What Happened
Grief · Psychiatry · Diagnostic Classification · History of Science · Evidence Brief1917 to 2022 · record to 17 Sep 2026
Grief theory, 1917 to the present
The Dying, Not the Bereaved
Denial, anger, bargaining, depression, acceptance. Elisabeth Kübler-Ross drew the five stages from people facing their own deaths, and published them in 1969. The book that puts the five stages of loss in its title and addresses the bereaved appeared in 2005, the year after she died. In between, the model became something almost nobody had tested: taught in 125,000 courses by 1982, and still with no study behind it. When a study finally came, in 2007, it said two things at once — and two of its four authors spent the next year taking one of them back.
- 1969the year the five stages were published, from Kübler-Ross’s contact with “over two hundred dying patients” — her own figure
- 2005the year a book titled for the five stages of loss and addressed to the bereaved appeared, published after her death
- 233bereaved people in the one study that set out to test stage theory directly, in Connecticut between 2000 and 2003
- 46% and 66.3%two published resilience shares drawn from a single widowhood dataset, ten years and one statistical method apart
Graded by how well each item stands up Sorted by standing. The model and the evidence about it are separate rows, never one row.
| Standing | What is recorded | Where |
|---|---|---|
| Confirmed | On Death and Dying, published November 1969, sets out five “stages” — denial, anger, bargaining, depression, acceptance — observed in dying patients. Kübler-Ross’s own figure for the contact behind them is “over two hundred dying patients” | Kübler-Ross (1969), p. 38, quoted in Stroebe, Schut and Boerner, Omega 74(4) |
| Confirmed | That same 1969 book already extends the stages to the anticipatorily bereaved family, in its chapter on the patient’s family. By 1975 a “stages of grief” ideology and a “currently popular conception of stages of widowhood” were in circulation | Stroebe, Schut and Boerner, Omega 74(4) (2017), quoting Lopata (1975), p. 50 |
| Confirmed | Kübler-Ross qualified the model herself in 1974: “Most of my patients have exhibited two or three stages simultaneously, and these do not always occur in the same order” | Questions and Answers on Death and Dying (1974) |
| Confirmed | On Grief and Grieving: Finding the Meaning of Grief Through the Five Stages of Loss, written with David Kessler, is published in 2005 — after Kübler-Ross’s death on 24 August 2004 | Scribner (2005); publication history from the encyclopaedic record of her life |
| Confirmed | Wortman and Silver name four assumptions the model rests on — that depression is inevitable, that distress is necessary and its absence pathological, that a loss must be worked through, and that recovery and resolution are to be expected — and conclude that “available empirical work fails to support and in some cases contradicts them” | Wortman and Silver, Journal of Consulting and Clinical Psychology 57(3) (1989) |
| Confirmed | The Yale Bereavement Study, 233 bereaved people in Connecticut with data collected from January 2000 to January 2003, reports both that “counter to stage theory, disbelief was not the initial, dominant grief indicator” and that the five indicators reached their maximum values in the sequence stage theory predicts. Its stated conclusion opens by speaking of “the normal stages of grief” | Maciejewski, Zhang, Block and Prigerson, JAMA 297(7) (2007) |
| Confirmed | Three letters disputing that study appear in the same journal four months later. Twenty-two months on, two of the study’s four authors write that “stages of grief might more accurately be described as states of grief” | JAMA 297(24) (2007); Prigerson and Maciejewski, British Journal of Psychiatry 193(6) (2008) |
| Confirmed | The field’s own review of the whole record concludes: “There is no scientific foundation, and decades of research have shown that most people do not grieve in stages” | Stroebe, Schut and Boerner, Omega 74(4) (2017) |
| Confirmed, not verifiable here | The 2005 book does both things at once. At p. 7 it warns the stages “are not stops on some linear timeline in grief”; at p. 12 it says “Anger is a necessary stage of the healing process.” Both passages are quoted with page numbers by the 2017 review; the book itself was not read for this piece | Kübler-Ross and Kessler (2005), quoted in Omega 74(4) |
| Confirmed, not verifiable here | The two most-quoted lines of the 2007 letters — Silver and Wortman on the model’s “devastating consequences”, Bonanno and Boerner on a “misguiding message” — reach this page through the review that carries them with page numbers. The letters themselves could not be opened at the publisher’s own page | JAMA 297(24) (2007), via Omega 74(4) (2017) |
| Unconfirmed | How many dying patients Kübler-Ross actually interviewed. “Over two hundred” is the book’s own wording and the only figure any source read here supports; larger counts circulate and none could be established | Kübler-Ross (1969), p. 38, via Omega 74(4) |
| Unconfirmed | The death toll of the 1942 Cocoanut Grove fire, the disaster from which grief research first took a clinical form. 492 is the headline figure; the memorial committee documents 490, with a capsule biography for each name, and the memorial under construction uses 490 granite bricks. Neither count supersedes the other | The encyclopaedic record of the fire, which carries both figures |
Timeline
1917 to 2022 Two threads run through it: what was written about grief, and what happened whenever anyone tested it. They are twenty-seven years apart at the start and four months apart in 2007.
