When Collateral Accounts Conflict: What the Drum Teacher Heard

October 11, 2026

Historical accuracy, informant disagreement, and clinical inference in pediatric neuropsychological assessment

Eli and his family are fictional, and every test score in this article is illustrative.

Thursdays

Consider a thirteen-year-old boy named Eli, and begin with the adult who knows him in the narrowest way. His drum teacher has seen him for half an hour every Thursday for four years, which comes to roughly two hundred lessons in the same small room. Eight weeks into the school year the teacher telephones each of Eli’s parents to say that something is different and that he cannot name it.

Eli’s playing is still technically sound, but he drifts in the middle of pieces he used to attack and flinches at corrections he once would have argued with. He has stopped joking between songs, and the teacher, who first noticed the change about five weeks into the term, watched for three more Thursdays before deciding that it was not a small one.

The call arrives fourteen months into a divorce in which a family business, two houses, and both parents’ professional standing are in dispute. Custody is being evaluated separately by another psychologist retained through the attorneys, and the parents jointly request a clinical neuropsychological and psychological evaluation for a different purpose. Eli has grown withdrawn and has developed routines he will not vary, and with his grades slipping both parents want to know whether he is autistic.

Before the first interview the evaluator reviews the custody orders, and both parents sign consent for the evaluation and for each outside contact. They are told that it offers no opinion on custody or on either of them as a parent, that each receives the same report, and that the report can still be subpoenaed in the litigation. Eli is told in plain language that his parents will read the findings and that a serious risk to his safety would be shared with them promptly. Other sensitive things he says would be handled with him, and the evaluator commits to telling him first, whenever possible, and to working out with him how they are put. With his assent, the evaluator contacts his grandmother, teachers, pediatrician’s office, and schools, telling each that what they say may appear in the report.

The referral question is whether Eli is autistic, and beneath it lies a harder question about how to weigh the accounts of the adults who know him. Each of them has seen something clinically real, and none of them has seen everything that matters.

Three histories of one boy

Each parent is interviewed alone and asked for dated examples of Eli’s early play, friendships, language, and routines. His mother sees this autumn as the surfacing of something she has watched for a decade. She describes a boy who hung back at birthday parties and preferred one friend at a time, and who could not bear shirt tags or the noise of a cafeteria.

His father describes a child who ran the lunch table in elementary school and was invited everywhere. He attributes the trouble to the divorce and to adolescence, and he adds that Eli’s mother has always been an anxious reader of their son. Neither account is treated in advance as deceptive or as the better of the two. Each parent is articulate and certain that the other is missing something important, and on that point, at least, both turn out to be right.

The third history belongs to Eli’s paternal grandmother, who looked after him on most weekdays from infancy through fourth grade while both parents worked long hours. She may have spent more waking hours with the young Eli than either parent, which makes her, on paper, the best developmental informant in the family. She remembers a bright, affectionate boy with the neighbors’ children at her kitchen table most afternoons. Asked about arguments between his parents, she says that every couple has words and that Eli was always well cared for.

Her interview has another feature, one that is harder to write about fairly than her warmth. She is unsure of several dates and of the order of some events, and she places one playground incident in kindergarten at first and in second grade an hour later. She has never had a cognitive evaluation, and neither her age nor one collateral interview licenses a conclusion about her memory. The interview shows that her dates need support from another source, which is a narrower finding than a verdict on her as an informant.

Uncertain dates do not erase years of afternoons that she saw at first hand, and her account is used for what she was positioned to see. Her favorable picture of the household could reflect what she saw, ordinary forgetting, protectiveness toward a family she loves, the fact that she went home before the evenings began, or several of these together.

What the instruments said

The battery, given over three sessions, is built around the questions the referral raises. Intellectual, academic, memory, and executive tests ask whether a learning or attention problem lies beneath the falling grades, and a performance validity test offers one check on those scores. Ratings from four adults and from Eli, a personality inventory, and two observational procedures address the rest, with the results that carry the argument in Tables 1 and 2.

The rating scales are where the disagreement within the family turns into numbers. On the School-Age Form of the SRS-2, which asks about social behavior over the preceding six months, his mother’s ratings yield a total T score of 78, in the severe range. His father’s total of 57 and his grandmother’s 55 are within normal limits. Because his homeroom teacher has known him for fewer than three of the six months the form covers, the evaluator elects to use her answers descriptively rather than interpret a standardized total. Read that way, they describe a boy who has gone quiet in class this term.

The adolescent forms of the BASC-3 repeat the pattern on Withdrawal, a scale that can rise in autism and also in anxious or unhappy children. His mother and his teacher place him in the clinically significant range, at 76 and 73, and his father and grandmother do not. On Depression three reports sit together in the at-risk range, Eli’s among them, and his own form carries the highest scores in the table, with Social Stress at 79 and Anxiety at 74.

The BRIEF-2, which asks about executive functioning in daily life, divides the same way. His mother and his teacher rate the Global Executive Composite at 67 and 64, against 52 from his father, and Eli’s own rating of 62 is mildly elevated. Every BASC-3 and BRIEF-2 protocol passes its validity scales, which the SRS-2 lacks, so both parents’ protocols pass every check the forms offer while describing very different boys.

