Articles

Objective and subjective experiences of childhood

maltreatment and their relationships with cognitive deficits:
a cohort study in the USA
Andrea Danese, Cathy Spatz Widom

Summary
Lancet Psychiatry 2024; Background Cognitive deficits might contribute to the elevated risk of life-course psychopathology observed in
11: 720–30 maltreated children. Leading theories about the links between childhood maltreatment and cognitive deficits focus on
Social, Genetic and documented exposures (objective experience), but empirical research has largely relied on retrospective self-reports of
Developmental Psychiatry
these experiences (subjective experience), and the two measures identify largely non-overlapping groups. We aimed
Centre and Department of
Child and Adolescent to test the associations of objective and subjective measures of maltreatment with cognitive abilities within the same
Psychiatry, Institute of individuals.
Psychiatry, Psychology &
Neuroscience, King’s College
Methods We studied a cohort of individuals from the US Midwest with both objective, court-documented evidence of
London, London, UK
(Prof A Danese MD PhD); childhood maltreatment and subjective self-reports of individuals’ histories at age 29 years. Between the ages
National and Specialist CAMHS of 29 years and 41 years, participants were assessed with a comprehensive set of cognitive tests, including tests of
Clinic for Trauma, Anxiety, and general verbal intelligence (Quick Test and Wide Range Achievement Test-Revised [WRAT]), non-verbal intelligence
Depression, South London and
Maudsley NHS Foundation
(Matrix Reasoning Test [MRT]), executive function (Stroop Test and Trail Making Test Part B [TMT-B]), and processing
Trust, London, UK speed (Trail Making Test Part A [TMT-A]). Participants were also assessed for psychopathology (Center for
(Prof A Danese); Psychology Epidemiologic Studies Depression Scale and Beck Anxiety Inventory). We tested the associations between objective or
Department, John Jay College, subjective measures of childhood maltreatment with cognitive functions using ordinary least squares regression. To
City University of New York,
New York, NY, USA
test whether cognitive deficits could explain previously described associations between different measures of
(Prof C S Widom PhD); Graduate maltreatment and subsequent psychopathology, we re-ran the analyses accounting for group differences in the Quick
Center, City University of Test. People with lived experience were not involved in the research or writing process.
New York, New York, NY, USA
(Prof C S Widom)
Findings The cohort included 1196 individuals (582 [48·7%] female, 614 [51·3%] male; 752 [62·9%] White, 417 [34·9%]
Correspondence to:
Prof Andrea Danese, Social,
Black, 36 [3·8%] Hispanic) who were assessed between 1989 and 2005. Of the 1179 participants with available data,
Genetic and Developmental 173 had objective-only measures of childhood maltreatment, 492 had objective and subjective measures, 252 had
Psychiatry Centre, Institute of subjective-only measures, and 262 had no measures of childhood maltreatment. Participants with objective measures
Psychiatry, Psychology & of childhood maltreatment showed pervasive cognitive deficits compared with those without objective measures
Neuroscience, King’s College
London, London SE5 8AF, UK
(Quick Test: β=–7·97 [95% CI –9·63 to –6·30]; WRAT: β=–7·41 [–9·09 to –5·74]; MRT: β=–3·86 [–5·86 to –1·87];
[email protected] Stroop Test: β=–1·69 [–3·57 to 0·20]; TMT-B: β=3·66 [1·67 to 5·66]; TMT-A: β=2·92 [0·86 to 4·98]). The associations
with cognitive deficits were specific to objective measures of neglect. In contrast, participants with subjective
measures of childhood maltreatment did not differ from those without subjective measures (Quick Test: β=1·73
[95% CI –0·05 to 3·50]; WRAT: β=1·62 [–0·17 to 3·40]; MRT: β=0·19 [–1·87 to 2·24]; Stroop Test: β=–1·41
[–3·35 to 0·52]; TMT-B: β=–0·57 [–2·69 to 1·55]; TMT-A: β=–0·36 [–2·38 to 1·67]). Furthermore, cognitive deficits
did not explain associations between different measures of maltreatment and subsequent psychopathology.

Interpretation Previous studies based on retrospective reports of childhood maltreatment have probably grossly
underestimated the extent of cognitive deficits in individuals with documented experiences of childhood maltreatment,
particularly neglect. Psychopathology associated with maltreatment is unlikely to emerge because of cognitive deficits,
but might instead be driven by individual appraisals, autobiographical memories, and associated schemas.

Funding National Institute of Justice, National Institute of Mental Health, Eunice Kennedy Shriver National Institute
of Child Health and Human Development, National Institute on Aging, Doris Duke Charitable Foundation, and
National Institute for Health and Care Research.

Copyright © 2024 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY
4.0 license.

Introduction associations is important because cognitive deficits

Childhood maltreatment has been associated with might contribute to the elevated risk of psychopathology
cognitive deficits, including lower IQ and impairment in observed in maltreated individuals. Cognitive deficits are
specific functions.1–3 A better understanding of these present in individuals at risk for, and patients diagnosed

