NOTE-1291: Surviving Early Childhood Trauma: Effects of the Holocaust on Survivors' Psychological and Physiological Well-Being

Contribution

A two-generation follow-up of the sample of Sagi-Schwartz et al. (2003, Am J Psychiatry), first studied 11 years earlier. It asks whether childhood Holocaust trauma still shows in survivors in old age, in mind and in the HPA axis, and whether anything reaches their daughters. It adds the stress hormone cortisol, measured over a day and under a standard stressor, and one candidate gene. Its answer is a split: marks in the survivors, almost none in the daughters.

Key insight

The trauma stays with the person who lived it and largely stops there. The thesis reads the daughters’ normal adjustment as the survivors’ achievement as parents: “they were unable to integrate their past and present challenges, but they nevertheless successfully managed to keep their children emotionally unharmed” (p. 40). It goes further and suggests that not resolving the trauma may be what protected the next generation: “Perhaps dissociation instead of resolution of extreme, incomprehensible trauma is the desirable solution as it paves the way for a more balanced life of generations to come” (p. 40; Proposition 8).

Assumptions

• The comparison group isolates the Holocaust. Comparison women were born in Europe and emigrated to pre-state Israel with their parents just before the war. Both groups then lived through the same post-war Israel (p. 39).
• Population-registry sampling avoids the bias of convenience and clinical samples, which the meta-analyses blame for earlier reports of transmission (pp. 16, 27).
• Cortisol indexes trauma “under the skin”. Both raised and lowered cortisol count as dysregulation (p. 50).
• The ADRA2B deletion marks vivid traumatic memory, from de Quervain et al.’s (2007) Rwandan refugees (p. 87).