Table 1. Ratings by informant, in T scores. Every value is invented, and none is patient data.

Rating scaleMotherFatherGrandmotherTeacherEli
SRS-2 Total785755Descriptive
BASC-3 Withdrawal76556073
BASC-3 Depression6751566568
BASC-3 Anxiety


7274
BASC-3 Social Stress



79
BASC-3 Sense of Inadequacy



67
BASC-3 Interpersonal Relations (adaptive)



35
BRIEF-2 Global Executive Composite6752
6462

Note. All values are T scores (mean 50, standard deviation 10). On the SRS-2 School-Age Form, totals of 59 and below are within normal limits, 60 to 65 mild, 66 to 75 moderate, and 76 and above severe (Constantino & Gruber, 2012). The homeroom teacher had known Eli for fewer than three of the form’s six reference months, and the evaluator chose to use her SRS-2 responses descriptively rather than interpret a standardized total. BASC-3 values come from the adolescent parent, teacher, and self-report forms, and the grandmother completed the parent forms (Reynolds & Kamphaus, 2015). On BASC-3 clinical scales, 60 to 69 is at-risk and 70 and above is clinically significant. Interpersonal Relations is an adaptive scale on which lower scores are poorer, with 31 to 40 at-risk. On the BRIEF-2, 60 to 64 is mildly elevated, 65 to 69 potentially clinically elevated, and 70 and above clinically elevated (Gioia et al., 2015). The BRIEF-2 also asks about six months, and the teacher’s BRIEF-2 is scored because its manual describes a month of daily contact as sufficient for a valid rating. The grandmother did not complete the BRIEF-2, and a blank cell elsewhere means that the scale is not on that form or was not part of this illustration. Every BASC-3 protocol had acceptable F, Consistency, and Response Pattern indexes, and the self-report also had acceptable L and V indexes. Every BRIEF-2 protocol had acceptable Negativity, Inconsistency, and Infrequency scales.

His mother’s 78 deserves to be taken seriously, although every elevated SRS-2 total carries a limitation that belongs to the instrument. In the Simons Simplex Collection, emotional and behavioral problems explained 20 to 26 percent of the variance in parent SRS raw scores, with a similar association in unaffected siblings and in children with autism (Hus et al., 2013). The authors concluded that unadjusted scores are better read as general impairment than as autism-specific severity. Their sample included no children referred for anxiety or depression, so applying the finding to an anxious, withdrawn adolescent is an inference, though a reasonable one.

The six-month window also complicates the family’s ratings, which span the months before the change and the months after it. The two lower totals do not cancel the mother’s, although they do bear on whether the behavior appears across settings, which the diagnosis expects.

The direct testing tells a different story from the ratings, and a narrower one. On the WISC-V, Eli’s five primary index scores run from 94 to 114, in the average to high average range. His word reading, reading comprehension, and mathematical problem solving on the WIAT-4 are average, as is his learning and recall of verbal and visual material on the ChAMP. On the D-KEFS he inhibits automatic responses and switches between rules at age level, which leaves the executive complaints on the BRIEF-2 without a counterpart in the testing room. That gap is common, because performance tests and ratings of executive function correlate weakly and appear to capture different aspects of functioning (Toplak et al., 2013). The difference is no evidence that either finding is wrong, since the D-KEFS measures efficiency on a structured task and the BRIEF-2 describes how he manages ordinary days.

His performance on the Memory Validity Profile, a performance validity test developed for ages 5 to 21 and conormed with the ChAMP, is within the valid range (Sherman & Brooks, 2015b). That result is one indicator from one point in three sessions, so it supports his scores without guaranteeing that every one of them reflects full effort.

He works steadily on tasks with a clear end and a single adult beside him. The most open-ended task is the WIAT-4 Essay Composition, a ten-minute expository essay, and his complete, on-topic essay earns an average standard score of 96 under the manual’s rules. The evaluator separately notes that the essay is dutiful and emotionally flat and that he asks twice whether it is enough, clinical observations that the scoring rules are not designed to capture. Intact scores of this kind show what he can do under structured conditions, and they cannot by themselves explain the falling grades, exclude an attention disorder, or identify what caused the decline.

On the PAI-A, an adolescent personality inventory normed on community adolescents aged 12 to 18, his validity scales are within acceptable limits and his profile shows considerable distress (Morey, 2007). Anxiety at 72 and Stress at 74 lie more than two standard deviations above the adolescent mean, and Depression at 68 and Nonsupport at 66 lie between one and two. Taken scale by scale, the profile describes an anxious, unhappy boy who reports heavy pressure in his present circumstances and feels poorly supported by the people around him. His Suicidal Ideation score is not elevated, although a nonelevated scale cannot rule out risk on its own. The evaluator asks him directly about suicide and self-harm as well, beginning in the first session, and he reports no thoughts of either. The inventory shows that something is weighing on him without being able to say what, a question Eli has so far answered for no one.