720 www.thelancet.com/psychiatry Vol 11 September 2024

 Articles

Research in context
Evidence before this study Added value of this study
Patients with depression, anxiety, and psychosis have small-to- This is the first study to test the relative associations of
moderate impairments in a broad range of cognitive domains, objective and subjective measures of childhood maltreatment
which have been posited to contribute to the risk of onset and with cognitive abilities within the same individuals. We studied
poor outcomes across psychiatric diagnoses. Childhood a unique cohort of 1196 children from the US Midwest with
maltreatment is a key risk factor for psychopathology, and it both objective measures of childhood maltreatment from
has been proposed that cognitive deficits could help to explain official court records and subjective measures from
the observed associations. Several papers have tested cognitive retrospective reports at a mean age of 29 years, as well as
abilities in individuals with versus without a history of multiple cognitive assessments in adult life. We found that
childhood maltreatment. We searched Ovid MEDLINE, Embase, participants with objective measures of childhood
and APA PsycInfo (updated to March 11, 2024), using terms maltreatment showed pervasive cognitive deficits in general
relating to childhood maltreatment and cognitive abilities to verbal intelligence, non-verbal intelligence, executive function,
identify previous systematic reviews and meta-analyses of the and processing speed compared with those without objective
literature (appendix p 10). We placed no restrictions on the measures. In contrast, participants with subjective measures of See Online for appendix
publication date or language. We identified only one relevant childhood maltreatment showed cognitive abilities similar to
meta-analysis, which included k=52 samples and those without subjective measures. The associations with
3919 participants. The meta-analysis found that, overall, cognitive deficits were specific to objective measures of neglect
individuals with a history of childhood maltreatment had and did not generalise to physical or sexual abuse. Finally,
moderate impairment in cognitive abilities compared with cognitive deficits did not explain associations between different
those without a history of childhood maltreatment (g=−0·50 measures of maltreatment and subsequent psychopathology.
[95% CI −0·60 to −0·41]), with significant heterogeneity
Implications of all the available evidence
beyond what would be expected based on sampling variation
Because of its reliance on retrospective measures, the current
(Q=291·93, p<0·001). Furthermore, effect sizes were greater in
evidence base regarding the association between childhood
studies where the cognitive assessment was undertaken in
maltreatment and cognitive abilities has probably grossly
children and maltreatment assessment was based on records
underestimated the challenges in children, young people, and
or informant reports (eg, at age 0–5 years; g=−0·66
adults with objective, documented exposures. Cognitive deficits
[−0·97 to −0·35]) compared with studies in adults relying on
might help to explain poor educational and employment
retrospective self-reports of childhood maltreatment (g=−0·56
outcomes and risk for adult antisocial behaviour in individuals
[−0·92 to −0·20]; Qbetween=9·93, p=0·019). The interpretation of
with documented exposure to childhood maltreatment and
these findings is, however, unclear. Although leading theories
should be the target of mitigatory interventions in educational
about the links between childhood maltreatment and
and social care settings. Although cognitive deficits are more
cognitive deficits have focused on documented exposures
likely in individuals with objective rather than subjective
(objective measures), empirical research has largely relied on
measures of childhood maltreatment, psychopathology is more
retrospective self-reports of these experiences (subjective
strongly related to subjective versus objective measures of
measures). Objective and subjective measures of childhood
childhood maltreatment. Consistent with these opposite
maltreatment identify largely non-overlapping groups of
patterns of association, cognitive deficits did not explain the
individuals and are differentially associated with
associations between maltreatment and subsequent
psychopathology. However, it is unclear whether the two
psychopathology. A better understanding of features of the
measures are also differentially associated with cognitive
subjective experience of childhood maltreatment, such as
abilities. Disentangling these associations is necessary to
individual appraisals, autobiographical memories, and schemas,
identify risk mechanisms for psychopathology and to target
could provide novel targets for effective treatments.
relevant interventions.

with, depression,4 post-traumatic stress disorder,5 and retrospective recall of childhood experiences in adults.11
psychosis.6 Cognitive deficits might also contribute to the This is problematic because prospective, documented
poor functioning7 and antisocial behaviour8 observed in measures of childhood maltreatment (indexing the
maltreated individuals. However, fundamental questions objective experience) and retrospective self-reports
remain unanswered. (indexing the subjective experience) identify largely non-
Most theories on the links between childhood overlapping groups of individuals and constructs,12 and are
maltreatment and cognitive abilities propose that differentially associated with psychopathology.13–15 However,
documented exposure to maltreatment triggers toxic it is unclear whether the two measures are also
biological responses leading to brain damage that is differentially associated with cognitive abilities. Disen­
detectable through neuropsychological inquiry.9,10 In tangling these associations is necessary to identify risk
contrast, most empirical research on these links relies on mechanisms for psychopathology, functional impairment,

www.thelancet.com/psychiatry Vol 11 September 2024 721

 Articles

and antisocial behaviour in maltreated children, and to Measures of childhood maltreatment