Key results

• The sample (pp. 18, 28–29, 51–52).
  • Of 106 first-generation women from the original study, 10 had died and 17 could not take part for health or personal reasons. 65 of the 79 eligible (82.3%) agreed: 32 survivors and 33 comparisons, aged 71 to 84 (M = 76.98).
  • Of 104 daughters, 79 took part (47 of survivors, 32 of comparisons), aged 38 to 59 (M = 47.46).
  • Survivors had to have lost both parents in the Holocaust (pp. 37, 51). Daughters had been recruited only if they had a child of 12 to 15 months (p. 104).
• Chapter 2: psychological outcomes (pp. 31–36).
  • Survivors against comparisons:
    • dissociation (DES) 7.13 against 5.03, t(62) = 2.19, p = .03;
    • satisfaction with life 23.91 against 27.64, p = .001;
    • cognition (TICS-m) 30.91 against 34.21, p = .041;
    • perceived stress of recent life events 5.50 against 3.91, p = .022, with no difference in the number of events;
    • mental health 54.59 against 59.15, p = .06; physical health, no difference.
  • In a logistic regression, only life satisfaction (odds 2.79, 95% CI 1.16 to 6.70) and dissociation (odds 2.39, 1.14 to 4.99) predicted being a survivor.
  • Daughters: no difference on any measure. Dissociation 7.23 against 7.19; life satisfaction 26.21 against 28.41 (p = .09). No variable predicted group in the regression.
  • Mothers and daughters correlated only in the survivor families, and only on life events: number of events r = .52, perceived stress r = .58 (Table 4; comparison families .17 and .09). The authors note that some events were shared family stressors (p. 37).
  • Attrition was selective: survivors who stayed in had reported more unusual beliefs 11 years earlier (t(44) = 2.07, p = .04), and comparisons who stayed in, higher well-being (p = .005).
• Chapter 3: diurnal cortisol (pp. 54–59). Six saliva samples on one day. Complete data for 58 mothers and 74 daughters.
  • Survivors’ cortisol was higher at every time point, with education as covariate: group F(1,55) = 6.10, p < .02, η² = .10; total output (AUCg) F(1,55) = 7.22, p = .01, η² = .12. Survivors had 5.1 years of schooling against 8.7.
  • Dissociation did not moderate it, and did not correlate with cortisol within the survivors (r = .07).
  • Daughters did not differ: group F(1,71) = 0.03, p = .86; AUCg p = .70.
  • In 42 mother–daughter dyads there was no group-by-generation interaction (p = .16).
  • Exploratory: daughters of survivors above the median on dissociation had lower AUCg (1.27, n = 17) than daughters of less dissociative survivors (1.39, n = 10): interaction F(1,42) = 4.44, p = .04; t(25) = 1.80, one-tailed p = .04, d = 0.73.
  • The authors read this through parenting. A dissociating parent is at once the attachment figure and a source of “fright without solution” (Hesse and Main), a precursor of disorganized attachment (p. 60).
• Chapter 4: response to the Trier Social Stress Test (pp. 72–76). The test was done at home, and the mothers’ speech argued that a grandchild deserved a scholarship. Complete data for 38 mothers (19 and 19) and 64 daughters (36 and 28).
  • Mothers: the only group difference was in the change from baseline to the first post-stress sample, F(1,35) = 4.10, p < .05. Survivors were high before the task and fell after it; comparisons rose.
  • Daughters: no difference, F(1,61) = 1.12, p = .29; both groups “showed similar declines” (p. 75). No mother–daughter correlations (Table 3, n = 16 to 22 per cell).
  • The authors suggest the survivors’ high baseline was anticipation, perhaps of “a research session about the Holocaust” (p. 77).
• Chapter 5: ADRA2B (pp. 91–95). 54 mothers genotyped (25 survivors, 29 comparisons).
  • Group by genotype, diurnal cortisol: F(1,32) = 4.98, p < .05.
  • Group by genotype, stress response: F(1,28) = 5.79, p < .05.
  • Among wildtype carriers survivors had higher cortisol than comparisons; among deletion carriers there was no difference.
  • The authors call the study “exploratory” (p. 95). They propose that deletion carriers, with more vivid memories, were “driven to resolve their vividly remembered experiences” and so ended less dysregulated (p. 95).
• What it reports of the meta-analyses (pp. 16, 26–27, 49, 59, 67, 101, 103). These are the two works the reading stands in for.
  • 2003, second generation: “32 samples with 4,418 children of Holocaust survivors”; “no evidence for secondary traumatization in studies with non-select recruitment and non-clinical samples” (p. 26). Transmission appeared in clinical samples and in weaker designs on convenience samples (pp. 16, 27): the clinical cases named are severe combat reaction (Solomon et al. 1988) and breast cancer (Baider et al. 2000) (p. 49).
  • The 2003 explanations, as reported: Paris’s stress-diathesis model and the social support available in an Israel full of survivors and their children (p. 59); and a “basic trust” formed before the war, when the survivors were 4 to 14, and then passed on (p. 103).
  • 2008, third generation: “another meta-analysis of 13 non-clinical samples of third generation participants (N = 1012). Again, there was no evidence for intergenerational transmission” (p. 67). It tested whether trauma “might skip the second generation and then be detected in the third” (p. 67).
  • The first phase of this sample (Sagi-Schwartz et al. 2003) found no difference in the third generation, then infants, nor in the daughters’ observed parenting of them (pp. 26, 49).
• Explanations offered for the daughters’ resilience (pp. 103–104).
  • Basic trust from before the war.
  • Immigration to Israel, where survivors “acquired a new sense of meaning, goals, and pride in building the new state of Israel” (from Barel et al. 2010).
  • Proposition 7: “Victims of post trauma might cope better with their devastating history when they can find meaning to their lives supported by their society.”

Claims

id	claim	strength	support
C1	Female child survivors who lost both parents show more dissociation, less life satisfaction and more cognitive impairment than matched women more than 60 years later	moderate: registry sample, matched comparisons, self-report, n = 32 and 33, selective attrition	Ch. 2
C2	Their daughters do not differ from comparison daughters in mental health, dissociation, cognition or physical health	moderate: a null at power the authors call sufficient for medium effects; daughters selected for having an infant	Ch. 2
C3	Survivors have higher diurnal cortisol than matched women; their daughters do not	moderate for survivors (one day of sampling); moderate for the daughters’ null	Ch. 3
C4	Daughters of more dissociative survivors have lower diurnal cortisol	weak: exploratory, after a null main test, median split, n = 17 and 10, one-tailed p = .04	Ch. 3
C5	Survivors respond anomalously to a laboratory stressor; their daughters do not	weak to moderate: one contrast of four, 19 per group	Ch. 4
C6	ADRA2B moderates survivors’ cortisol: the excess appears only in wildtype carriers	weak: exploratory by the authors’ description, about 12 per cell	Ch. 5
C7	In non-clinical samples with adequate designs there is no meta-analytic evidence of second- or third-generation traumatization	not assessable here: reported at second hand, without effect sizes	Chs. 1–4, 6, citing the 2003 and 2008 meta-analyses

Concepts

• secondary and tertiary traumatization: the transmission of a survivor’s trauma to children and to grandchildren respectively.
• non-select sample: one drawn from population-wide records, as against a convenience sample from survivors’ organizations or a clinical one.
• “under the skin”: trauma carried in physiology, here the HPA axis, beneath normal-looking adjustment (pp. 49, 101).
• AUCg: total cortisol output over the day, the area under the curve with respect to ground.
• “historian” rather than “victim”: a role through which survivors may reach ego integrity without integrating the trauma, by bearing witness (p. 37; Proposition 2).