Direct observation comes next, with Module 3 of the ADOS-2, the module for verbally fluent children and younger adolescents. Eli’s overall total is 4, all from the Social Affect domain, against an autism spectrum cutoff of 7, which gives a nonspectrum classification and a comparison score of 2. The codes rate his behavior across the whole administration, so the four points cannot be traced to any single topic. Through most of the session he converses reciprocally, coordinates gesture with eye contact, shifts easily between activities, and shares his enjoyment of them with the evaluator.

Separately from the scoring, the evaluator’s notes describe him growing tense when the questions turn to friendships and to trouble with other kids, and his answers shorten. Those notes stand as observations in their own right, and they cannot by themselves distinguish a social difference from a guarded boy on a painful subject. A nonspectrum classification does not rule autism out, since the instrument’s developers note that clinical judgment can overrule it and call the severity score behind the comparison score one part of a broader picture (Gotham et al., 2009).

The evaluator also uses the MIGDAS-2, a qualitative interview built around conversation and the examinee’s own interests, which yields a descriptive profile with no scores, norms, or cutoffs. In that conversation Eli teaches the evaluator a paradiddle on the edge of the table and laughs, kindly, at the result. In the area of sensory use and interests, he describes hating sudden noise that he does not control, such as fire drills and a full cafeteria, while seeking out loud sound that he makes himself. His interest in drumming is intense, shared with his two oldest friends, and set aside without strain when the evaluator changes the subject.

The ADOS-2 and the MIGDAS-2 cannot be counted as two independent findings against autism, because the same evaluator conducted both in the same office and knew the result of the first. The MIGDAS-2 questionnaires and the SRS-2 likewise come from the same four adults, so a mother’s narrative that matches her ratings is one informant heard twice. The MIGDAS-2 therefore enters the file as further clinical evidence about Eli at ease, with no standing as confirmation of the ADOS-2, and neither procedure can say what he was like at four. Eli’s anxiety on the PAI-A and on the BASC-3 is the same boy answering twice, which strengthens the finding about his present state without adding a second source.

Table 2. Eli’s test performance and observed behavior. Every value is invented, and none is patient data.

MeasureResult
WISC-V primary indexesVerbal Comprehension 112, Visual Spatial 108, Fluid Reasoning 114, Working Memory 101, Processing Speed 94
WIAT-4, selected subtestsWord Reading 106, Reading Comprehension 103, Math Problem Solving 99, Essay Composition 96
ChAMPVerbal and visual memory, immediate and delayed, in the average range
D-KEFS, selected conditionsInhibition 11 and Inhibition/Switching 10 on Color-Word Interference, and Number-Letter Switching 10 on Trail Making
Memory Validity ProfileWithin the valid range
PAI-AValidity scales within acceptable limits. Anxiety 72, Stress 74, Depression 68, Nonsupport 66. Suicidal Ideation not elevated.
ADOS-2, Module 3Social Affect 4, Restricted and Repetitive Behavior 0, overall total 4 (autism spectrum cutoff 7), nonspectrum classification, comparison score 2
MIGDAS-2Qualitative profile of language and communication, social relationships and emotional responses, and sensory use and interests. No scores, norms, or cutoffs.

Note. WISC-V and WIAT-4 values are standard scores (mean 100, standard deviation 15), and D-KEFS values are scaled scores (mean 10, standard deviation 3). PAI-A values are T scores (mean 50, standard deviation 10) against the inventory’s norms for community adolescents, and each elevation is read in terms of what its scale measures. The cognitive and academic measures are the WISC-V (Wechsler, 2014), the WIAT-4 (NCS Pearson, 2020), the ChAMP (Sherman & Brooks, 2015a), and the D-KEFS (Delis et al., 2001). The others are the Memory Validity Profile (Sherman & Brooks, 2015b), the PAI-A (Morey, 2007), the ADOS-2 (Lord et al., 2012), and the MIGDAS-2 (Monteiro & Stegall, 2018).

Together, the measures document substantial reported emotional distress and broadly intact performance under structured conditions. They cannot establish when the difficulties began or what caused them, resolve the conflicting developmental histories, or reveal what is happening to Eli at school. At this point the file still supports two readings, a lifelong social difference now showing under stress or an ordinarily social boy who has recently stopped behaving like one. The scores cannot choose between them, because every instrument describes the present and the two readings differ about the past.

What he had not told anyone

The formulation changes in the third session, about three months into the school year, as the MIGDAS-2 conversation about drumming is winding down. Asked why the lessons have stopped being fun, Eli says that he cannot hold a beat in his head any more because of everything else that is in there. After a long silence he takes out his phone and slides it across the table.

The messages on the phone begin in the second week of the school year, after a classmate overheard his own parents discussing the dispute over the family business. Within days Eli had a nickname drawn from it, and a group chat followed, devoted to jokes about which parent would end up with the house. Since then, he says, five boys have imitated his voice whenever he answers in class, and his seat at the lunch table has been given away. He has never answered back, and since the fourth week of school he has eaten lunch alone in the band room.