target relevant interventions. Details of the measures of objective and subjective
To characterise the differential associations of experiences of childhood maltreatment are reported in
objective and subjective experience of childhood the appendix (pp 6–8).
maltreatment with cognitive deficits, we studied a To identify objective experience of childhood
unique cohort13,14 with both objective measures of maltreatment, official reports of child abuse or neglect,
childhood maltreatment from official court records and based on records of county juvenile (family) and adult
subjective measures from retrospective reports at a criminal courts between 1967 and 1971, were used. Only
mean age of 29 years, as well as multiple assessments court-substantiated cases involving children younger
of cognitive functions and psychopathology in adult than 12 years at the time of abuse or neglect were
life. We capitalised on this sample to test two competing included. Findings based on this objective measure of
a-priori hypotheses. First, if cognitive deficits are more maltreatment have been replicated by several, more
strongly associated with objective experience, then contemporaneous cohorts, showing that (1) the
cognitive abilities should be impaired in participants construct of childhood maltreatment captured in our
with official records of childhood maltreatment, sample is similar to the construct captured by other,
regardless of whether they also retro­spectively report it. more recent measures, and (2) the findings are relevant
Second, and alternatively, if cognitive deficits are more to modern society. Of note, substantiated court records
strongly associated with subjective experience, then provide the legal standard on which child protection
cognitive abilities should be impaired in participants actions are based and, thus, provide the strongest
who retrospectively report a history of childhood possible evidence for the objective experience of
maltreatment, regardless of whether they were also childhood maltreatment.
identified through official records. Furthermore, we To identify subjective experience of childhood
tested whether cognitive deficits could explain maltreatment, retrospective self-reported measures were
previously described associations between different chosen to include a broad set of maltreatment experiences
measures of maltreatment and subsequent representative of the experiences cited in the original
psychopathology.14 (objective) court cases. During interview 1, participants
were asked to respond about experiences that occurred
Methods before age 12 years to make the retrospective reports as
Study design and participants similar as possible to the court cases. In addition, because
This prospective cohort study was initiated in 1986 with no single retrospective assessment measure is universally
a large group of documented cases of childhood endorsed by researchers, multiple measures of each type
physical and sexual abuse and neglect (n=908) and a of maltreatment were included to be as comprehensive as
comparison group that did not contain any known possible. Four measures were used to assess self-reports
cases of child abuse or neglect and was matched on the of childhood sexual abuse during interview 1, all of which
basis of age, sex, race or ethnicity, and approximate were adapted from previous work.13 Two measures were
family social class at the time of the childhood used to assess retrospective self-reports of childhood
maltreatment (n=667).16 Details regarding sample physical abuse during interview 1: the Conflict Tactics
selection are reported in the appendix (pp 2–5). Scale and the Self-Report of Childhood Abuse Physical.
For this paper, we use data from three follow-up Retrospective assessments of neglect were more
phases, which involved tracing, locating, and challenging because, at the time these retrospective
interviewing the abused or neglected children and reports were collected, the field lacked a validated neglect
comparison group members in 1989–1995 (interview 1), instrument. Lacking such an instrument at the time,
in 2000–02 (interview 2), and in 2003–05 (interview 3). questions were designed to cover a range of neglect
All interviews took place in the participant’s home or experiences (ie, inadequate provision of food, clothing,
other convenient place. The interviewers were not shelter, and supervision) that were similar to the charges
aware of the purpose of the study and the inclusion of in the official neglect petition. Of note, measures of child
an abused or neglected group. Similarly, the participants neglect that are currently commonly used, such as the
were not aware of the purpose of the study and were Conflict Tactics Scale or the Childhood Trauma
told they had been selected to participate as part of a Questionnaire, ask questions that are not dissimilar to
large group of individuals who grew up in that area in those used in our study.
the late 1960s and early 1970s. All participants signed a
consent form acknowledging that they were Measures of cognitive function
participating voluntarily. Institutional Review Board The Quick Test and the Wide Range Achievement Test-
approval was obtained at each wave of the study and Revised Level II (WRAT) were assessed at interview 1
participants provided written, informed consent. People (1989–95). The Stroop Test was assessed at interview 2
with lived experience were not involved in the research (2000–02). The Trail Making Test (Parts A and B; TMT-A
or writing process. and TMT-B) and the Weschler Adult Intelligence

722 www.thelancet.com/psychiatry Vol 11 September 2024

 Articles

Scale—Third Edition’s Matrix Reasoning Test (MRT) maltreatment and cognitive functions using ordinary
were assessed at interview 3 (2003–05). least squares (ie, linear) regression. We then identified
The Quick Test is a measure of receptive vocabulary four groups: (1) participants who were identified as
and verbal information processing that can be used to victims of childhood maltreatment by virtue of official
approximate IQ.17 The Quick Test is easily administered records but did not retrospectively recall the experience
and shows good reliability and validity.18 The Quick Test (objective-only measure); (2) participants who were
has an average normed score of 100 (SD 10). The WRAT identified as victims of childhood maltreatment by virtue
is a measure of reading level and classroom academic of official records and also retrospectively recalled
achievement with adequate concurrent, content, and the experience (objective and subjective measure);
construct validity.19 (3) participants who retrospectively recalled being
The MRT is a measure of non-verbal reasoning.20 In the maltreated in childhood but were not identified as
MRT, participants must recognise a pattern and select its victims of childhood maltreatment in official records
missing part from several different options. Higher (subjective-only measure); and (4) participants who did
scores on the MRT represent correctly completed items not have official records or retrospective recall (no
and better performance. measure). We tested differences in cognitive functions
The Stroop Test is a measure of cognitive control and between participants with no measure of maltreatment
inhibition.21 The Stroop Test assesses the ease with which and those in the three groups with objective or subjective
an individual keeps a goal in mind and suppresses a measures (or both) using ordinary least squares
routine response while prioritising an uncommon regression to examine relative differences associated
response. The Stroop Test Interference Score was with either measure compared with the same baseline.
computed according to Golden’s methods.22 Higher Missing values in cognitive measures collected at
scores on the Stroop Test represent greater ability to interviews 2 and 3 (Stroop Test, TMT, and MRT) were
inhibit interference and better performance. imputed assuming missing at random (MAR) or missing
The TMT-A is a measure of processing speed.23 In the completely at random (MCAR) mechanisms through
TMT-A, participants must draw lines between 25 randomly multiple imputation by chained equations in Stata 18,
arranged numbers in sequential order. Higher scores using 20 imputed datasets combined through Rubin’s
indicate more time needed to complete the task and rules. All statistical tests were two-sided. All analyses
indicate poorer performance. In contrast, the TMT-B is were carried out in Stata 18. To test the sensitivity of the
a measure of cognitive flexibility.23 In the TMT-B, results to the MAR or MCAR assumptions, we also re-
participants must draw lines between 25 randomly ran group comparisons on the overall maltreatment
arranged numbers and letters in alternating order of measures in the subset of participants with complete
letters or numbers. Higher scores indicate more time data.
needed to complete the task and indicate poorer
performance. TMT-A and TMT-B scores were natural log- Records Interview 1 Interview 2 Interview 3
transformed to correct for high positive skew.
Dates 1967–71 1989–95 2000–02 2003–05
Sample size 1575 1196 896 807
Measures of psychopathology
Mean age at interview, years (SD) ·· 29·2 (3·8) 39·5 (3·5) 41·2 (3·5)
As in our previous work, to measure the course of
14

