Thin Places

I. Prologue: A Correction

A companion essay of mine, written for the salon and pitched to a room of adult readers, opens with an appeal to an old Irish phrase. A thin place, I say there, is a spot on the earth where the membrane between this world and the next has worn through; my grandmother’s people knew the concept even if they lacked the term. It is a good opening. It is also, in one respect, false, and since the present essay means to be scholarly rather than hortatory, the falsehood should be dispatched in the first paragraph rather than smuggled through to the last.

The phrase “thin place,” used as a named category of Celtic or Celtic-Christian sacred geography, does not appear to be ancient. No pre-twentieth-century textual attestation of it has been located in the Irish folklore corpus or in histories of Celtic Christianity; the term is rare in print before the twenty-first century and becomes a fixed idiom only around its turn, propelled by popular spirituality writing, tourism copy, and eventually mass-market travel journalism.¹ The earliest identified print use, in Evelyn Underhill’s Collected Papers of 1946, refers to Iona — and is itself a transcription of a single Scotsman’s offhand remark. A friend of Underhill’s, about to travel there, was told by her gardener: “Iona is a very thin place.” Asked what he meant, he answered, “There is not much between Iona and the Lord.”²

That is not the transmission of doctrine. That is a gardener being unimprovably eloquent on a Tuesday, and a well-read Anglican writing it down because she recognized that he had said something she could not have said better. The phrase’s subsequent career belongs to the broader phenomenon that Celtic-studies scholars have spent thirty years documenting: the modern construction of “Celtic Christianity” as an invented tradition, assembled largely by writers without access to the Celtic languages and projecting contemporary spiritual, ecological, and anti-institutional preoccupations backward onto a romanticized early medieval past. Donald Meek, a native Gaelic speaker and Professor of Celtic at Aberdeen, made the case at book length; Ian Bradley made it independently.³

I want to be precise about what this correction does and does not cost me, because the distinction is the methodological hinge of everything that follows.

What it costs: the packaging. I may not present “thin places” as inherited lore, and I will not.

What survives intact: the belief complex the phrase was later draped over. Irish and Scottish traditions of death omens, second sight, the fetch or wraith, the phantom death-coach, and the bean sídhe are genuinely old and are documented to a high evidentiary standard by Patricia Lysaght, whose study of the supernatural death-messenger draws on the University College Dublin Folklore Archives, the Ulster Folk and Transport Museum holdings, questionnaire fieldwork, and her own interviews, and whose conclusion is that these beliefs “were fully developed in Gaelic Ireland,” with negligible contamination from printed sources.⁴ Her published fieldwork contains testimony of exactly the kind my grandmother’s household would have recognized. One informant describes a woman watching her husband walk up the garden path and waiting for him to come in:

she saw him coming up the garden and she was waiting for him to come in and he didn’t come… that night when she went to get his bed ready for him word came that he had been found dead where he’d gone off for a walk across the roads. He was found dead: he died, and he came back to her to let her know that he was going.

So the argument of this essay does not require an ancient phrase. It requires only that a particular structure of experience — the dead as proximate rather than departed, the past as present tense, the border between states as a membrane rather than a wall — be old, widespread, and documented. It is all three. What I add is a proposal about where such a structure of experience comes from in the individual case, and what happens when it is handed to a person who writes.

II. The Construct

In 1989 Ernest Hartmann, a Boston psychiatrist who had spent two decades studying chronic nightmare sufferers, published the first report of an instrument he called the Boundary Questionnaire. He had noticed something his clinical population kept doing: they were not simply people with bad dreams. They were people whose mental categories were, across the board, less firmly separated than other people’s. He built a self-report measure to test whether that generalization would hold, administered it to more than a thousand respondents, and found that nightmare sufferers and art students scored at one end of the resulting distribution while a comparison group of naval officers scored at the other. Thinness correlated with dream-recall frequency at r = .40.⁵

The mature instrument, described in the original validation study and in Hartmann’s later review of the research program, runs to 138 items scored 0–4 and distributed across twelve content categories: sleep, dreams, and waking states; unusual experiences; thoughts, feelings, and moods; childhood, adolescence, and adulthood; interpersonal distance; sensitivity; neatness and order; clear edges and lines; opinions about children and adults; opinions about organizational authority; opinions about groups and nations; and opinions about abstract concepts such as beauty and truth. The summary score, “SumBound,” excludes seven items and returns an internal-consistency alpha of 0.925. In the original principal-components analysis, all 138 items loaded positively on the first component, supporting the existence of a single general boundary dimension; a twelve-factor solution then separated the theoretically thin groups from the theoretically thick one.⁶