The particulars matter diagnostically, because what separates bullying from ordinary conflict between peers is an imbalance of power (Juvonen & Graham, 2014). Here the imbalance is one of numbers and social standing, and by his account it has been exercised almost daily for about eleven weeks.

He says that he told no one because he was ashamed and because he expected adult intervention to make the boys worse. He gives a third reason that belongs to this family, and he states it as a prediction of what the adults would do. “If I tell one of them, it goes to a lawyer,” he says, “and then it turns into a thing about the other one.”

Before anything is decided about telling, the evaluator asks him again, in words suited to his age, about suicide and self-harm. He describes no wish to die and no thought of hurting himself, and he reports no threats of violence and no physical aggression from the boys. Nothing he describes suggests serious or immediate danger, so the bullying is treated as sensitive information to be handled with him under the confidentiality terms agreed at the outset. With those terms and his assent in mind, the evaluator works with him toward telling the adults who can support and protect him, at home and at school. The rest of the hour is spent deciding with him how each parent will be told, in whose words, and which adult at school he would want to hear it first.

Near the end of the session he adds something that nobody had asked about. “They say what I just said back to me in a stupid voice,” he says, “and that’s what it sounds like at my house when both my parents are in it.” He speaks warmly of each parent taken singly, and by his account he has listened for years to the two of them mimic and belittle each other.

The screenshots on his phone, dated from the second week of school, already document the nickname and the jokes in the group chat. Within the week, after both parents have been told and have authorized a follow-up with the school, the band director confirms that Eli has eaten lunch in the band room since the fourth week. The mimicry in class, the lost seat at lunch, and how the nickname began remain his account alone, since his homeroom teacher has noticed his silence without noticing its cause. His account is taken seriously and acted on before any confirmation arrives, because a disclosure of this kind cannot be held in suspense until a second source agrees with it.

His account of the voices at home enters the file as a statement of his experience. The emotional security hypothesis holds that a child’s past experience of conflict between parents shapes how the child regulates emotion and responds when they fight again (Davies & Cummings, 1994). A review links conflict between parents to children’s later emotional problems (Harold & Sellers, 2018), and a separate review links being bullied to the same outcomes (Arseneault, 2018). Those findings make his account plausible, but none of them tests the step from conflict at home to humiliation at school, for Eli or for children in general.

Reading the record

A disclosure that moving is also a hazard for the evaluator, who now has a story that seems to explain everything. The autism question is therefore put to the developmental record as though the phone had stayed in his pocket, using documents written at the time and collected from the pediatrician, the schools, and both parents.

Under DSM-5-TR the diagnosis requires persistent deficits, currently or by history, in all three areas of Criterion A, which are social-emotional reciprocity, nonverbal communication in social interaction, and the capacity to build and keep relationships. The features must also have been present in the early developmental period, although the manual allows that they may become fully apparent only when social demands outgrow a child’s capacities. It adds that they may be masked later by learned strategies, and both allowances describe what his mother has in mind (American Psychiatric Association, 2022).

The pediatric chart shows developmental surveillance at each well-child visit and a toddler autism screen in the low-risk range, which counts for little on its own. A preschool conference note describes a four-year-old who recruits other children into pretend games and assigns the parts, and who covers his ears during fire drills. His kindergarten report card calls him a leader at the block table who notices when a classmate has been left out. Team rosters show him playing alongside the same two boys in most years since first grade, and both parents and the grandmother, asked separately, name those boys as his closest friends.

Read against Criterion A, the preschool and kindergarten documents show reciprocity at four and five, in the recruiting of playmates and in noticing a classmate left out. The rosters and the adults’ separate accounts point to two friendships kept from first grade until this autumn. No document or informant, his mother included, describes unusual eye contact, gesture, or expression at any age, and during the ADOS-2 they were well integrated. His current withdrawal is real, but it began this autumn and is confined to the places and peers he associates with the five boys. The available record shows no persistent deficit in any of the three areas, although documents written for other purposes can miss subtle difficulties that a bright child learns to cover. Nothing in the records contradicts his mother’s description of a cautious boy who prefers a few friends to many, which is a temperament common among children who are not autistic.

Criterion B requires at least two of four kinds of restricted or repetitive behavior, and his dislike of sudden noise, documented at four and described again in the MIGDAS-2 conversation, could count as one. His drumming is an intense interest of an ordinary kind, and no source describes repetitive movements, rigid routines, or distress at small changes before this autumn. The new rigidity that prompted the autism question has a pattern, because by Eli’s account each routine keeps him away from the five boys. He refuses the bus and arrives one minute before the bell for the same reason that he eats lunch in the band room. A routine that exists to avoid a threat serves a different function from insistence on sameness, even when a parent watching cannot tell the two apart.

The evidence therefore does not support autism spectrum disorder, which is a weaker statement than proof of absence and the strongest one the method allows. It would have been different if the early records had described a child who talked at length on one subject and had to be steered into groups by adults. It would be reopened if difficulties of the Criterion A kind persisted after the bullying had stopped. Either way the bullying would still be real, and the formulation would need both findings, because adversity does not exclude autism.