Mean age at court petition for 6·4 (3·3) 6·3 (3·3) 6·2 (3·3) 6·3 (3·3)
depression and anxiety after the assessment of the
childhood maltreatment, years (SD)
subjective experience of childhood maltreatment
Sex
(interview 1), we examined whether participants met the
Female 799 (50·7%) 582 (48·7%) 457 (51·0%) 425 (52·7%)
clinical symptom severity threshold for depression or
Male 776 (49·3%) 614 (51·3%) 439 (49·0%) 382 (47·3%)
anxiety diagnosis across the subsequent interviews.
Race or ethnicity*
Depression was assessed at interviews 2 and 3 using the
White 1042 (66·1%) 752 (62·9%) 557 (62·2%) 487 (60·4%)
Center for Epidemiologic Studies Depression Scale
Black 515 (32·6%) 417 (34·9%) 315 (35·2%) 301 (37·3%)
(CES-D)24 with a past-week reporting period. Anxiety was
Hispanic 5 (0·3%) 36 (3·8%) 36 (4·0%) 32 (4·0%)
assessed at interviews 2 and 3 using the Beck Anxiety
Abuse or neglect† 908 (57·7%) 676 (56·5%) 500 (55·8%) 458 (56·8%)
Inventory (BAI)25 with a past-week reporting period. The
number of depressive or anxiety episodes was defined as Any physical abuse 160 (10·2%) 110 (9·2%) 79 (8·8%) 78 (9·7%)

the count of assessments when participants met the Any neglect 697 (44·3%) 543 (45·4%) 406 (45·3%) 370 (45·9%)

symptom severity clinical threshold for depression Any sexual abuse 153 (9·7%) 96 (8·0%) 68 (7·6%) 61 (7·5%)
(CES-D cutoff >20) or anxiety (BAI cutoff >16) across *Race or ethnicity information was not available for 13 individuals in the official records. For interviews, the
interviews 2 and 3. percentages add up to more than 100 because participants self-identified race and had the option to reply to race and
ethnicity separately and could have reported being both Black and Hispanic, for example. †The numbers of cases of
specific types of abuse and neglect add up to more than the total number in the abuse or neglect group total (n=676 at
Statistical analysis interview 1) because approximately 11% of the sample had more than one type of abuse or neglect.
To test group differences, we first tested the associations
Table: Characteristics of the study sample over four waves
between objective or subjective measures of childhood

www.thelancet.com/psychiatry Vol 11 September 2024 723

 Articles

To test the sensitivity of the results, we re-ran group

comparisons controlling for key matching variables
A
originally used to develop this sample (age, sex, and race
10 Objective measures
or ethnicity)26 and using individual types of maltreatment
as independent measures. To test whether cognitive
deficits could explain previously described associations
5
Maltreatment versus no maltreatment (β)

between different measures of maltreatment and

subsequent psychopathology, we re-ran the main
analyses and any statistically significant sensitivity
0 analyses accounting for group differences in the Quick
Test. We selected the Quick Test because it is the broadest
measure of cognitive functions we collected, had the
–5 most observed values, and was assessed before the
psychopathology outcomes, thereby providing optimal
timing to examine mediation effects. These tests rely on
–10 Poisson regression models, and differences in estimates
were examined through an interaction test.27
To facilitate interpretation and comparisons, the raw
scores from all cognitive measures were standardised
B with a mean of 100 and an SD of 15. Group differences
10 Subjective measures are presented here as β (95% CI) or IRR (95% CI) from
the regression analyses. Group differences in cognitive
abilities are also described as SD differences to facilitate
5
interpretation. A p-value of less than 0·05 is regarded as
Maltreatment versus no maltreatment (β)

significant.

Role of the funding source

0
The funders of the study had no role in study design,
data collection, data analysis, data interpretation, or
writing of the report.
–5

Results
The cohort included 1196 individuals who were assessed
–10 between 1989 and 2005. Characteristics of participants at
each wave of the study (including the 1575 individuals
who were originally identified in 1986) are shown in the
table. Despite attrition associated with death, refusals,
C and inability to locate individuals over the various waves
10 Objective-only measures of the study, the composition of the sample remained
Objective and subjective measures about the same, except that females were more likely to
Subjective-only measures
participate in the 2003–05 interview than males (β=0·50,
5 p<0·001).
Maltreatment versus no maltreatment (β)

Of the 1179 participants with available data on

maltreatment measures, 173 had objective-only measures
0
of childhood maltreatment, 492 had objective and

–5
Figure 1: Differences in cognitive test scores between study groups
identified by childhood maltreatment variables
(A) Associations of objective measures with cognitive test scores (objective
–10 measures versus no objective measures). (B) Associations of subjective measures
with cognitive test scores (subjective measures versus no subjective measures).
(C) Relative associations of objective-only, objective and subjective, and
subjective-only measures (versus no measures) with cognitive test scores. All
Quick Test WRAT MRT Stroop Test TMT-B TMT-A
cognitive measures were standardised with a mean of 100 and an SD of 15. Error
(age 29 years) (age 29 years) (age 41 years) (age 39 years) (age 41 years) (age 41 years) bars display 95% CIs. WRAT=Wide Range Achievement Test-Revised.
Cognitive test MRT=Matrix Reasoning Test. TMT-B=Trail Making Test Part B. TMT-A=Trail
Making Test Part A.

724 www.thelancet.com/psychiatry Vol 11 September 2024

 Articles

subjective measures, 252 had subjective-only measures,

A
and 262 had no measures of childhood maltreatment. 10 Objective measures
With regard to the cognitive tests, there were 1185 valid
observations for the Quick Test (with a mean score of
90 [SD 13]), 1178 for the WRAT (mean score 49 [16]),
5

Physical abuse versus no physical abuse (β)