Hartmann’s own summary of the correlational picture is worth stating exactly, because it is stronger than the usual run of personality findings offered to literary audiences. Dream recall correlated with boundary thinness at r = .58 in a sample of forty-two members of the Association for the Study of Dreams and at r = .40 in a general sample of six hundred; Hartmann described thin boundaries as, at that time, the only well-replicated correlate of dream recall in the literature. Boundary thinness correlated with the Tellegen Absorption Scale at r = .67 and with Openness to Experience at r = .73. Nightmare sufferers and lucid dreamers both scored thin, differing chiefly on a factor Hartmann labelled Overinvolvement. And he added, in a sentence I regard as the single most consequential line in the entire literature for anyone interested in art: “Thus, we all have thinner boundaries when we are dreaming.”⁷

The construct has since been subjected to independent psychometric replication. Working from a six-subscale short form, a 2022 study in a Russian-speaking sample (total N = 387) derived a thirty-three-item version across five factors — Unusual Experience, Need for Order, Childlikeness, Sensibility, and Trust — with test–retest reliability of r = 0.86 and a correlation of r = 0.66 with Hartmann’s own eighteen-item short form. Its Big Five correlations were more modest than Hartmann’s: Openness r = .326, Neuroticism r = .262, Conscientiousness r = −.486.⁸

That same study reports a finding I will return to twice more in this essay, because it is the empirical fact with the most direct literary consequence. Boundaries thicken with age, significantly and substantially: F = 52.18, p < .001, partial η² = .12.⁸

Hold that number. A poet asserted it first, and we will get to him.

III. What the Construct Is Made Of

A trait measure is only as interesting as its nomological network — the web of other constructs it hangs among. Boundary thinness sits inside a well-mapped one.

Its nearest neighbour is absorption, the disposition Auke Tellegen and Gilbert Atkinson introduced in 1974: a capacity for episodes of total attention that fully engage one’s perceptual, imaginative, and ideational resources, producing a heightened sense of the reality of the attentional object, imperviousness to distraction, and an altered sense of reality generally.⁹ Anyone who has lost three hours to a paragraph will recognize the description as occupational.

Adjacent to absorption lies fantasy proneness, introduced by Wilson and Barber in 1983 to describe a small subgroup — they estimated two to four per cent of the population — who spend much of waking life inside vivid internal fantasy and report quasi-hallucinatory sensory experience.¹⁰ A 2021 meta-analysis of 139 samples totalling N = 24,007 established the trait’s correlational profile with unusual precision: absorption r = .64, hallucinatory experiences r = .64, magical ideation r = .59, perceptual aberration r = .58, dissociative symptoms r = .52, excessive daydreaming r = .51.¹¹

That meta-analysis also settles, or at least badly weakens, a hypothesis that literary criticism finds irresistible. If a writer possesses unusual imaginative permeability, the assumption goes, some early wound must have opened it. The data do not support the inference: global childhood trauma correlated with fantasy proneness at only r = .26, and violent trauma at r = .10. The authors conclude that a history of childhood trauma “appears to be insufficient to fully account for the expression of fantasy proneness in adulthood.”¹¹

This matters for how one reads a body of work. The trauma-etiology reading of a permeable imagination is the most available critical move and one of the least empirically supported. Something in the constitution is prior.

IV. The Case Against My Own Argument

A scholarly essay that recruits an empirical literature owes its reader the parts of that literature which cut against it. Four bodies of evidence do.

The variance problem

The most important corrective comes from Michael Schredl and colleagues, who in 2003 fitted a structural equation model to dream-recall data from N = 444 participants. Four factors were significantly related to dream-recall frequency: personality (openness, thin boundaries, absorption), creativity, nocturnal awakenings, and attitude toward dreams. Together they explained 8.4 per cent of the total variance. Each single factor explained less than three per cent. Schredl notes explicitly that this is considerably lower than earlier single-factor studies had implied, and raises the possibility that expectancy effects inflated the earlier estimates.¹²

Eight per cent. That is the honest size of the thing. Boundary thinness is a real, replicable, reliably measured trait that accounts for a small fraction of the variance in even its flagship outcome. Any essay claiming that this trait explains a literary practice is overreaching, and I want to name the overreach before a reader does. What the trait can support is weaker and still worth having: a vocabulary, a mechanism, and a testable family of predictions about where a certain kind of writing comes from.

The unflattering alternative

A second corrective concerns what else predicts thinness. A pair of studies published in 2003 found that thin boundaries were most strongly predicted not by artistic sensibility but by insecure attachment and interpersonal dependency; in the replication, openness and insecure attachment jointly accounted for thirty-one per cent of the variance in boundary score.¹³

This is the reading I would least like to be true and cannot dismiss. A writer who finds that the past will not stay past, who cannot maintain a firm partition between himself and another person, who returns compulsively to the same grandmother in the same kitchen — that writer may be describing a poetics, or he may be describing an attachment style with a literary career attached to it. The two accounts are not mutually exclusive, and the second is less flattering, better supported in the specific studies at issue, and therefore deserves a paragraph rather than a footnote.

