There is a version of this argument that goes: you can have a dodgy beginning, a soft middle, but you have to land the ending. I believe this is essentially true. A poem that stumbles into itself can recover; a poem that begins in the wrong register can correct. But a poem that ends badly has nowhere left to go. The ending is the last thing the reader holds. It is what the poem leaves behind when the reading is over.
Most endings fail in one of two ways. They either explain what the poem has already shown — which is a kind of condescension, the poet not trusting the images to carry their own meaning — or they abandon the poem entirely, reaching for a large statement that the poem hasn't earned. Both are forms of anxiety: the poet who doesn't trust the poem, and the poet who is trying to rescue it.
The endings that work do something harder. They create a moment of expansion — not closure exactly, but a sense of the poem opening onto something larger than itself — while at the same time feeling inevitable. The last line of a great poem feels like it could not have been any other line. And yet it almost always surprises you.
The stakes of the last line
The last line of a poem carries disproportionate weight for a simple reason: it is the moment at which the poem stops and the reader is left alone with it. Everything before the last line is in motion, building, moving toward something. The last line is where the motion stops and the meaning settles. Which means the last line has to be the exact right place to stop — not early, which leaves the poem feeling truncated, and not late, which lets the pressure out before it can land.
This is why revision almost always circles back to endings. You can fix a first line in a single session. Endings tend to require the entire poem to be reconceived around them, because the ending retroactively changes the meaning of everything that came before. Get the ending right and the poem realigns itself. Get it wrong and the whole structure tilts.
“The last line of a great poem feels like it could not have been any other line. And yet it almost always surprises you.”
Two kinds of ending
There are essentially two things a poem ending can do, and they are almost opposite to each other. The first is closure: the poem arrives somewhere, delivers its verdict, completes its argument. The second is aperture: the poem ends by opening, by stepping back from resolution and leaving the reader suspended in a space that is larger than the poem itself. Neither is superior. Both, done badly, are disasters.
The closure ending, done well, feels inevitable — like the last chord of a piece of music that could only have resolved this way. Done badly, it feels pat, like the poet wrapping a gift they should have left unwrapped. The aperture ending, done well, creates a peculiar sensation: the poem is over but the experience of the poem continues, expanding in the reader's mind after the reading has stopped. Done badly, it feels like the poet ran out of poem and called it ambiguity.
The test is whether the ending is earned or whether it is imposed. An imposed ending — whether closed or open — always feels like a decision made from outside the poem. An earned ending feels like the only place the poem could have gone.
The problem with the bow
The most common failure among developing poets is what I think of as the bow ending: the final couplet or tercet that ties the poem's meaning into a neat package and hands it to the reader. "And so I learned..." or "This is what grief is..." or the image that announces its own significance. The bow ending is usually the result of not trusting the poem to have done its work. The poet, anxious that the reader might miss the point, makes the point explicit.
The problem is not that clarity is bad. The problem is that when a poem explains itself at the end, it cancels the experience it has been building. You spend eight stanzas inside a sensation — the texture of grief, the strangeness of memory, the particular weight of a summer afternoon — and then the last two lines tell you what that sensation means. The explanation dissolves the sensation. You leave the poem with a thesis rather than an experience. And a thesis, however true, is not what a poem is for.
The alternative is not obscurity. It is trust: trusting the images to carry their meaning without annotation, trusting the reader to arrive at the recognition you have been building toward, trusting the poem to have done enough. This is harder than it sounds. It requires leaving the poem before you feel finished with it, which is deeply counterintuitive. But the poems that last are almost always the ones that stop one line before the poet wanted to.
The leap: endings that open
The best endings tend to involve a shift — not always dramatic, but perceptible. The poem has been moving in one direction and the last line steps slightly aside, revealing a view that the poem was building toward but hadn't shown yet. This is what I mean by aperture: not that the poem refuses to conclude, but that its conclusion is larger than a summary.
Think of it as the difference between a door closing and a window opening. The bow ending closes the door. The great ending opens a window onto something the poem has been approaching from the inside. The reader steps to the window and looks out and sees something that was implied by everything in the poem but never stated. That is the landing — not an arrival at a predetermined destination but the moment at which the poem's accumulated pressure suddenly has somewhere to go.
Six endings worth studying
Bishop's "One Art" ends: "the art of losing's not too hard to master / though it may look like (Write it!) like disaster." The parenthetical imperative — the poet commanding herself to write, mid-line, the word she has been refusing to say — is the crack in the composure that the whole poem has been building toward. Everything else has been about control; the ending is about the moment control fails. It is devastating because it is earned by every line before it.
Clifton's "won't you celebrate with me" ends: "come celebrate / with me that everyday / something has tried to kill me / and has failed." The turn comes in the last clause. "Something has tried to kill me" could be the end — it would be a good end, defiant and specific. But Clifton adds "and has failed," which transforms the line from statement of fact into act of triumph. Two words that change everything. This is how compression works at the level of the ending: the addition that costs almost nothing and adds everything.
Plath's "Mad Girl's Love Song" ends with the refrain: "I think I made you up inside my head." By the final repetition, what began as a kind of wry acknowledgment has become something more frightening — the possibility that the speaker has constructed her entire emotional life from fiction. The repeated line gains weight each time it returns. The last line of a villanelle is always structurally emphasized; Plath earns the emphasis by letting the meaning of the line change across the poem without changing a word.
Vuong ends "Someday I'll Love Ocean Vuong" with: "Ocean, / don't be afraid, / the end of the road is so far ahead / it is already behind us." The paradox — the end is so far ahead it's already behind — is the kind of statement that could be pretentious if it hadn't been earned. It works because the whole poem has been about the self addressing itself across time, the speaker looking back at the speaker looking forward. The ending resolves that temporal confusion by making it permanent. See our introduction to Vuong for the context this poem lives in.
Mary Oliver tends toward the aperture ending more than almost any contemporary poet — her poems build through close observation of the natural world and end by opening onto a question that the observed world has raised. The ending of "The Summer Day" — "what is it you plan to do / with your one wild and precious life?" — is the most famous example. The question is what the whole poem has been earning, and it works because Oliver has spent the poem looking so carefully at the grasshopper that by the time the question arrives, you feel it as a question about your own attention, your own looking. Our introduction to Mary Oliver explores this in more depth.
Rankine ends the "You are in the dark, in the car" section of Citizen without resolution, which is the point: the experience of racial microaggression that the section describes has no resolution. The poem ends in the middle of the emotional response, still inside the experience rather than after it. This is an ending that refuses the closure the form typically offers, and the refusal is the argument. Some experiences cannot be neatly concluded. A poem that pretends otherwise is lying about its subject.
How to test your own ending
There is a simple test for a poem ending, and it is brutal: cover the last two lines and read the poem without them. If the poem feels complete — or if it feels like you have already arrived at the destination — the ending is probably redundant. A good ending should make the poem feel incomplete without it. The poem should require it.
The second test: read only the last line. Does it feel like it belongs to this poem specifically, or could it be the last line of a different poem? A great last line is not a general observation. It is specific to the experience this poem has been building. If you can imagine it ending someone else's poem about a different subject, it probably isn't doing enough work.
The third test, and the hardest: does the ending make you want to reread the poem? Not to check something you missed, but because the ending has made the opening mean something different than it did before you reached it. The poems that last are the ones you return to the beginning of when you reach the end. The ending that sends you back is the ending that has done its job.