874 for the Stroop Test (mean score –2·4 [8·5]), 782 for
the TMT-A (mean time 37 s [19]), 770 for the TMT-B
(mean time 88 s [48]), and 792 for the MRT (mean score
14 [6]). Correlations among the tests are shown in the 0

appendix (p 28). Missing values for Stroop Test, TMT,

and MRT were imputed, resulting in an analytical sample
of 1179. –5
Study participants with official records (objective
experience) of childhood maltreatment showed, on
average, cognitive deficits in most tests compared with –10
participants without objective experience of childhood
maltreatment, with differences ranging from half an SD
on the Quick Test and WRAT to one-fifth of an SD on the
TMT-A (Quick Test: β=–7·97 [95% CI –9·63 to –6·30];
B
WRAT: β=–7·41 [–9·09 to –5·74]; MRT: β=–3·86
10 Subjective measures
[–5·86 to –1·87]; TMT-B: β=3·66 [1·67 to 5·66]; TMT-A:
β=2·92 [0·86 to 4.98]; figure 1A; appendix p 13). There was
no significant difference on the Stroop Test (β=–1·69
[–3·57 to 0·20]). In contrast, participants with retrospective 5
Physical abuse versus no physical abuse (β)

recall (subjective experience) of childhood maltreatment

did not show cognitive deficits compared with participants
without subjective experience of childhood maltreatment 0
(Quick Test: β=1·73 [–0·05 to 3·50]; WRAT: β=1·62
[–0·17 to 3·40]; MRT: β=0·19 [–1·87 to 2·24]; Stroop Test:
β=–1·41 [–3·35 to 0·52]; TMT-B: β=–0·57 [–2·69 to 1·55]; –5
TMT-A: β=–0·36 [–2·38 to 1·67]; figure 1B; appendix p 13).
The agreement of measures based on official records or
retrospective recall of childhood maltreatment was low
–10
(Cohen κ=0·25; appendix p 14), highlighting the need to
test the relative contribution of the two different constructs
by investigating four subgroups.
Compared with participants with neither objective nor
subjective experience of maltreatment (no maltreatment), C
10 Objective-only measures
those with objective-only experience showed cognitive Objective and subjective measures
deficits in most tests, with differences ranging from Subjective-only measures
two-thirds of an SD on the Quick Test and WRAT, to
one-quarter of an SD on the TMT-A (Quick Test: β=–9·72 5
Physical abuse versus no physical abuse (β)

[95% CI –12·47 to –6·97]; WRAT: β=–9·50 [–12·30 to –7.34];

MRT: β=–5·87 [–9·25 to –2·48]; TMT-B: β=5·72
[2·37 to 9·07]; TMT-A: β=4·34 [0·82 to 7·86]; figure 1C; 0

–5
Figure 2: Differences in cognitive test scores between study groups
identified by childhood physical abuse variables
(A) Associations of objective measures with cognitive test scores (objective
measures versus no objective measures). (B) Associations of subjective measures –10
with cognitive test scores (subjective measures versus no subjective measures).
(C) Relative associations of objective-only, objective and subjective, and
subjective-only measures (versus no measures) with cognitive test scores. All
cognitive measures were standardised with a mean of 100 and an SD of 15. Error Quick Test WRAT MRT Stroop Test TMT-B TMT-A
bars display 95% CIs. WRAT=Wide Range Achievement Test-Revised. (age 29 years) (age 29 years) (age 41 years) (age 39 years) (age 41 years) (age 41 years)
MRT=Matrix Reasoning Test. TMT-B=Trail Making Test Part B. TMT-A=Trail Cognitive test
Making Test Part A.

www.thelancet.com/psychiatry Vol 11 September 2024 725

 Articles

appendix p 15). There was no significant difference on the

A
10 Objective measures Stroop Test (β=–1·92 [–5·12 to 1·28]). Compared with
participants with no maltreatment, those with objective
and subjective experiences showed some cognitive deficits,
with differences of one-third of an SD on the Quick Test
5
and WRAT and one-fifth of an SD on the Stroop Test and
Sexual abuse versus no sexual abuse (β)

TMT-B (Quick Test: β=–4·75 [–6·89 to –2·60]; WRAT:

β=–4·57 [–6·75 to -2·40]; MRT: β=–3·19 [–5·84 to –0·54];
0 Stroop Test: β=–2·58 [–5·03 to –0·14]; TMT-B: β=2·76
[0·12 to 5·39]; figure 1C; appendix p 15). There was no
significant difference on the TMT-A (β=2·33 [–0·45 to
–5 5·11]). Finally, compared with participants with no
maltreatment, those with subjective-only experience
showed a small elevation in some cognitive functions, with
a difference of one-fifth of an SD on the Quick Test and
–10
WRAT (Quick Test: β=3·55 [1·07 to 6·02]; WRAT: β=2·96
[0·46 to 5·47]; figure 1C; appendix p 15). There were no
significant differences on the other tests (MRT: β=–0·07
[–3·15 to 3·00]; Stroop Test: β=–1·38 [–4·25 to 1·49];
B
10 Subjective measures TMT-B: β=–0·48 [–3·81 to 2·85]; TMT-A: β=–0·50 [–3·45
to 2·45]).
Similar findings to those in imputed datasets, described
above, were observed in analyses in complete cases
5
(appendix pp 16–17). Similar findings were also observed
Sexual abuse versus no sexual abuse (β)

when controlling for key matching variables originally

used to develop this sample (appendix pp 18–19). Because
0 the agreement between objective and subjective
measures was inconsistent across different types of
childhood maltreatment (appendix p 14), we tested
–5 whether the associations of objective or subjective
experiences of childhood maltreatment with cognitive
functions varied as a function of maltreatment type.
Neither participants with objective experience nor
–10
participants with subjective experience of childhood
physical abuse showed cognitive deficits across all tests
(figure 2A, B; appendix p 20). Testing the relative
contribution of the objective and subjective experiences,
C
10 Objective-only measures
no differences in cognitive functions were observed
Objective and subjective measures between participants with objective-only, objective and
Subjective-only measures subjective, or subjective-only experiences of childhood
physical abuse and participants with no childhood
5 physical abuse (figure 2C; appendix p 21).
Sexual abuse versus no sexual abuse (β)

Neither participants with objective experience nor

participants with subjective experience of childhood
0 sexual abuse showed cognitive deficits across all tests

–5
Figure 3: Differences in cognitive test scores between study groups
identified by childhood sexual abuse variables
(A) Associations of objective measures with cognitive test scores (objective
–10 measures versus no objective measures). (B) Associations of subjective measures
with cognitive test scores (subjective measures versus no subjective measures).
(C) Relative associations of objective-only, objective and subjective, and
subjective-only measures (versus no measures) with cognitive test scores. All
Quick Test WRAT MRT Stroop Test TMT-B TMT-A
cognitive measures were standardised with a mean of 100 and an SD of 15. Error
(age 29 years) (age 29 years) (age 41 years) (age 39 years) (age 41 years) (age 41 years) bars display 95% CIs. WRAT=Wide Range Achievement Test-Revised.
MRT=Matrix Reasoning Test. TMT-B=Trail Making Test Part B. TMT-A=Trail
Cognitive test
Making Test Part A.