The artists problem

Third, the literature on whether artists actually score thin is genuinely mixed. Hartmann’s review reports that art and music students scored significantly thinner than controls, but that established professional artists scored in the normal range rather than the extreme-thin range.⁷ Against that, a 2005 doctoral study of practising artists found writers significantly thinner than student norms (t(22) = 5.76, p < .001) and visual artists likewise (t(36) = 4.62, p < .001), while musicians did not differ from norms at all (t(30) = 1.37, p = .18) — with writers scoring significantly thinner than musicians, and the largest subscale effects falling on Unusual Experiences and Childlikeness.¹⁴

Two caveats are required. The 2005 study is an unpublished thesis rather than a peer-reviewed article, and the discrepancy with the earlier finding may reflect differences in artistic domain, in sample composition, or in how “established” was operationalized. What one can say is that the domain specificity is suggestive: on this evidence writers, not artists generally, are the thin-scoring group, and musicians — whose art is the least referential and the least dependent on the retention of biographical particulars — are not.

The retrospective coherence problem

Fourth and last, the methodological objection that applies to this essay specifically rather than to the literature. I have not been administered the Boundary Questionnaire. No instrument has been applied to me, no score exists, and a group-level trait measure cannot in any case adjudicate an individual case. What I am doing here is using a psychometric vocabulary to describe patterns I had already identified in my own work by other means — which is precisely the procedure most likely to generate a flattering and unfalsifiable account.

I do not think this vitiates the exercise, but it constrains what the exercise can claim. Read what follows as a hypothesis about a mechanism, framed so that a reader could disagree, rather than as a diagnosis that settles anything.

V. The Literary Genealogy: The Trait Had Other Names

Here is the claim I think is actually novel, and it is a claim about intellectual history rather than about psychology. Hartmann’s construct was not new in 1989. It had been described, with considerable precision, by poets working two centuries earlier, who lacked the instrument but not the observation. What the Boundary Questionnaire measures, Romantic poetics had already theorized.

Keats: the self/other boundary

In December 1817 Keats wrote to his brothers that he had realized “what quality went to form a Man of Achievement especially in Literature and which Shakespeare posessed so enormously — I mean Negative Capability, that is when man is capable of being in uncertainties, Mysteries, doubts, without any irritable reaching after fact and reason.”¹⁵ Read that as a personality description rather than an aesthetic doctrine and it becomes a near-paraphrase of the thin end of Hartmann’s continuum: a tolerance for unresolved category boundaries, and a corresponding absence of the drive to firm them up.

Ten months later Keats went further, in a letter that is the single best pre-psychometric account of a permeable self:

As to the poetical Character itself… it is not itself— it has no self— it is every thing and nothing — It has no character… What shocks the virtuous philosopher, delights the camelion Poet… A Poet is the most unpoetical of any thing in existence; because he has no Identity.¹⁶

“It has no self.” Hartmann’s interpersonal-boundary items ask, in flat questionnaire English, whether the respondent experiences a firm sense of where he ends and another person begins. Keats answers the item in 1818 and answers it at the extreme. And note the contrast he draws: his own chameleon character is defined against “the Wordsworthian or egotistical sublime; which is a thing per se and stands alone” — which is to say, against a thicker interpersonal boundary in a contemporary. Keats is not merely describing permeability. He is describing individual differences in permeability, and locating himself and a rival at opposite ends. That is a trait psychology in everything but name.

Coleridge: the mechanism

Coleridge supplies the mechanism. In Chapter XIII of the Biographia Literaria he coins a word for the imagination’s unifying power — “esemplastic” — and explains the coinage: “I constructed it myself from the Greek words, eis en plattein, to shape into one; because, having to convey a new sense, I thought that a new term would both aid the recollection of my meaning.”¹⁷ To shape into one. A faculty defined by its capacity to dissolve the partitions between separate things.

His distinction between primary and secondary imagination then describes the operation in terms a dream researcher would recognize. The secondary imagination, he writes, “dissolves, diffuses, dissipates, in order to recreate: or where this process is rendered impossible, yet still at all events it struggles to idealize and to unify.”¹⁷ Dissolution followed by recombination is, almost exactly, the process later proposed for what REM sleep does to memory traces — the activation, recombination, and recontextualization of memory elements. Coleridge names as the highest poetic faculty the operation the sleeping brain performs nightly and involuntarily.

Wordsworth: the developmental prediction

And Wordsworth, in the Intimations Ode, makes the developmental claim — the one that turns out to be empirically testable.