The evidence supports adjustment disorder with mixed anxiety and depressed mood, which the manual codes as F43.23. The earliest documented changes date from about three weeks after the nickname appeared, when his grades began to slip and his drum teacher noticed the silence between songs. His grades and his playing had held steady through the first year of the divorce, by the school’s reports and the drum teacher’s account, which is why the bullying is identified as the stressor. The drum teacher’s account matters most here, because his lessons give the only steady record of Eli on either side of the change, though he knows nothing of home and cannot name a cause. The symptoms are impairing and do not meet the criteria for another mental disorder, and nothing in the history suggests an earlier disorder they could be exacerbating. The impairment shows in slipping grades and a retreat from friends of seven years, and the diagnosis makes no claim that his distress is out of proportion to what was done to him. It will be revisited if his symptoms deepen, or if they persist more than six months after the bullying ends.

The criteria for major depressive disorder are not met at present, because he has fewer than five of the nine symptoms and his sleep, appetite, and energy are largely intact. His low mood and poor concentration also lift when school is out, on weekends at his father’s house and equally on the school holidays that fall on his mother’s days. Social anxiety disorder requires fear out of proportion to the actual threat, typically lasting six months or more, and his fear is recent and attached to boys who are in fact humiliating him.

Attention-deficit/hyperactivity disorder requires several symptoms before age twelve and requires that they not be better explained by another disorder, and no record or informant describes inattention before this autumn. His trouble concentrating arrived with the anxiety and low mood and lifts with them on free days, which fits a reaction to his circumstances better than a lifelong attention disorder. Intact test scores would not exclude the disorder on their own, since structured tests given by one adult in a quiet room can miss difficulties that appear in a classroom. Posttraumatic stress disorder requires exposure to actual or threatened death, serious injury, or sexual violence, none of which careful inquiry finds in his history. Chronic humiliation can do serious harm without meeting that definition, and calling it trauma in the diagnostic sense would misdescribe both the boy and the category.

Adjustment disorder carries no implication of mildness, because the category is defined by its link to a stressor and sets no upper limit on distress. The report also names the years of conflict between his parents, as Eli describes living with them, as a possible source of added vulnerability.

Five questions inside one

No adult in this family is the best source on every question, and the formulation comes from matching each question to the evidence able to answer it. That is worth examining, because the adults’ accounts raise five separate questions that clinicians routinely fold into one.

  1. Did the informant understand the questions and attend to them?
  2. Were the informant’s answers internally consistent across the questionnaire or interview?
  3. Did the informant accurately observe the behavior that the answers describe?
  4. Is the informant’s retrospective account of past events historically accurate?
  5. How much should the clinician rely on that account in resolving this differential diagnosis?

The validity indexes on a rating scale bear only on the first two questions, and the third concerns vantage point, which in this family is set by the calendar. Under the temporary schedule Eli spends school nights at his mother’s house and most weekends at his father’s, where his grandmother also sees him. The two adults who rate him as withdrawn see him at school or within hours of it, and the two whose ratings are mostly unremarkable see him on Saturdays.

The evaluator’s working hypothesis is that the ratings divide by school day more than by household. It has support, since his mood also lifts on school holidays spent in either home, and it remains a hypothesis because no one has rated him on matched days in both settings. In the terms of the Operations Triad Model, the disagreement is a candidate diverging operation, one that reflects real variation across settings (De Los Reyes et al., 2013). Rater and measurement explanations stay open, and none of them requires a theory about anyone’s motives.

The fourth question separates two kinds of memory in the grandmother’s account, her repeated first-hand experience and her placing of events in time. The school documents corroborate her general picture of a sociable young boy, and the rosters show years of shared activity with the same two boys, though not what those friendships were like. Her particular memories of the neighbors’ children at her table have no independent confirmation, and they are weighed as first-hand observations that fit the documented pattern. Her picture of the marriage carries little weight because she was seldom present for it.

The fifth question is the clinician’s, and it has to be answered one claim at a time. The mother’s account weighs heavily on what Eli is like on a school night and less heavily on his social development at five. There a kindergarten report written at the time carries more weight than a memory eight years old, although a teacher’s note is one adult’s view and can be mistaken as well. The father’s account weighs heavily on Eli’s early friendships and lightly on this autumn, since he sees his son on weekends and knows the elementary lunch table at second hand. The drum teacher is the strongest witness to change and the weakest to cause, and the grandmother has the widest view of the early years and the least certain calendar. A single verdict on each person’s overall credibility would be easier to write, and it would discard accurate observations from whichever adult it ranked lowest.

The same questions apply to Eli, whose self-reports have acceptable validity indexes and whose account of the bullying is partly corroborated, although neither fact shows that the rest of his report is accurate. Four kinds of evidence in a file like this one are easily mistaken for one another. A performance validity test supports the interpretation of Eli’s test scores, and response-quality indexes describe selected features of how a questionnaire was completed. Agreement between informants describes how their reports compare, and historical accuracy concerns whether a particular claim matches what happened. The first two come with published cutoffs, and a professional consensus statement recommends performance validity tests routinely for school-aged children (Sweet et al., 2021). No check of comparable standing has been established for the fourth, whether the claim is a mother’s memory of preschool or a grandmother’s memory of a playground. Adding instruments strengthens the evidence about Eli’s present state, and none of them settles whether a particular historical claim is true.