726 www.thelancet.com/psychiatry Vol 11 September 2024

 Articles

(figure 3A, B; appendix p 22). Testing the relative A

contribution of the objective and subjective experiences, 10 Objective measures
no differences in cognitive functions were observed
between participants with objective-only, objective and
subjective, or subjective-only experiences of childhood
5
sexual abuse and participants with no childhood sexual
abuse (figure 3C; appendix p 23).

Neglect versus no neglect (β)

Participants with objective experience of childhood
neglect showed, on average, cognitive deficits in most 0

tests compared with participants without objective

experience of childhood neglect, with differences ranging
from half an SD on the Quick Test and WRAT to about –5
one-seventh of an SD on the TMT-A, and no significant
differences on the Stroop Test (figure 4A; appendix p 24).
In contrast, participants with subjective experience of –10
childhood neglect did not show any cognitive deficits
compared with participants without subjective experience
of childhood neglect (figure 4B; appendix p 24).
Compared with participants with no neglect, those
B
with objective-only experience of neglect showed, on 10 Subjective measures
average, cognitive deficits in most tests, with differences
of about two-thirds of an SD on the Quick Test and
WRAT, to about one-third of an SD on the MRT and
5
TMT-B, and no significant differences on the Stroop
Test (figure 4C; appendix p 25). Compared with
Neglect versus no neglect (β)

participants with no neglect, those with objective and

subjective experiences of neglect showed some cognitive 0

deficits, with differences of one-third of an SD on the

Quick Test and WRAT, and one-fifth of an SD on the
MRT and TMT-B. Finally, compared with participants –5
with no neglect, those with subjective-only experience of
neglect showed no significant changes in cognitive
functions. –10
As shown in our previous work,14 participants with
objective and subjective or subjective-only measures of
childhood maltreatment had a greater number of
subsequent follow-up phases with depression and
C
anxiety than controls over the following decade, but 10 Objective-only measures
participants with objective-only measures did not Objective and subjective measures
Subjective-only measures
(figure 5; appendix p 26). We tested whether cognitive
functions could explain these associations. In analyses
5
adjusted for Quick Test scores, the strength of the
aforementioned associations was similar to that of
Neglect versus no neglect (β)

unadjusted analyses, with no significant differences

(figure 5; appendix p 26). The same results were observed 0

for childhood neglect (appendix p 27).

–5

Figure 4: Differences in cognitive test scores between study groups

identified by childhood neglect variables
(A) Associations of objective measures with cognitive test scores (objective –10
measures versus no objective measures). (B) Associations of subjective measures
with cognitive test scores (subjective measures versus no subjective measures).
(C) Relative associations of objective-only, objective and subjective, and subjective-
only measures (versus no measures) with cognitive test scores. All cognitive Quick Test WRAT MRT Stroop Test TMT-B TMT-A
measures were standardised with a mean of 100 and an SD of 15. Error bars display (age 29 years) (age 29 years) (age 41 years) (age 39 years) (age 41 years) (age 41 years)
95% CIs. WRAT=Wide Range Achievement Test-Revised. MRT=Matrix Reasoning Cognitive test
Test. TMT-B=Trail Making Test Part B. TMT-A=Trail Making Test Part A.

www.thelancet.com/psychiatry Vol 11 September 2024 727

 Articles

First, the results might be affected by misclassification.

A Depressive episodes
4 Because objective measures have low sensitivity, they
Analysis
Unadjusted might not have captured all study participants with an
Adjusted objective experience of childhood maltreatment. Therefore,
some participants who had objective-only measures would
3 have been classified as having no measure of childhood
maltreatment. This misclass­ification would have led to
Incidence rate ratio (95% CI)

artifactual lowering of cognitive functions in the control

group, with underestimation of effect sizes across all
2 comparisons. Furthermore, some participants who had
objective and subjective measures would have been
classified as having a subjective-only measure of childhood
maltreatment. As participants with objective and subjective
1 measures had more cognitive deficits than those with a
subjective-only measure, this misclassification would have
led to an overestimation of cognitive deficits in participants
with a subjective-only measure. Another type of misclass­
0
ification might have arisen from non-disclosure of a
B Anxiety episodes subjective experience of childhood maltreatment (eg,
4 because of shame or fear of harming or being harmed by
the perpetrator). Therefore, some participants who had
objective and subjective measures would have been
classified as having an objective-only measure of childhood
3 maltreatment. As participants with objective and subjective
measures had similar or less severe cognitive deficits than
Incidence rate ratio (95% CI)

those with an objective-only measure, this misclassification

would have either not affected the overall conclusions or
2 led to an underestimation of cognitive deficits among
those with an objective-only measure. Furthermore, some
participants who had a subjective-only experience would
have been classified as having no measure of childhood
1
maltreatment. As participants with subjective-only
measures had similar cognitive function to those with no
measure of childhood maltreatment, this type of
misclassification would not have affected the overall
0
Objective only Objective and subjective Subjective only conclusions. As such, none of these potential mechanisms
Measure of maltreatment of misclassification is likely to have significantly biased our
interpretation of the results.
Figure 5: Associations between study groups identified by childhood
maltreatment variables and psychopathology Second, the observational study design did not enable us
Analyses unadjusted or adjusted for cognitive function (indexed here by Quick to infer causal effects: objective experience of maltreatment
Test at age 29 years). (A) Relative associations of objective and subjective might have caused cognitive deficits; cognitive deficits
measures with the number of depressive episodes. (B) Relative associations of
might have increased the risk of maltreatment;
objective and subjective measures with the number of anxiety episodes.
unmeasured variables might have led to spurious
associations between the objective experience of
Discussion maltreatment and cognitive deficits.3,28 Third, the cognitive
In this prospective longitudinal study, we found that assessment undertaken in this cohort did not include a
cognitive deficits are concentrated among individuals specific assessment of memory function. Participants
with objective experience of childhood maltreatment with subjective measures might have had normal or
and not observed in those with subjective measures of heightened autobiographical memory, enabling them to
maltreatment. Cognitive deficits did not explain the better recall childhood experiences of maltreatment. Test
associations between different maltreatment measures scores for memory correlate moderately with test scores
and later psychopathology. Differential associations of for general intelligence and executive functions like those
objective and subjective experiences of maltreatment measured in this cohort.3 Fourth, because the objective
with cognitive deficits point to differences between measures were based on court records, we assume that
the constructs identified by either measure.11 The the maltreatment cases were all severe enough to come to
findings should be interpreted in the context of study the attention of the court but we do not have more
limitations. granular measures to discriminate severity within the