Our birth is but a sleep and a forgetting:
The Soul that rises with us, our life’s Star,
Hath had elsewhere its setting,
And cometh from afar:
Not in entire forgetfulness,
And not in utter nakedness,
But trailing clouds of glory do we come
From God, who is our home:
Heaven lies about us in our infancy!

And the question the poem is built to ask: “Whither is fled the visionary gleam? / Where is it now, the glory and the dream?”¹⁸

For two centuries this has been read as consolation literature, a lyric about the melancholy of maturing. Read it instead as an empirical assertion and it says something specific: that intensity of inward image is maximal in early childhood, universal at that age, and lost progressively with development in most but not all people.

That assertion has been tested. Hartmann asked thick- and thin-boundaried adults for their earliest remembered dream; both groups located it around age six or seven, and — uniquely among his comparisons — the two groups did not differ in the rated intensity of those earliest dreams. His inference was that everyone begins with that intensity and that in most people it attenuates as boundaries thicken, while in some it does not.⁷ And the thickening itself has now been measured directly: F = 52.18, p < .001, partial η² = .12, boundaries growing thicker across the lifespan.⁸

Wordsworth’s ode is a correct developmental hypothesis about a personality trait, stated in 1807, confirmed in outline by questionnaire studies in 1991 and 2022. The visionary gleam does not depart. It departs from most. And the writer, on this account, is not someone who acquired a faculty but someone in whom the ordinary hardening failed to complete — which reframes literary vocation from an achievement into a retention.

Emerson: the metaphysical extension

Emerson completes the genealogy by taking the permeability of self and other and elevating it from psychology to metaphysics. The Over-Soul is defined as “that Unity, that Over-soul, within which every man’s particular being is contained and made one with all other; that common heart of which all sincere conversation is the worship.”¹⁹

I have written for years out of that sentence and would not now retract a word of it. But the scholarly point is a different one. If a mind is constituted such that the partition between self and other is thin as a matter of ordinary perception, then Emerson’s doctrine will not present itself to that mind as a doctrine requiring assent. It will present itself as a description. The metaphysics is what the phenomenology looks like once it is written down by someone who trusts it.

VI. Time, and Why Recurrence Is Not Repetition

The strongest version of my thesis concerns the temporal boundary, and it is here that the philosophical literature does more work than the psychological.

Bergson is the indispensable source. In the Stanford Encyclopedia’s summary, duration for Bergson is “the prolongation of the past into the present,” and the account of consciousness he gives in Time and Free Will describes “several conscious states” that “are organized into a whole, permeate one another, [and] gradually gain a richer content” — a “mutual penetration, an interconnexion and organization of elements, each one of which represents the whole.”²⁰

Permeate. Mutual penetration. Bergson is describing, as the fundamental structure of lived time, exactly what Hartmann measures as an individual difference. And crucially, Bergson supplies the distinction on which my entire defence of recurrence depends: “duration implies a conservation of the past… memory conserves the past and this conservation does not imply that one experiences the same (re-cognition), but difference.”²⁰

Conservation that yields difference rather than sameness. That single clause separates a poetics of recurrence from a symptom.

Proust furnishes the phenomenological instance. In the madeleine episode the past does not arrive as a representation labelled “past”; it arrives as a present sensation whose recognition and whose retrieval are effectively simultaneous. In the most careful cognitive-literary treatment of the passage, retrieval of associations “only minimally precedes, or occurs simultaneously with, identification.”²¹ There is no interval in which the narrator knows he is remembering. That is what a thin temporal boundary produces in prose.

Hegel supplies the logic, though here I must correct a formulation I have used loosely in the past and which appears in my own earlier essays. I have described Aufhebung as a triple operation — cancelling, preserving, and lifting up. The Stanford Encyclopedia, checked directly, attributes to Hegel a double sense: aufheben “means both to cancel (or negate) and to preserve at the same time.”²² The third sense, the raising or lifting, is widely attested in the wider literature on the term but should not be attributed to that source, and I withdraw the confident triple formulation in favour of the double one, which is sufficient for my purposes anyway. Cancelled and preserved at once is exactly the condition of a stolen chalice that is later returned: the theft is not erased by the restitution, and it is not left standing either.

The poems as evidence

With that apparatus in place, three passages from my own published verse can be read for what they are doing formally rather than for what they mean.

The close of “Dark City,” a poem for a grandfather I met once, at two years old, in a dark hallway:

I press my palm against the homestead,
just as you touched my head so many years ago.
I see you move from light into darkness and beyond.