Connections

• The two meta-analyses (LIT-1718, LIT-1770). The thesis’s promotors are their three authors, and the restatements are theirs at one remove. It gives sample sizes and conclusions but no effect size for either.
• Yehuda’s group. It found lower cortisol in US survivors with PTSD and in offspring of survivors with PTSD. The thesis found higher cortisol in Israeli survivors and explains the gap by chronic stress in post-war Israel (p. 59). Its exploratory offspring result runs the same way as Yehuda’s: lower cortisol in children of more symptomatic parents (p. 102). Yehuda et al.’s FKBP5 paper (LIT-1692) and Yehuda and Lehrner (LIT-1715) postdate the thesis.
• Danieli’s handbook (LIT-1701) is cited for the clinical tradition (Felsen 1998, on daughters’ greater vulnerability, p. 27). The thesis does not discuss the “conspiracy of silence”.
• Heard and Martienssen (NOTE-1254). The daughters were born in Israel after the war to mothers traumatized as children. This is an intergenerational case in their sense, and every route is open for it. The thesis tests none apart from the others.
• Costa et al. (NOTE-1255). Their untested line, that a return “to a grateful nation” protected sons, has a counterpart here: the meaning and pride of building Israel, offered for survivors (p. 104).
• Horsthemke (NOTE-1265). The thesis is an example of his point. Its one offspring signal is explained by parenting with no means of excluding other routes.

Bearing on the record

• The intergenerational trauma question (QUESTION-297). The thesis bears on three of its four routes. These are inferences; the thesis does not frame its findings by route.

  • Parenting. This is the route the authors favour, and the evidence for it is thin. The daughters show no psychological difference. Their one physiological difference, lower cortisol in daughters of more dissociative mothers, rests on 27 people in an exploratory split, and the authors explain it by dissociative parenting. The mother–daughter link in stressful life events fits shared family circumstance as well as transmission.
  • Germline and womb. Not tested. The daughters were conceived years after the war, so on Heard and Martienssen’s rule (NOTE-1254) anything they carry is intergenerational, and the design cannot tell a germline mark from care.
  • Narrative and identity. The thesis does not measure identity, but its design matters for that route. The comparison mothers were European Jews who reached Palestine before the war. Their daughters grew up in the same state, under the same national memory of the Holocaust. So the design compares family exposure while holding the collective story fixed. An injury carried as part of a shared Israeli or Jewish identity would act on both groups and be invisible here. The null therefore says the family route adds nothing detectable on top of whatever the shared identity carries. It says nothing about the identity route itself.
  • Meaning as a buffer. Where the thesis touches a group’s story, the story protects. Survivors may reach ego integrity as “historians” rather than “victims” (p. 37), and building the new state may have given meaning that protected them (p. 104). This is the second time the record has met this idea, after Costa et al.’s “grateful nation” (NOTE-1255). Neither work tests it.
  • Silence protects, on this account. Proposition 8 says that “lack of resolution and setting aside the trauma on the part of survivors may protect the offspring”. The clinical tradition the question cites (Danieli, LIT-1701) treats silence as the means of transmission. Read with the identity route, the inference is this: in a family, what an ancestor’s injury passes on may depend on whether it is told, and here the untold injury was not visibly passed on.

• The owner’s thesis (CLAIM-491). Only indirectly, through the question’s identity route. If a descendant carries an ancestral injury as part of an extended self, these data put that self in the group, not the family: being a survivor’s daughter added nothing measurable. That is an inference from a null, and the thesis measures no self.

• The triviality objection (CLAIM-492) and Batson’s objection (CLAIM-494). Not addressed.

• A dated note should be added to QUESTION-297. Proposed text:

  Note of 2026-10-29: the parenting route, read through Fridman (2011). The two meta-analyses are still unread, but their authors’ student restates them: no second-generation traumatization in non-clinical, well-designed samples (32 samples, 4,418 offspring), and none in the third (13 non-clinical samples, N = 1,012). Her own registry sample of daughters of child survivors showed no psychological or cortisol difference. The one signal, lower cortisol in 17 daughters of more dissociative mothers, is exploratory. The comparison families shared the survivors’ national memory, so the design tests the family routes with the identity route held fixed. Where the thesis speaks of a group’s story, as meaning found in building a state, it treats the story as a buffer, not a carrier.