- 1917Freud publishes “Trauer und Melancholie”, distinguishing mourning as a conscious, healthy process from melancholia as an unconscious, pathological one. The grief-work idea it founds — that attachment must be withdrawn from what was lost — governs the field for most of the century
- 1942On 28 November, a fire at the Cocoanut Grove nightclub in Boston kills 490 or 492 people, depending on which record is consulted
- 1944Erich Lindemann, a Massachusetts General Hospital psychiatrist who studied the fire’s survivors and the bereaved, reads “Symptomatology and Management of Acute Grief” at the American Psychiatric Association’s centenary meeting. Grief becomes a clinical observation rather than a psychoanalytic argument
- Nov 1969On Death and Dying appears, with five “stages” drawn from dying patients. Its own chapter on the patient’s family already extends them to those waiting
- 1974Kübler-Ross’s own qualification, five years in: “Most of my patients have exhibited two or three stages simultaneously, and these do not always occur in the same order”
- 1975Helena Lopata records a “stages of grief” ideology and a “currently popular conception of stages of widowhood” among Chicago women. The model is folk wisdom about the bereaved thirty years before a book of it is addressed to them
- Jun 1989Wortman and Silver publish “The myths of coping with loss”, and find that the available empirical work fails to support the assumptions the stage model rests on
- 1999Stroebe and Schut publish the dual process model, naming the grief-work hypothesis’s faults — imprecise definition, no account of dynamic processing, no validation across cultures or periods — and replacing progression with oscillation between loss-oriented and restoration-oriented coping
- 2002A prospective study of 205 people, with data gathered some years before a spouse’s death and again at six and eighteen months after, finds resilience the most frequent pattern and common grief relatively infrequent
- 2005Two publications, one year after Kübler-Ross’s death. On Grief and Grieving, with the five stages of loss in its title and the bereaved as its readers. And a randomised trial in which complicated grief treatment beats interpersonal psychotherapy, 51% responding against 28%, with a number needed to treat of 4.3
- Feb 2007The Yale Bereavement Study is published, claiming the first empirical examination of stage theory, and reporting results that cut both ways
- Jun 2007Three letters dispute it in the same journal — from Silver and Wortman, from Weiner, and from Bonanno and Boerner — with an author reply
- Dec 2008Two of that study’s four authors propose that “stages of grief might more accurately be described as states of grief”, and that they be relabelled multidimensional states that evolve and diminish over time
- 2010A conceptual replication in 614 bereaved young adults, 441 bereaved by natural and 173 by violent causes, finds “limited support” for stage theory — and sense-making a far stronger predictor of grief than time since the loss
- 2012The 2002 dataset is re-analysed by Latent Class Growth Analysis. The resilient share moves from 46% to 66.3%, nothing about the bereaved having changed — and what is being measured is a trajectory of depression
- 2017Stroebe, Schut and Boerner review the whole record and ask that the stages be “relegated to the shelves of history”. They add a complaint about their own side: a range of alternatives “presents a weak, nonunited front to stage theory”
- 2018A review of 54 studies of potentially traumatic events of every kind — not only bereavement — finds resilience the modal response at 65.7%, with a 95% confidence interval of 61.6 to 69.8
- Jan 2022ICD-11, adopted by the 72nd World Health Assembly in 2019, comes into effect carrying prolonged grief disorder. DSM-5-TR appears the same year with a differently worded prolonged grief disorder of its own
- 2023Maarten Eisma sets out the case against both: a non-linear history that breaks the link to earlier research, no agreed definition of normal grief to measure prolonged grief against, and the risk of medicalising a common response
The Argument
Three disagreements, named on both sides None of them is settled by this page. One of them is not settled at all.