The conjunction is doing the philosophical work. Just as is not deployed here as simile; there is no comparison being drawn between two like things. It asserts identity between a hand on Irish stone in the present and a hand on a child’s head six decades earlier — one gesture, occurring twice, or occurring once in a duration that does not distinguish. And the final verb is present indicative: I see you move. Not I imagine, not I remember. The grammar refuses the pastness of the past, which is precisely Bergson’s prolongation rather than Proust’s retrieval.

“The Crosswalk” performs the collapse with a harsher instrument. The poem moves from a present-tense street scene to a dead father to a dispute about a memory the father had attributed to the wrong son, and then:

Someone crossed the room and paused to take a picture.
It was on the table by your coffin. Your hands on mine.
Proof that we had closeness for a moment,
and that is enough.

Two sentences, two times, no transition and no grammatical marking of the shift. The photograph exists simultaneously in the moment of its taking and on the table beside the coffin; the poem declines to specify which sentence is the flashback. And the closing verdict, that is enough, is delivered in the present tense about something the poem has just refused to consign to the past.

“Honeymoon” runs the mechanism forward rather than back. A son kneels to work ointment into his mother’s cracked heel, sees her honeymoon photograph on the end table, and registers that his parents in the picture could be his children. Then:

Someday later, I’ll hold the photograph
and remember this night —
the rain and the darkening eve.
I will see my mother waving to me.
Maybe I will feel the cut of pain, a sigh of love,
or nothing at all.

The poem steps into its own future and regards its own present as already-memory. What is remarkable is not the prolepsis, which is a common enough device, but the honesty of the final line: the possibility that the future grief will register as nothing at all is admitted rather than warded off. A mind for which past and future interpenetrate the present does not get to choose which of them arrives, or in what condition.

VII. Fusion: The Sentence in Which Thought Arrives Wearing Its Feeling

One category on Hartmann’s questionnaire deserves separate treatment, because it bears on style rather than on structure. Among the thin profile’s defining features is a reduced separation between thought and feeling: such a respondent will report, if pressed, that he cannot readily imagine a thought without an accompanying feeling.⁷ Stated as a questionnaire item this sounds trivial. Stated as a compositional fact it is close to what we mean, imprecisely, by voice.

The relevant literary theory is again Coleridge’s. If the secondary imagination “dissolves, diffuses, dissipates, in order to recreate,” then its characteristic product is the figure in which two orders of thing — a physical observation and an emotional state — have been shaped into one and cannot be separated again by analysis.¹⁷ What distinguishes such a figure from ordinary metaphor is that it does not present itself as a comparison. Nothing is being likened to anything. The feeling has arrived already wearing a physical form.

“Brother” supplies the clearest instance I can offer. The poem sets nine lines of memory — two boys, a fire under a bridge — against five lines of consequence, and the hinge is a failure of the sleep/waking boundary:

In the blackness of my bedroom, sometimes a fire
blazes and I see our pink faces before the flames.

The fire is thirty years extinguished. It blazes in the bedroom regardless, and the poem does not label this as imagination; it reports the event in the flat register one uses for a draft or a noise in the pipes, and the adverb sometimes does the work of making it habitual rather than visionary. Then the fused sentence:

Your arm crosses my cold neck and hugs my shoulder,
and I dream we never stepped outside our hallowed space.
The snow was so cold and the streetlights too strong.

That final line is a statement about weather that is entirely a statement about grief, and I know of no instrument that separates the two. The streetlights too strong is an observation about candlepower carrying the loss of a brother. Asked to render the same feeling abstractly, I could not do it — not from reticence, but because the feeling did not present itself in an abstract form to begin with.

“Piano” pushes the fusion to a limit and is worth reading for its technique. A man dismantles a grand piano with an axe in a December yard; the poem inventories the wreckage in a carpenter’s vocabulary — resin-impregnated paper, dovetail joints, wooden ribs, polished mahogany — and then converts the entire inventory in a single move:

If I had left the window to join you in the yard, I would have seen wetness
on your cheeks, glistening strings from the soundboard of your broken soul.

The strings are literal, lying in the leaves. They are simultaneously the man’s interior made visible. The line declines to choose, and the refusal is not coyness: a perception of this kind does not encounter the two as separate items requiring a choice. This is the mechanism behind a claim I have made elsewhere in plainer language — that the ordinary is holy. The proposition is not that a writer elects to sanctify a cracked object. It is that the object arrives already carrying its freight, and the labour of composition is confined to finding language adequate to a significance that was present before the writer got there.

VIII. The Objection From Pathology

There is a serious counter-reading of everything above, and it has a distinguished pedigree. It says: what you are calling a poetics of recurrence is repetition compulsion with better prose.