What the informant literature already explains

The researchers who study informants have known most of this for decades, and it should not be presented as a discovery. Achenbach et al. (1987) pooled 269 samples from 119 studies and found a mean correlation of .60 between informants of the same type, such as two parents or two teachers. Informants of different types, such as a parent and a teacher, agreed at .28, a gap the authors read as evidence that children’s behavior varies by situation. Children’s own reports agreed with those of other informants at .22, and a meta-analysis of 341 later studies put overall agreement between informants at .28 (De Los Reyes et al., 2015).

The field has also built frameworks for interpreting disagreement, beginning with the argument that informants should be chosen for the contexts and perspectives they represent (Kraemer et al., 2003). The Attribution Bias Context Model, proposed by De Los Reyes and Kazdin (2005), traces discrepancies to informants’ differing explanations of a child’s behavior and to the settings in which they see it. Later work tied disagreement between parents and teachers to how preschoolers actually behaved with a parent and with a clinical examiner (De Los Reyes et al., 2009).

In the Dunedin cohort, agreement between what eighteen-year-olds recalled and what had been documented as they grew up was poorest for family processes and psychological states (Henry et al., 1994). The authors advised against using retrospective reports where precise dates and frequencies are needed. Parents of children referred for possible autism or developmental delay, interviewed repeatedly as the children grew from two to nine, reported progressively later ages for early language milestones (Hus et al., 2011).

Yet a review of adults’ recollections of their own childhoods concluded that claims of general unreliability are overstated (Brewin et al., 1993). Belli (1998) argued that recall improves when questions follow the organization of autobiographical memory into life periods and connected events, as an event history calendar does. The grandmother is therefore asked which house she lived in at the time of the playground incident, and with that anchor she gives second grade twice. A steadier answer is not thereby a correct one, and the date goes into the file as her best estimate.

The literature therefore explains why informants disagree, and it offers sound guidance on choosing and questioning them. The best-known method for appraising individual statements, criteria-based content analysis, was developed to evaluate children’s testimony in sexual offense cases, which is a different task from judging whether an honest adult remembers a childhood accurately (Vrij, 2005).

The index I did not build

I have spent more time than I would like to admit trying to build a tool for that task. The idea was a brief response-consistency check for collateral informants that borrowed the logic of validity testing, one of the most successful quality-control traditions in our field. I had ruled out from the start any suggestion that an inconsistent informant is a deceptive one. The version that survived longest used five original questions about the examinee’s history, asked before an intake questionnaire and again after it. Four matching answers out of five would yield 80 percent agreement, a figure that is easy to compute and easier still to put in a report.

The trouble began with the question of what the figure would mean. Validity belongs to the interpretation of a score for a particular use, and each interpretation needs its own evidence (American Educational Research Association et al., 2014). As a description of five answers the percentage is beyond reproach, and as an estimate of how stably the informant generally responds it is too coarse to use. Each changed answer moves it by twenty percentage points, so the scale has only six possible values. If the five answers were independent and each had the same nine-in-ten chance of being repeated, a careful informant would still change at least one about four times in ten. Neither assumption is likely to hold for five different questions about a childhood, so the figure illustrates the arithmetic and nothing more.

Answers change for reasons that can belong to the question as much as to the person answering it. Bond (1987) showed that inconsistency on repeated personality items depends on properties of the item itself, including its content and its ambiguity. Fekken and Jackson (1988) found that how long a person took to answer an item predicted a later change of answer better than the item’s own properties did. Those studies involved people describing themselves, so their bearing on an informant recalling a child’s history is by analogy. A grandmother who revises an answer after forty minutes spent thinking about her grandson’s childhood may be the most conscientious person in the building.

The short interval creates the opposite problem, because a respondent who simply remembers the first answer will repeat it whether or not it was true. Identical answers therefore do not establish accuracy, and changed answers do not establish carelessness. Five questions that sit outside the standardized questionnaire also describe only themselves, and they say nothing about how consistently the informant completed the instrument in between.

The survey literature on careless responding looked at first like a foundation. Meade and Craig (2012) classified roughly 10 to 12 percent of undergraduates completing a lengthy survey for course credit as careless responders. Wood et al. (2017) found that consistency in online samples drops sharply when people answer faster than about one second per item. A later review, citing that work, noted that about thirty repeated items are recommended for a reliable consistency index of that kind (Zickar & Keith, 2023). Indexes like these are built to catch people who are not reading the items, which is a different problem from an attentive grandmother’s shifting memory of a date. The closest clinical study I found, of 65 psychiatric patients, found only weak evidence that person-reliability indexes moderated the validity of clinical criteria (Fekken et al., 1999).

I then held the idea against the standards of the validity literature it was imitating. None of its researchers has commented on a collateral check of this kind, and what follows is my own application of their published principles. That literature assesses test performance and reported symptoms with separate measures, each sampled with several indicators across an examination (Sweet et al., 2021), because response bias can vary within a single evaluation (Boone, 2009). Cutoffs are expected to be developed against criterion groups and checked in clinical samples, and they are set to keep false positives rare among credible examinees, by convention at about one in ten or fewer.