728 www.thelancet.com/psychiatry Vol 11 September 2024

 Articles

group of cases. Fifth, sex-specific analyses were not (r approximately 0·2),34 that cognition and psychopathology
performed as there was no clear indication in the literature appear as independent constructs in structural models,35
that the associations tested would be different in males and that treatments that are effective in reducing
and females. Sixth, people with lived experience were not psychopathological symptoms might not improve
involved in the research or writing process. Despite these cognitive functions.36,37 Therefore, psycho­pathology, which
limitations, the findings have implications for is more strongly associated with the subjective experience
understanding the associations between childhood of childhood maltreatment, is unlikely to emerge because
maltreatment and cognitive functions and their of cognitive deficits. Instead, better understanding of
explanatory role in psychopathology. features of the subjective experience, such as individual
Despite notable exceptions,3,29–31 most cognitive and appraisals, autobio­graphical memories, and schemas,
cognitive neuroscience studies on childhood maltreatment could provide novel targets for effective treatments.11
have relied on retrospective reports in adulthood. The Contributors
reliance on the subjective experience of childhood AD contributed to conceptualisation, formal analysis, investigation,
maltreatment is likely to have grossly underestimated the methodology, project administration, visualisation, writing the original
draft, and reviewing and editing the manuscript. CSW contributed to
extent of cognitive and cognitive neuroscience findings in conceptualisation, data curation, funding acquisition, investigation,
individuals with prospective and documented measures of methodology, project administration, and reviewing and editing the
maltreatment. This cautionary observation is consistent manuscript. AD and CSW directly accessed and verified the underlying
with—and substantially expands through the use of official data reported in the manuscript, take responsibility for the integrity of
the data and the accuracy of the data analysis, and had final
court records—previous findings about the associations of responsibility for the decision to submit for publication.
prospective measures of childhood maltreatment with
Declaration of interests
cognitive deficits3 and structural brain-imaging findings32 We declare no competing interests.
in other cohorts. The findings suggest that more attention
Data sharing
should be paid to cognitive abilities in children with The data reported in the current article are not publicly available because
documented histories of maltreatment, and particularly they contain extremely sensitive information that could compromise
neglect, to mitigate negative consequences in education research participant privacy and confidentiality. We cannot provide
and employment7 and to reduce the risk of adult antisocial individual-level data from this project because our confidentiality
agreement with the participants in this study precludes this.
behaviour.8
Acknowledgments
The observed associations between maltreatment and
This research was supported in part by grants from the National Institute
cognitive deficits in our sample emerged because of of Justice (86-IJ-CX-00333), the National Institute of Mental Health
specific links between childhood neglect and cognitive (MH49467 and MH58386), the Eunice Kennedy Shriver National Institute
deficits and did not generalise to childhood physical or of Child Health and Human Development (HD40774), the National
Institute on Aging (AG058683), and the Doris Duke Charitable Foundation
sexual abuse. Because of the marked overlap between
to CSW. The opinions, findings, and conclusions or recommendations
family poverty and neglect in particular,32 the specificity expressed are those of the authors and do not necessarily reflect those of
might reflect confounding by family poverty, which the US Department of Justice or agencies of the US National Institutes of
should be investigated in future studies. Because the Health. AD received funding from the National Institute for Health and
Care Research (NIHR) Biomedical Research Centre at South London and
prevalence of neglect was greater than the prevalence of
Maudsley National Health Service (NHS) Foundation Trust and King’s
physical or sexual abuse in our sample, the specificity College London (NIHR203318) and is supported by Medical Research
might also reflect limited statistical power to detect Council grant P005918. The content is solely the responsibility of the
differences between those with and without abuse authors and does not necessarily represent the official views of the NHS,
the NIHR, or the UK Department of Health and Social Care.
subtypes. However, the group differences in cognition
linked to neglect were also greater than those linked to References
1 Perez CM, Widom CS. Childhood victimization and long-term
physical or sexual abuse. The specificity of these intellectual and academic outcomes. Child Abuse Negl 1994;
associations is in line with previous work emphasising 18: 617–33.
the distinction between abuse and neglect.33 Measure­ 2 Masson M, Bussières E-L, East-Richard C, R-Mercier A, Cellard C.
Neuropsychological profile of children, adolescents and adults
ment heterogeneity in previous cognitive and cognitive experiencing maltreatment: a meta-analysis. Clin Neuropsychol 2015;
neuroscience studies (ie, the mixture of studies based on 29: 573–94.
childhood physical abuse, sexual abuse, and neglect) 3 Danese A, Moffitt TE, Arseneault L, et al. The origins of cognitive
deficits in victimized children: implications for neuroscientists and
might hamper replication of the findings. clinicians. Am J Psychiatry 2017; 174: 349–61.
Cognitive deficits and psychopathology are associated 4 Rock PL, Roiser JP, Riedel WJ, Blackwell AD. Cognitive impairment
with different measures of maltreatment (ie, objective and in depression: a systematic review and meta-analysis. Psychol Med
2014; 44: 2029–40.
subjective measures, respectively), suggesting that the risk
5 Scott JC, Matt GE, Wrocklage KM, et al. A quantitative meta-
pathways for the two outcomes are, at least partly, separate. analysis of neurocognitive functioning in posttraumatic stress
In our analyses, cognitive deficits did not explain disorder. Psychol Bull 2015; 141: 105–40.
associations between different measures of maltreatment 6 Kahn RS. On the origins of schizophrenia. Am J Psychiatry 2020;
177: 291–97.
and subsequent psychopathology. This is consistent 7 Jaffee SR, Ambler A, Merrick M, et al. Childhood maltreatment
with evidence that cognition and psycho­ pathology predicts poor economic and educational outcomes in the transition
show only small correlations in population studies to adulthood. Am J Public Health 2018; 108: 1142–47.