Freud’s formulation, in the Hubback translation, is exact and damning. The patient “is obliged rather to repeat as a current experience what is repressed, instead of, as the physician would prefer to see him do, recollecting it as a fragment of the past.”²³ Repeating as current experience rather than recollecting as a fragment of the past — that is a description of failed temporal boundary maintenance, and it is not offered as a compliment. Every claim I have made about the present-tense grandmother could be restated in this vocabulary without altering a single observation, only its valence.

Cathy Caruth extends the point into literary theory. Trauma, she argues, is “not locatable in the simple violent or original event in an individual’s past, but rather in the way that its very unassimilated nature — the way it was precisely not known in the first instance — returns to haunt the survivor later on”; the wound of the mind is “a breach in the mind’s experience of time.”²⁴ A breach in the experience of time is, near enough, what I have been celebrating.

The dream-science literature supplies the same structure in neurocognitive terms. Domhoff’s continuity principle, paired with what he calls a repetition principle, holds that dream content both reflects waking conceptions and repeats across years, tied to amygdala-mediated vigilance and fear systems; he cites one journal of 904 dreams collected across fifty years in which six recurring themes accounted for seventy-six per cent of the material.²⁵ Levin and Nielsen’s model attributes disturbed dreaming to dysfunction in an affect network comprising the amygdala, medial prefrontal cortex, hippocampus, and anterior cingulate, modulated by situational “affect load” and dispositional “affect distress”; on their account normal dreaming performs a fear-extinction function and nightmares mark its failure.²⁶ Revonsuo’s threat-simulation theory holds that dreaming exists to rehearse threat perception and avoidance in the first place.²⁷

Put the three together and a deflationary account of my whole body of work is readily available. A man with an overactive affect network rehearses the same six threats for fifty years; the rehearsal fails to extinguish anything; he writes it down and calls it the Oversoul.

I want to concede most of the mechanism and contest the conclusion, and the ground on which I contest it is Bergson’s clause.

Repetition compulsion, in Freud’s account, repeats the same. Its signature is invariance; it is a mechanism precisely because it does not learn. Bergsonian conservation, by contrast, “does not imply that one experiences the same (re-cognition), but difference.”²⁰ Hegelian sublation cancels and preserves simultaneously, producing a third thing that neither prior term could have been alone.²²

The empirical question, then, is not whether material returns in a body of work — it does, in mine, conspicuously — but whether it returns transformed or unchanged. And that is a question a critic can actually adjudicate by reading, which is why I regard it as the right question. When a gold chalice is stolen from Mission Church in one story and returned to it in another, the second appearance is not a re-run of the first; it is the first event cancelled and preserved at once, which is the technical definition of the thing Hegel named. When a blindfolded woman on a globe appears first as a print above a psychiatric patient’s bed and later as a person standing on a hospital lawn, the image has been metabolized, not repeated. When a grandmother recurs across four stories and two novels as a different woman each time — thief, saint, liar, intercessor, and in one book all four inside a page — the recurrence is generative.

Repetition compulsion produces the same dream for fifty years. A poetics of recurrence produces a different book from the same material every time it returns. I cannot prove from the inside which of those I have been doing. I can specify the criterion by which a reader might decide, and I would rather be judged against it than exempted from it.

IX. The Threshold: Composition at the Edge of Sleep

If the trait argument has an empirical frontier, it is the sleep-onset boundary, where the literature has recently become genuinely exciting.

Andreas Mavromatis’s monograph on hypnagogia — the transitional state between waking and sleep — identifies as its central feature what he terms a loosening of ego boundaries, and his underlying doctoral work ties those phenomena to subcortical activity.²⁸ Hartmann’s thin profile includes, as ordinary questionnaire items, exactly the states Mavromatis studies: being half-awake and half-asleep, entering reverie easily, having difficulty leaving it.

What is new is that the state has been shown to do measurable cognitive work. In a 2021 study of 103 participants asked to discover a hidden mathematical rule, spending as little as fifteen seconds in N1 — the lightest, hypnagogic stage of sleep — tripled the rate of insight: 83 per cent solved the rule after touching N1, against 30 per cent of those who remained awake and 14.29 per cent of those who descended into N2. The benefit vanished once sleep deepened, and the productive window was characterized electrophysiologically by moderate alpha power together with low delta.²⁹ The authors call sleep onset a creative sweet spot; a writer would call it the hour before the alarm.