The Yale Bereavement Study is cited by both camps, and neither is misquoting it. Its abstract opens against the model it set out to test: “Counter to stage theory, disbelief was not the initial, dominant grief indicator. Acceptance was the most frequently endorsed item and yearning was the dominant negative grief indicator from 1 to 24 months postloss.” Then, in the same paragraph, it reports that the five indicators reached their maximum values in exactly the sequence stage theory predicts — disbelief highest at one month, yearning peaking at four, anger at five, depression at six. Its stated conclusion opens by speaking of the normal stages of grief.
Three letters in the same journal four months later disputed the design and the reading. Silver and Wortman wrote that a mistaken belief in the stage model “can have devastating consequences”, leading bereaved people to feel they are not coping properly and producing unhelpful responses from their networks and from clinicians. Bonanno and Boerner wrote that grief stages “tell us little about how people might cope with the loss” and that, set against evidence contradicting any “average” normal response, this is “a misguiding message”. Both passages reach this page from the 2017 review that carries them with page numbers; the letters themselves could not be opened.
Then the authors moved. In December 2008, two of the four wrote that “stages of grief might more accurately be described as states of grief”, proposing they be relabelled multidimensional states that evolve and diminish in intensity over time. A conceptual replication in 614 bereaved young adults found “limited support” two years after that.
- 4months from the study to the letters disputing it, in the same journal
- 22months to two of its four authors preferring “states” to “stages”
- 614bereaved young adults in the replication that found limited support
There is no date on which the five stages stopped being about dying and started being about grieving, and looking for one is how the story gets told wrong in both directions.
The observations are of dying patients. But the 1969 book does not stop there. Stroebe, Schut and Boerner call it “important to note” that Kübler-Ross extended the stages to the anticipatorily bereaved family inside that same book, in its chapter on the patient’s family. By 1975, Helena Lopata was recording a “stages of grief” ideology and a “currently popular conception of stages of widowhood” among Chicago women — thirty years before any book of the five stages was addressed to them.
What 2005 supplies is the cover. On Grief and Grieving: Finding the Meaning of Grief Through the Five Stages of Loss, written with David Kessler and published the year after Kübler-Ross died, is the first book to put the five stages of loss in front of the bereaved under that name. It also does both things at once: at p. 7 it warns that the stages “are not stops on some linear timeline in grief” and that not everyone goes through all of them or in a prescribed order; at p. 12 it says “Anger is a necessary stage of the healing process.” The review that quotes both notes that anger is a common grief symptom and not a universal one.
Kübler-Ross had made the same qualification in 1974 and it did not travel with the model. That is the ordinary fate of a memorable idea, and it is the reason the stages were described as descriptive and used as prescriptive.
- 1969the stages published, and already extended to the waiting family in the same book
- 1975stages of widowhood recorded as folk wisdom among Chicago women
- 2005the five stages of loss on a cover, addressed to the bereaved
That most bereaved people come through without professional help is the field’s best-travelled finding, and the number behind it needs handling.