Even a failed validity test is no finding of malingering under those standards. The most widely used criteria, proposed by Slick et al. (1999) and revised by Sherman et al. (2020), require an external incentive together with positive evidence of fabrication or exaggeration that other conditions cannot explain. The revision counts as one kind of evidence a marked discrepancy between an examinee’s account and the report of a reliable collateral informant with minimal stakes in the outcome. That safeguard asks the clinician to weigh an informant’s stake without treating it as evidence of distortion, and the criteria offer no validated means of judging whether a collateral history is accurate. Involvement in a contested divorce does not by itself make either of Eli’s parents an unreliable informant, and the stakes in this family are a reason to seek independent records and nothing more.

Measured against those standards, my index had no criterion group, since it is hard to assemble a sample of informants who are known to be inaccurate. It had no means of estimating a false-positive rate and no reference data for grandmothers as distinct from fathers. Nor was there evidence of ecological validity, meaning any link between five answers in an office and the accuracy of an account of a childhood.

I tried narrowing the claim, so that one or two changed answers would serve only as a prompt for further questions. That is a sensible habit with no validated cutoff behind it, and it is what a good interviewer does without a form. I then tried rating individual historical claims against whatever independent evidence existed, which aimed at the right target and became a worksheet that no number could summarize. Each narrowing made the procedure easier to defend, and each left it with less to tell me that a careful interview would not already have shown.

Eli’s evaluation supplies a last and more practical argument, because it is clinical but takes place beside a lawsuit. In that setting a line stating that one parent’s response consistency was 60 percent would be read as a verdict on that parent’s overall credibility, drawn from five answers, whatever caveats surrounded it. The evaluator would be asked, fairly, for an error rate and a validation sample that do not exist. The careful interview has no published error rate either, which is why the third study proposed below tests it. The problem is real, and although the goal was a reasonable one, no instrument came out of that work.

Where the gap is

What remains insufficiently established is how the accuracy of an individual informant’s historical account can be quantified without confusing consistency with truthfulness, or turning observations that depend on context into a global credibility score. Part of the gap is psychometric, and larger samples and better-designed items could address some of its measurement limitations without establishing whether consistency predicts historical accuracy. Another part is methodological, because accuracy cannot be validated without a criterion and records verify only the public surface of a childhood, never the tone of voice at a dinner table. The deepest part is evidentiary, because accuracy is a property of a particular claim and consistency a property of a particular set of answers. A judgment about an informant’s credibility in general is a third matter, and the accuracy or consistency of a few statements cannot settle it for the rest of what that informant reports. A method faithful to those distinctions would work at the level of the claim and would state its conclusions in words.

Studies worth running

A research program that respected those limits would be modest, and its first two studies would be feasible in ordinary clinics. The first would be descriptive, because I did not identify published base-rate data on how often conscientious collateral informants change their answers to clearly worded historical questions within one session. It would vary the ambiguity of the questions and the interval between askings, and it would measure each informant’s memory, emotional involvement, and first-hand access to the events. Its product would be the base rates without which no individual change can be interpreted.

The second study would ask whether changed answers have anything to do with wrong ones, using pediatric and school records to check events that can be dated, such as moves and hospitalizations. When the Dunedin investigators compared documented childhood adversity with adults’ later recollection of it, agreement was only moderate, with a correlation of .47 and a weighted kappa of .31 (Reuben et al., 2016). If inconsistency within a session fails to predict disagreement with the record in more than one sample, the case for a consistency index weakens considerably. If it does predict disagreement, an index would still have to show that it adds something to a careful interview.

The third study would test the interview itself and leave the informant unscored, comparing a structured collateral interview that documents how the informant knows each thing with a standard intake in a randomized design. Its outcomes would be agreement among clinicians blind to condition and agreement with a longitudinal expert diagnosis (Spitzer, 1983) built from later data by experts who never see the interview under test, which guards against incorporation bias. Reporting standards for such best-estimate designs now exist, and they warn that the method does not guarantee validity, which depends on how carefully it is carried out (Eijsbroek et al., 2025). Even a well-built expert diagnosis is a fallible reference standard, and agreement with it would show only whether the interview improves diagnosis. That is a different property from historical accuracy, which only a check against independent records, as in the second study, can test. A further line of work would ask when a disagreement is itself the finding, building on the observational design used with preschoolers.

Each study should state its hypotheses and analysis plan in advance and be replicated at an independent site before anyone relies on it. Any brief method should also be tested for harm, meaning whether it leads clinicians to label informants unreliable more often than the evidence warrants. I would not recommend a screening score for informants, and nothing here should be read as a step toward one.