REM contributes something related but distinct. In a nap-paradigm study using the Remote Associates Test, a REM-containing nap improved performance on primed items by nearly forty per cent over morning baseline, while mere passage of time did not, and no condition improved performance on unprimed items — indicating that REM does not manufacture insight from nothing but recombines material already present.³⁰ The broader review literature situates this within a two-stage model of sleep-dependent consolidation and credits REM specifically with facilitating access to weak associative links and flexible recombination.³¹

The temptation at this point is to reach for Coleridge, and the temptation should be resisted, or at least heavily qualified. The 1816 preface to “Kubla Khan” claims composition of two or three hundred lines during three hours of profound sleep, with the whole distinctly recollected on waking and then irrecoverably lost after the interruption of the person on business from Porlock.³² It is the most quoted composition anecdote in English and it is not reliable evidence. Rosemary Ashton calls it “the most famous, but probably not the most accurate, preface in literary history”; Richard Holmes calls the account one of the most disputed ever written; Charles Rzepka reads the preface not as documentary report but as an analogue of poetic creation. The Crewe manuscript, moreover, describes not a dream but “a sort of Reverie brought on by two grains of Opium,” which is a materially different claim — and closer, as it happens, to hypnagogia than to REM. Elisabeth Schneider’s monograph argues on metrical and literary-historical grounds that the poem bears the marks of deliberate composition, and states flatly: “I have also never shared the belief that it is a product of the unconscious.”³³

The methodological lesson is one I would press on any graduate student tempted by this material. A poetics should not be founded on a composition legend, because composition legends are authorial self-fashioning and are the least trustworthy genre in the archive. What can be used instead is form — the structural decisions a text makes about thresholds, which are visible on the page and require no testimony about how they got there.

By that standard “Copacetic” is better evidence than any anecdote I could offer about my own working hours. The poem begins as a word game, summons a father framed in a doorway asking his sons whether everything is copacetic before a trip, moves to a hospital where the father sleeps and intermittently asks the time, and then reverses the tableau: the son now stands in the doorway and uses the father’s word.

“Everything copacetic?” I say.
He turns and looks.
“It was tough,” he says.

The father’s answer had been spoken decades earlier about Korea. It returns here attached to something else, or perhaps to the same thing; the poem does not disambiguate, and the refusal is the point. Then the ending:

I guide him to the bed and sleep in the chair beside him.
When I wake, I find that he has gone.

The death of the father is delivered across the boundary of the speaker’s own sleep. He does not witness the event; he wakes into it. That is a formal choice, available to inspection, and it is the choice a text makes when the mind composing it treats sleeping and waking as adjacent rooms rather than separate jurisdictions. No claim about how the poem was drafted is required.

X. The Central Image as a Unit of Composition

Hartmann’s late work isolated a construct he called the Central Image: the single dominant picture around which a dream organizes itself, the element that survives when the dream’s narrative has evaporated. He developed an intensity scoring for it and reported that thin-boundaried subjects produce more intense Central Images, both in recent dreams and in the dreams that stand out across a lifetime.⁷

The construct has an obvious literary cognate, and the convergence is more than terminological. Dream-content analysis in the Hall and Van de Castle tradition, built from more than ten thousand reports, established that dreams are amenable to systematic content coding at all — that they have stable, quantifiable structure rather than being noise.³⁴ Domhoff’s continuity principle established that the content so coded is continuous with waking concerns, and his more recent synthesis positions dreaming as an intensified form of mind-wandering rather than a separate mental kingdom.³⁵ Put those together and the dreaming mind is doing what a lyric poem does: selecting from waking preoccupation and organizing the selection around a dominant image.

The final page of “Blindfolded” can be read as a Central Image constructed by condensation in precisely the dream sense. Leaving the hospital, the boy and his mother see the dead grandfather on the grass:

My grandfather was sitting on the grass beneath a tree. He smiled and waved to us. One star shone in the twilit sky.

The grandmother, who sees nothing, asks what they are looking at; the mother answers, “Hope”; the boy asks how anyone can see hope; and the grandmother replies, “Hope is sitting right beside me” — meaning her daughter, whose name it is. Then the boy opens the mother’s sketchbook:

A blindfolded woman wearing a green gown sat atop a light brown globe, her head bent to the left as she played a lyre with a single string. In the background, one star shone in the gray-blue sky. Printed underneath the reproduction was “Hope, 1886, George Frederic Watts.”

Count the layers held in one frame: a dead man visible to two of three witnesses; a single star above a lawn; the word hope operating simultaneously as common noun, proper name, and painting title; and inside the painting, the same single star in the same twilit sky above a blindfolded woman playing one remaining string. The star inside the artwork and the star above the hospital are the same star. The scene is not built by argument; it is built by folding one image inside another until they cannot be separated, which is condensation in the technical sense and the compositional signature of the dreaming mind.

This also furnishes the honest answer to why the same objects recur across my books. A Central Image of sufficient intensity is not consumed by a single use. The star above the hospital lawn appears above another lawn in Give Them Unquiet Dreams; Yeats’s Innisfree — whose own closing lines locate an absent lake “in the deep heart’s core” while the speaker stands on grey pavement, a compact instance of the same permeability³⁶ — arrives first as a poem pasted into an album, then as a boat within the album, then as a scarf caught on a branch. A mind that will not file the past as concluded will not file an image as spent.