www.thelancet.com/psychiatry Vol 11 September 2024 729

 Articles

8 Nikulina V, Widom CS. Higher levels of intelligence and executive 23 Reitan RM. Validity of the Trail Making Test as an indicator of
functioning protect maltreated children against adult arrests: organic brain damage. Percept Mot Skills 1958; 8: 271–76.
a prospective study. Child Maltreat 2019; 24: 3–16. 24 Radloff LS. The CES-D Scale: a self-report depression scale for
9 Shonkoff JP, Boyce WT, McEwen BS. Neuroscience, molecular research in the general population. Appl Psychol Meas 1977;
biology, and the childhood roots of health disparities: building a 1: 385–401.
new framework for health promotion and disease prevention. 25 Beck AT, Epstein N, Brown G, Steer RA. An inventory for
JAMA 2009; 301: 2252–59. measuring clinical anxiety: psychometric properties.
10 Danese A, McEwen BS. Adverse childhood experiences, allostasis, J Consult Clin Psychol 1988; 56: 893–97.
allostatic load, and age-related disease. Physiol Behav 2012; 26 Pearce N. Analysis of matched case-control studies. BMJ 2016;
106: 29–39. 352: i969.
11 Danese A. Annual Research Review: rethinking childhood 27 Altman DG, Bland JM. Interaction revisited: the difference between
trauma—new research directions for measurement, study design two estimates. BMJ 2003; 326: 219.
and analytical strategies. J Child Psychol Psychiatry 2020; 61: 236–50. 28 Fisher JH, Widom CS. Child maltreatment and cognitive and
12 Baldwin JR, Reuben A, Newbury JB, Danese A. Agreement between academic functioning in two generations. Child Abuse Negl 2021;
prospective and retrospective measures of childhood maltreatment: 115: 105011.
a systematic review and meta-analysis. JAMA Psychiatry 2019; 29 Sonuga-Barke EJS, Kennedy M, Kumsta R, et al. Child-to-adult
76: 584–93. neurodevelopmental and mental health trajectories after early life
13 Danese A, Widom CS. Objective and subjective experiences of child deprivation: the young adult follow-up of the longitudinal English
maltreatment and their relationships with psychopathology. and Romanian Adoptees study. Lancet 2017; 389: 1539–48.
Nat Hum Behav 2020; 4: 811–18. 30 Humphreys KL, King LS, Guyon-Harris KL, et al. Foster care leads
14 Danese A, Widom CS. Associations between objective and to sustained cognitive gains following severe early deprivation.
subjective experiences of childhood maltreatment and the course of Proc Natl Acad Sci USA 2022; 119: e2119318119.
emotional disorders in adulthood. JAMA Psychiatry 2023; 31 Gehred MZ, Knodt AR, Ambler A, et al. Long-term neural
80: 1009–16. embedding of childhood adversity in a population-representative
15 Baldwin JR, Coleman O, Francis ER, Danese A. Prospective and birth cohort followed for 5 decades. Biol Psychiatry 2021; 90: 182–93.
retrospective measures of child maltreatment and their association 32 Drake B, Pandey S. Understanding the relationship between
with psychopathology: a systematic review and meta-analysis. neighborhood poverty and specific types of child maltreatment.
JAMA Psychiatry 2024; published online May 1. https://doi. Child Abuse Negl 1996; 20: 1003–18.
org/10.1001/jamapsychiatry.2024.0818.
33 Berman IS, McLaughlin KA, Tottenham N, et al. Measuring early
16 Widom CS. The cycle of violence. Science 1989; 244: 160–66. life adversity: a dimensional approach. Dev Psychopathol 2022;
17 Ammons RB, Ammons CH. The Quick Test (QT): provisional 34: 499–511.
manual. Psychol Rep 1962; 11: 111–61. 34 Caspi A, Houts RM, Belsky DW, et al. The p factor: one general
18 Zagar RJ, Kovach JW, Busch KG, et al. Ammons Quick Test validity psychopathology factor in the structure of psychiatric disorders?
among randomly selected referrals. Psychol Rep 2013; 113: 823–54. Clin Psychol Sci 2014; 2: 119–37.
19 Jastak S, Wilkinson G. Wide Range Achievement Test-Revised: 35 Rotstein A, Fund S, Levine SZ, Reichenberg A, Goldenberg J.
administration manual. Wilmington, DE: Jastak Associates, 1984. Is cognition integral to psychopathology? A population-based cohort
20 Wechsler D. WAIS–III administration and scoring manual. study. Psychol Med 2023; 53: 7350–57.
San Antonio, TX: Psychological Corporation, 1997. 36 Kahn RS, Keefe RSE. Schizophrenia is a cognitive illness: time for a
21 Strauss E, Sherman EMS, Spreen O. A compendium of change in focus. JAMA Psychiatry 2013; 70: 1107–12.
neuropsychological tests: administration, norms, and commentary, 37 Semkovska M, Quinlivan L, O’Grady T, et al. Cognitive function
3rd edn. New York, NY: Oxford University Press, 2006. following a major depressive episode: a systematic review and meta-
22 Golden C. Stroop Color and Word Test: a manual for clinical and analysis. Lancet Psychiatry 2019; 6: 851–61.
experimental uses. Chicago, IL: Stoelting Co, 1978.

730 www.thelancet.com/psychiatry Vol 11 September 2024