In 2002, a prospective study gathered data on 205 people some years before their spouses died, then again at six and eighteen months after. It found resilience the most frequent pattern and common grief relatively infrequent; the classification put 46% in the resilient group. Ten years later the same data were re-analysed by Latent Class Growth Analysis, and the resilient share came out at 66.3%. Nothing about the bereaved had changed. The method had — and the 2012 paper says so of its predecessor in as many words, calling the earlier procedures rudimentary.
Two cautions follow, and the page holds both. A resilience headline is a statement about an analysis as much as about grief. And the frequently quoted “about 65%” is a different number from a different study: 65.7%, from a 2018 review of 54 studies of potentially traumatic events of every kind, not of bereavement. The 66.3%, meanwhile, is a trajectory of depression symptoms, which that paper’s own title says. That the two land within a percentage point of each other is a coincidence, and it is not read here as a confirmation.
- 46%resilient, by the original classification of 205 bereaved spouses
- 66.3%resilient, from the same data re-analysed ten years later
- 65.7%resilient across 54 studies of potential trauma of every kind, a different question
| Figure | What it measures | In whom |
|---|---|---|
| 46% | A resilient trajectory, classified by the study’s original procedure | 205 bereaved spouses in the United States, from before the death to 18 months after |
| 66.3% | A resilient trajectory of depression symptoms, by Latent Class Growth Analysis | The same 205 people, re-analysed ten years later |
| 65.7% | Resilience as the modal response, 95% confidence interval 61.6 to 69.8 | 54 studies of potentially traumatic events of every kind, bereavement among them |
What Others Add
What replaced the stages, and what the diagnosis asks The field did not stop at saying no, and it is not united about what to say instead.
1989
Four assumptions, none of them supported
- That depression is inevitable after a loss
- That distress is necessary, and its absence a sign of pathology
- That a loss must be worked through
- That recovery and resolution are to be expected — and, on all four, that “available empirical work fails to support and in some cases contradicts them”
1999
Oscillation, not progression
- Names the grief-work hypothesis’s faults: imprecise definition, no account of dynamic processing, and no validation across cultures or historical periods
- Two stressors rather than a sequence: loss-oriented, which is the grief itself, and restoration-oriented, which is the rebuilt life
- Argues for dosage: respite from grieving is part of adaptive coping, not a failure of it
1944
Where grief became clinical
- A Boston nightclub fire on 28 November 1942 killed 490 or 492 people; the two counts are both current and neither has displaced the other
- Erich Lindemann, a Massachusetts General Hospital psychiatrist, studied the survivors and the bereaved and read his paper at the American Psychiatric Association’s centenary meeting in 1944
- It is where grief stops being a psychoanalytic argument and starts being something a clinician writes down — twenty-five years before the five stages
On offer today
Seven alternatives, and that is the complaint
- Trajectories, cognitive stress theory, meaning making, the psychosocial transition model, the two-track model, the dual process model, the task model
- The review’s own objection to its own side: a range of alternatives “presents a weak, nonunited front to stage theory”
- What it asks for instead is one accessible substitute that attempts an explanation, rather than a list — which is why the stages, simple and memorable, keep their place
Both current diagnostic manuals now carry a prolonged grief disorder, and they are not quite the same disorder.
ICD-11 was adopted by the 72nd World Health Assembly in 2019 and came into effect on 1 January 2022; DSM-5-TR appeared that year, listing prolonged grief disorder as a new diagnosis. Set side by side they differ in the places that decide who is a case. DSM-5-TR requires the death to have been at least twelve months ago and at least three of eight named symptoms. ICD-11 sets no minimum in its event criterion — only that responses lasting under six months, and longer in some cultural contexts, do not qualify — and it publishes no diagnostic algorithm at all. A content analysis put the overlap between the two at a Jaccard index of 0.47, moderate on the scale it used, with each resembling its own predecessor more closely: 0.58 and 0.63.
The difference is measurable in people. In 211 Dutch and 222 German bereaved adults, fewer were probable cases under DSM-5-TR than under ICD-11. But a comparison of six criteria sets in 855 bereaved individuals found prevalence between about 10% and 20% and agreement between the sets substantial — so the definitions disagree about wording rather more than they disagree about people.
