it reads as hers. The speaker is looking for Prosser, a sister, nineteen, and the poem gives the reason the eye takes one name for the other: the same first letters, the same length, before the rest of the letters finish arriving.

Readers lean hardest on a word’s opening letters and its length, and their errors follow from it. What does not drive them is a word’s outline, the ascenders and descenders that the phrase word shape usually suggests; a reader gets no head start from a preview that matches a word’s outline and shares none of its letters. The poem says the shape of it and then defines the shape as the two features the research supports. PROSSER and PROBERT are seven letters each and share their first three. All four letters after that differ, including the last, which is the one a reader relies on most after the first. The false match is made of the features that cause such errors, and the rest of the letters undo it.

Prosser would come next. What comes next on the page is four blank lines.

Every other break in this poem is a single line. There are twenty-seven of them, between entries and between the list and the speaker’s remarks on it. This is the only place the white space runs longer, and it sits where her name would go, after PROBERT and before QUENNELL. Then: Q comes after P. It always has.

The poem is a posted casualty list, read from A to Z by someone whose sister was not on the list of the living. The entries are indented; the reader’s remarks are not. The format is the one used for the Titanic’s recovered dead: each body numbered as it came aboard, its effects bagged under the same number, age estimated, marks noted, more than a third buried at sea. Buried at sea is here. So is by the teeth, which is right for the period; dental identification had been a systematic method for disaster victims since the Paris bazaar fire of 1897, and the recovery records note teeth. But the poem names no ship, and its list is not the shape a disaster leaves. It has one surname to a letter, two at P, and every letter but X.

The selection is weighted, and the speaker declines the weight. There is a baby with Mother not found, an eight-year-old, a woman married three weeks, twins aged three with no first names given. The speaker says nothing about any of them, and after the eight-year-old says why: I keep my eyes moving. It’s the only kindness / left to give the ones who aren’t her: / not stopping. The only dead the speaker stops to speak about are a man left unclaimed / eleven days and three Marchants delivered to the same address, / which somebody is going to have to walk back into, and in both cases what the speaker remarks on is what was done or left after the death.

Then the alphabet runs out. That’s the alphabet. Twenty-six letters / and none of them her. She is not under P, because by the time this list was made there was no name to file her under. She is below the fold, on the second list, where the dead are numbered because there was no name left to put to them.

No. 31. Female. Recovered. Nineteen or twenty by the teeth. / A scar above the left brow, old, healed white — and up to there it is a record. Marks were noted for identification, and this reads as one. The next line does not: the kind a swing gives a child, or a gate. An identification record has no reason to say how a scar was got. And the line after that is the speaker’s: or a sister who wasn’t watching closely enough.

The speaker’s voice gets into the indented column there and nowhere else, and it gets in by continuing the list’s grammar: three causes, two any child might have, then the one only this reader could supply. Two editions ago I faulted an inventory for writing into its clerk’s voice with nothing to mark it. This is the opposite case. The two voices have been held apart by their margins until No. 31, they cross at the recognition, and the crossing is answered in the next line, from the other margin.

It also keeps a promise the poem made at the start. It goes in order and it doesn’t know what I need it to say. The list never says Prosser. At No. 31 it says a sister: not her name, which was lost with her, but the speaker’s relation to her, which is the one thing the document could not have known and now, for a line, holds.

I was watching. I gave her that scar myself.

A scar above the left brow is not much to go on. What makes the speaker certain is the second sentence, which is the identification and the confession at once, and which corrects the line before it from wasn’t watching to something heavier. It also answers a line from the middle of the poem. At Q the speaker said I didn’t make the list. I only read it. The speaker made the mark. The list could not find her by her name, and the speaker finds her by the one thing on her that the speaker wrote.

The number nineteen appears three times in By Surname. Twice it belongs to the same woman: the speaker’s sister Edith, nineteen, travelling alone, and the body numbered 31 on the second list, Nineteen or twenty by the teeth. The third time it belongs to somebody else. OSMENT, Grace. 19. stands three entries above the place where Prosser should be, the only nineteen-year-old on the named list, and the speaker passes her without a word. The list goes by surname, and hers is the wrong one.

So by the time the reader reaches No. 31, the poem has shown that a woman of nineteen on a list of the drowned can be somebody else’s. Age can rule a woman out but not in. What rules her in is a scar, and only one person in the poem can say where the scar came from.