XI. Self, Other, and the Limit Case

I have argued that reduced separation between self and other is, in a novelist, the whole apparatus of characterization — the technical capacity Keats named when he said the poetical character has no self. The clearest evidence I can offer is not a sympathetic character but an unsympathetic one, because the inverse case is where the argument is falsifiable.

Consider Nanna in “Blindfolded,” who fakes her grandson’s leukemia to move through a healing line, cuts ahead of a humpbacked woman, coaches the boy to collapse, steals a chalice while he moans on the floor, and lies to a nurse about the wine — and who also says: “I often doubt the possibility of miracles, but then I find myself thinking that every moment is miraculous.” A judging narrator must choose between the thief and the theologian. The story does not experience the choice as pending. Both are simply true of her at once, which is possible in the prose because the author is not stationed far enough outside her to arbitrate.

Now the limit case. Molly Bonamici, narrating “A Perfect Hit,” murders her roommate with cocaine baked into her grandmother’s cookies, performs competent CPR on the body, accepts praise for her composure, and observes that a perk of being a sociopath is learning to gather information by manipulating people. Earlier, listening through a wall, she says something more diagnostically precise than the murder:

I’ve always been intrigued by the inflections of emotion in people’s voices. Most emotions felt like foreign languages to me, but over the years, I had become very good at translating them and reading people. Sometimes, I would reenact conversations I’d overheard, watching my facial expressions in the mirror.

That is a portrait of maximal thickness: an absolute wall between self and other, across which affect does not pass and must instead be translated, acquired as a second language, rehearsed before a mirror. Where the boy in the Aiden stories cannot keep the dead out, Molly cannot let the living in.

The critical claim is that this portrait requires the opposite constitution to produce. A writer for whom emotional traffic between persons is itself effortful would have no particular reason to find Molly’s condition remarkable, and would likely render her as a villain — that is, from outside, morally rather than phenomenologically. Rendering thickness from the inside, with attention to the specific labour it involves, is available chiefly to someone for whom that labour is unnecessary and therefore visible. Permeability is what makes impermeability legible.

This is why I hold that mercy in fiction is a technical position before it is a moral one. It is not primarily a virtue the author exercises toward the character. It is the precondition of being able to occupy a person one is not — and it has the same source as the nightmares.

XII. A Poetics of the Unfinished

Let me state the resulting position as compactly as I can, with its warrants and its limits attached.

Hartmann’s boundary construct is a reliably measured, independently replicated personality dimension describing the permeability of partitions within the mind — between sleep and waking, self and other, thought and feeling, past and present, real and imagined. It correlates strongly with absorption and openness, is elevated in nightmare sufferers and in at least some populations of writers and visual artists, and thickens measurably with age. It also explains a modest share of variance in its flagship outcome, is predicted as well by insecure attachment as by artistic sensibility, and has never been administered to me.

Romantic poetics described the same dimension without instruments: Keats as the self that has no self, Coleridge as the faculty that shapes into one by dissolving in order to recreate, Wordsworth as a developmental gleam that fades in most people, Emerson as a unity in which every particular being is contained. Bergson gave it a metaphysics of duration, and supplied the criterion that distinguishes conservation from mere repetition: what returns returns as difference.

And a practice follows, which is the part that concerns writers rather than psychologists. If the past does not file itself away as concluded, then material is never finished, and the recurrence of a grandmother or a chalice or a single star across a body of work is not a failure of supply but a property of retention. The images do not recur because nothing new arrives. They recur because nothing old departs. Those are opposite conditions, and criticism — which has long treated the second as evidence of the first — has been mistaking one for the other for a long time.

Whether any given writer’s returns are generative or merely compulsive is a real question with a real answer, and the answer is available on the page rather than in the biography: does the material come back changed? I have tried to make my own work answerable to that test rather than exempt from it.

What remains, after the correction with which this essay opened, is not an ancient Celtic doctrine. It is something better attested and less picturesque: a documented folk psychology of permeable boundaries between the living and the dead, recorded from Irish informants by a folklorist working from archives and fieldwork⁴; a personality dimension that describes the same permeability in the individual case; and a literary tradition that theorized it two centuries before anyone thought to score it.

My grandmother, who had a sixth-grade education and walked a mile to a well, would have found the questionnaire beside the point and the finding obvious. She believed the dead were in the next room. Hartmann’s contribution, translated out of Boston clinical English, is the observation that for some people the wall to that room was never fully built — and that those people begin as everyone begins, at six or seven, when the images are enormous for all of us, and simply fail to complete the ordinary work of thickening.

Most people set it down. A few keep having to write it.

— James Mulhern