The objection is not that severe, disabling grief is unreal. Maarten Eisma names the gap: there is no consensus definition of normal grief to measure prolonged grief against, two taxometric studies support a dimensional rather than a categorical picture, and the estimated prevalence of pathological grief after unnatural deaths is 49% — a figure that turns diagnosis into the labelling of a common response. Against that stands the trial that is the diagnosis’s clinical argument: 51% responded to complicated grief treatment against 28% to interpersonal psychotherapy, with a number needed to treat of 4.3.
- 0.47Jaccard overlap between the two current definitions — moderate, on the scale used
- 6 vs 12months since the death before either manual will consider a case
- 49%estimated prevalence of pathological grief after unnatural deaths, which is the medicalisation argument in one figure
| Where they differ | ICD-11 | DSM-5-TR |
|---|---|---|
| Time since the death | No minimum in the event criterion; responses under six months, and longer in some cultural contexts, do not qualify | At least twelve months; six for children and adolescents |
| Accompanying symptoms | Listed without a count, and with no published diagnostic algorithm at all | At least three of eight, each one named |
| Other disorders | Not carried in the tabulated criteria | A criterion of its own: not better explained by major depressive disorder, post-traumatic stress disorder or another disorder |
| Which it most resembles | Its own earlier draft, at a Jaccard index of 0.58 | Persistent complex bereavement disorder, at 0.63 — each is closer to its predecessor than to the other, at 0.47 |
| Who it picks out | More probable cases, in a sample of 433 Dutch and German bereaved adults | Fewer, in the same sample |
Conclusion
What to hold, and how loosely A model this comfortable is not dislodged by a finding. It is dislodged by knowing what it was for.
The five stages came from a psychiatrist sitting with people who were dying, in a hospital where nobody would talk to them about it. That work changed how dying is done, and nothing on this page touches it.
What followed it was a drift with no date. From the dying, to the family waiting. From the family, to widowhood as folk wisdom by 1975. From folk wisdom to a book title in 2005, and from there to job loss, divorce, addiction and, on the same list, a cancelled insurance policy. Kübler-Ross qualified the model herself in 1974 and again in 2005, and neither qualification travelled with it — which is the ordinary fate of a memorable idea and not a charge against her.
The charge, such as it is, belongs downstream. A model taught in 125,000 courses by 1982 was tested directly for the first time thirty-eight years after it was published. That study was disputed in the same journal within four months. Two of its four authors preferred states to stages within two years. And the field’s own review of the whole record, in 2017, concludes that there is no scientific foundation and asks that the stages be shelved.
Three things to hold loosely, because this page holds them loosely. The 2007 study genuinely points both ways, and anyone citing it in either direction is using half of it. The resilience figures are statements about methods as well as about people, and the two best-known ones measure different things. And the newest institutional fact here — that severe, disabling grief now has a name in both diagnostic manuals — is the one with the most live argument attached, not the least.
- 38years between the stages being published and the first study that set out to test them directly
- 125,000courses in which the stages had already been taught, by Kübler-Ross’s own 1982 estimate
What is established
The stages were drawn from dying patients and published in 1969; the same book already extended them to the waiting family, and 2005 is when a book of the five stages of loss was addressed to the bereaved. Wortman and Silver found no support for the assumptions beneath them in 1989. The 2017 review of the whole record finds no scientific foundation.
What is genuinely contested
Whether the 2007 study supports stage theory. It reports findings counter to it and a peak sequence consistent with it, and its own conclusion uses the phrase “the normal stages of grief”. Three letters disputed it and two of its four authors later preferred “states”. Citing it for either side means dropping half the abstract.
What to watch
Whether the two definitions of prolonged grief disorder converge. They overlap at a Jaccard index of 0.47, each resembles its own predecessor more closely, and there is still no agreed definition of normal grief to measure either against.