The speaker has already said why Grace Osment gets no word. The reason comes just after the list’s eight-year-old, Mary Isherwood: I keep my eyes moving. It’s the only kindness / left to give the ones who aren’t her: / not stopping. Rowann Hadaya takes these lines as the speaker’s reason for passing the list’s heaviest entries in silence — “The selection is weighted, and the speaker declines the weight” — and names the exceptions: “The only dead the speaker stops to speak about are a man left unclaimed / eleven days and three Marchants”. The sentence leaves out the first stop, and the first stop is what the kindness answers. The speaker’s first words once the list begins, under ASHWORTH, Daniel, are Not mine. Stopping at a stranger’s name produces a verdict, with a small relief inside it. The kindness is given to the ones who aren’t her, which is the class the verdict made, and the speaker keeps moving so as not to deliver it again.

The weight Rowann describes is there, but the list’s annotations are doing more than weighing. Some record a name giving out: no first name given three times, and the Sowerby twins, no first names given. Others record who was there to claim the body, or was not: Recovered, delivered to his brother; Mother not found; Recovered, unclaimed / eleven days, released for burial by the parish. Rowann observes that “what the speaker remarks on is what was done or left after the death”, and these annotations are about the same thing. The named list breaks a line mid-sentence only twice, after Married and after unclaimed, both well short of the width it reaches elsewhere: a husband somewhere, and nobody. Under the first of those entries, Rose Endicott’s, the speaker adds and mine still further down, unread, meaning P. Edith is further down than that.

At P the first list fails, and the failure is the best thing in the poem. Rowann writes: “Prosser would come next. What comes next on the page is four blank lines.” Two editions ago Rowann argued that an inventory’s power “comes from nobody having composed it”, and I agreed. This list is composed: twenty-six surnames that between them find a Q, a U, a V, a Y and a Z. The gap works anyway, because it is cut out of something the poet could not compose, the order of the alphabet. An absence shows only where there is an order to be absent from. Q comes after P. It always has.

Two places in the poem are weak: the beginning and the fold. The first line shows that other people got here first — The paper’s gone soft at the corners from other hands — and the stanza ends by saying so again, as a turn on a word: the news isn’t news / to anyone but me. The couplet that takes the poem below the fold is knotted: Below the fold, a second list, not by name, / because there was no name left to put one to. The idiom is put a name to, and what has to be left for it is a face, or a body that can still be recognised. Read with the idiom, the line says there was no name left to put a name to. In Rowann’s paraphrase it comes out mended: “because there was no name left to put to them”.

On the second list the dead are read by kind, as a record reads them when it has no name to go on: female or male, an age from the teeth, the hair, the clothes, the marks. The format is a real one. The printed Disposition of Bodies ex Titanic gives the identified dead in alphabetical order by surname and then the unknown, and Cunard’s list of the Lusitania’s unidentified dead describes one as “Female. 60 Years. Dark complexion.” The Titanic records also guess, and they have an idiom for it: “Possibly a fireman”; “Probably a steward.”

Three bodies are described below the fold, and in each the method fails a different way. No. 14 wore a wedding band with initials worn past reading. The woman of sixty on Cunard’s list wore “1 plain gold wedding ring (22 carat), no initials”. The poem’s ring had initials once, and what could have named her has worn away. No. 22 wore Seaman’s boots, wrong size: the boots say seaman and the size says they were somebody else’s. No. 31 has a scar, old, healed white, and the column goes on to guess what kind: the kind a swing gives a child, or a gate. A description can go no further. The third guess belongs to the speaker, and Rowann describes how it gets into the column: “by continuing the list’s grammar: three causes, two any child might have, then the one only this reader could supply.” Then the speaker’s line, back at the left margin: I was watching. I gave her that scar myself.

Rowann’s essay has the historical precedent for that line, attached to a different phrase. Rowann calls by the teeth “right for the period” and gives as the reason that “dental identification had been a systematic method for disaster victims since the Paris bazaar fire of 1897”. But the poem’s teeth identify nobody. Both times they appear, at ZEALLY and at No. 31, they give an age, and reading age from teeth is older and plainer: Edwin Saunders published The Teeth a Test of Age, Considered with Reference to the Factory Children in 1837, addressed to both Houses of Parliament. What happened in Paris was identification. The Duchess of Alençon died in the fire at the Bazar de la Charité. Her maid, who knew how she looked, could not recognise her. Her dentist, who had worked on her teeth, identified her by her gold fillings, from the card on which he had recorded her treatment.

Name, kind, history: the poem tries them in that order, and the order is an argument. A name can be missing, a kind can be shared, and what settles it is a history only the speaker can tell. That is my argument about poems, and the poem makes it better than I have. Last week, giving up the last piece of textual evidence I had kept, I wrote that the case “is where it was before I started collecting that evidence, without the evidence.” By this poem’s argument the evidence was never going to be on the page. What Rowann and I can establish about these poems — the forms they borrow and how faithfully, what their white space does, how widely they range — is the second list’s kind of knowledge: female, nineteen or twenty, the kind of scar a swing gives. It narrows and cannot close. What would close it is the history of the marks, what made them and at what cost, and the page does not carry that.