• No THEORY filed. One sample, women only, with the transmission result exploratory.

• No ML instruction.

Limitations

• Not the meta-analyses. No effect size, moderator test or study list is given for either. The restatements come from the meta-analyses’ own authors’ group and vary slightly from chapter to chapter.
• One attribution goes beyond the meta-analyses. The General Discussion says the meta-analyses “show … that intergenerational transmission may be present under the skin, when risk factors such as higher parental posttraumatic symptomatology exist” (p. 103). The “under the skin” result is this thesis’s own (Chapter 3). The 2003 meta-analysis, as the thesis reports it elsewhere, concerned psychological outcomes.
• Women only, and selected. Mothers had to have become grandmothers, and daughters to have had an infant at first contact. The authors grant this may have picked survivors able to “create new generations” (p. 104).
• Small cells. Chapters 4 and 5 rest on about 19 and about 12 per cell. The transmission finding is one exploratory split.
• The stressor may not have stressed. In the unadjusted log means, comparison mothers rose only from 0.55 to 0.57 after the test, and both groups of daughters fell (0.42 to 0.37 and 0.48 to 0.46). The discussion nonetheless says the procedure was “sufficiently stressful for both second generation groups to generate increases” (p. 78). The results section says the daughters “showed similar declines” (p. 75).
• The survivors’ stress response differs between chapters. In Chapter 4, survivors fall after the stressor (0.57 to 0.48). In Chapter 5’s table, from an overlapping subsample with different n per sample, they rise (0.61 to 0.65).
• One-day cortisol sampling (p. 60).
• A tension between chapters. Chapter 2 suggests dissociation rather than resolution may be “the desirable solution” (p. 40). Chapter 5 suggests that working through vivid memories leaves survivors less dysregulated (p. 95). The authors reconcile them as short-term against long-term costs (pp. 40, 95), without data on either.
• Internal inconsistencies.
  • Chapter 2 text gives the mother–daughter correlations as r = .53 and r = .42 (p. 35); Table 4 gives .52 and .58 (p. 36).
  • The mortality test reports χ²(1) = .44, p = .74 (p. 31); with one degree of freedom that χ² gives p of about .51.
  • Table 2’s note gives 33 comparison daughters (p. 33); the text and abstract give 32.
  • Chapter 3’s discussion puts the dissociation in the “adult offspring of Holocaust survivors showing elevated levels of dissociation” (p. 60). The result was for the mothers’ dissociation.
  • Chapter 5 gives 79 of 106 as “75.4%” (p. 88); it is 74.5%.
  • The Dutch summary says 106 and 104 participants “took part” in the present study (p. 112); the English chapters say they were contacted.
  • Sagi-Schwartz et al. 2008 is cited as Attachment and Human Development, 10, 1–17 (p. 21 and elsewhere). The record’s LIT, from Crossref, gives 10(2), 105–121.
• Typos.
  • “physiolocial” in the repository’s title of the thesis (cover sheet).
  • “have past” for “have passed” (p. 15); “Holcaust” (p. 17).
  • “Kudileka” and “Kirrschbaum” for Kudielka and Kirschbaum (p. 17, p. 68).
  • “Chodoff, 1968” in the text, 1986 in the references (pp. 16, 26; pp. 19, 41).
  • “comparison offspring of Holocaust survivors” for comparison offspring (p. 28).
  • “Goloer” and “Seckel” for Golier and Seckl (p. 68); “preformed” for “performed” (p. 72).
  • Chapter 5’s running head reads “Secretaion” (odd pages, pp. 87–95).
  • “Backround” in the list of tables (p. 11).
  • In the Dutch summary, “cortosolniveaus”, “opmerkeljke”, “controlgroep” and “stresvolle” (pp. 111–113).
  • The CV’s first page is numbered 120, repeating the acknowledgements’ page number.

Open questions

• What effect sizes did the 2003 and 2008 meta-analyses find, and how large was the clinical-sample effect? Only the papers can say.
• Does the dissociation result replicate in a sample large enough to test it as a prediction, with the daughters’ attachment measured?
• Would a comparison group outside the shared national memory, such as descendants of non-Jewish European emigrants or of survivors outside Israel, show an identity-route effect that this design holds fixed? A study measuring descendants’ identification with the group, separately from family exposure, would answer the question’s second part.