What each of them had right

Eli hears the conclusions first, and each parent then hears them separately and in the same words. Eli’s mother was right that something was seriously wrong, and she hears that the early records do not show the lifelong pattern she inferred from his caution. His father’s sociable boy is there in the early school records, and he hears that the weekends he sees are the part of the week in which his son’s distress is least visible. His grandmother’s afternoons at the kitchen table stand in the report as her own first-hand account, consistent with the school records, with the dates left approximate. None of them had known about the lunch table, and the person who did was thirteen and had concluded from experience that handing information to adults was dangerous.

The recommendations go to both parents in identical form, beginning with the school. The school is asked for a plan with a named adult Eli can go to, supervision where the boys gather, and action on the group chat under its bullying policy. The burden of the plan should fall on what the five boys do, and Eli should not have to rearrange his day in order to be safe. Psychotherapy is recommended for the anxiety, low mood, and avoidance that have narrowed his days, including a graded return to the places he avoids once adults have made them safe. His therapist is asked to screen for suicidal thinking at intervals while his mood stays low, and his parents are told which changes should prompt a call. Both parents are asked to keep the litigation out of his hearing and out of his hands, with no messages carried between houses and no use of the bullying in their dispute.

Until better methods exist, the work available to clinicians is unglamorous, and it consists largely of asking each person for the source of each claim and reading what was written at the time. Clinical integration of this kind is not unscientific for lacking a validated score, although clinical experience does not give it an independently established error rate. It also requires staying curious about the observation that does not fit the leading explanation.

The school has since intervened with the five boys and with the group chat, and Eli has begun seeing a therapist whom both parents accepted. His drum teacher reports that the jokes between songs have not returned, although Eli stayed late last Thursday to work out a fill he had heard and wanted to get right. The statement in this file that has held up best is still the first, from a drum teacher who said only that something was different and that he did not know what.

Eli is fictional. The details belong to no patient, assembled because this pattern arrives often, wearing different clothes.

References

Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and emotional problems: Implications of cross-informant correlations for situational specificity. Psychological Bulletin, 101(2), 213–232. https://doi.org/10.1037/0033-2909.101.2.213

American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for educational and psychological testing. American Educational Research Association.

American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.). https://doi.org/10.1176/appi.books.9780890425787

Arseneault, L. (2018). Annual research review: The persistent and pervasive impact of being bullied in childhood and adolescence: Implications for policy and practice. Journal of Child Psychology and Psychiatry, 59(4), 405–421. https://doi.org/10.1111/jcpp.12841

Belli, R. F. (1998). The structure of autobiographical memory and the event history calendar: Potential improvements in the quality of retrospective reports in surveys. Memory, 6(4), 383–406. https://doi.org/10.1080/741942610

Bond, J. A. (1987). The process of responding to personality items: Inconsistent responses to repeated presentation of identical items. Personality and Individual Differences, 8(3), 409–417. https://doi.org/10.1016/0191-8869(87)90042-0

Boone, K. B. (2009). The need for continuous and comprehensive sampling of effort/response bias during neuropsychological examinations. The Clinical Neuropsychologist, 23(4), 729–741. https://doi.org/10.1080/13854040802427803

Brewin, C. R., Andrews, B., & Gotlib, I. H. (1993). Psychopathology and early experience: A reappraisal of retrospective reports. Psychological Bulletin, 113(1), 82–98. https://doi.org/10.1037/0033-2909.113.1.82

Constantino, J. N., & Gruber, C. P. (2012). Social Responsiveness Scale (2nd ed.). Western Psychological Services.

Davies, P. T., & Cummings, E. M. (1994). Marital conflict and child adjustment: An emotional security hypothesis. Psychological Bulletin, 116(3), 387–411. https://doi.org/10.1037/0033-2909.116.3.387

Delis, D. C., Kaplan, E., & Kramer, J. H. (2001). Delis-Kaplan Executive Function System. The Psychological Corporation.

De Los Reyes, A., Augenstein, T. M., Wang, M., Thomas, S. A., Drabick, D. A. G., Burgers, D. E., & Rabinowitz, J. (2015). The validity of the multi-informant approach to assessing child and adolescent mental health. Psychological Bulletin, 141(4), 858–900. https://doi.org/10.1037/a0038498

De Los Reyes, A., Henry, D. B., Tolan, P. H., & Wakschlag, L. S. (2009). Linking informant discrepancies to observed variations in young children’s disruptive behavior. Journal of Abnormal Child Psychology, 37(5), 637–652. https://doi.org/10.1007/s10802-009-9307-3

De Los Reyes, A., & Kazdin, A. E. (2005). Informant discrepancies in the assessment of childhood psychopathology: A critical review, theoretical framework, and recommendations for further study. Psychological Bulletin, 131(4), 483–509. https://doi.org/10.1037/0033-2909.131.4.483

De Los Reyes, A., Thomas, S. A., Goodman, K. L., & Kundey, S. M. A. (2013). Principles underlying the use of multiple informants’ reports. Annual Review of Clinical Psychology, 9, 123–149. https://doi.org/10.1146/annurev-clinpsy-050212-185617

Eijsbroek, V. C., Kjell, K., Schwartz, H. A., Boehnke, J. R., Fried, E. I., Klein, D. N., Gustafsson, P., Augenstein, I., Bossuyt, P. M. M., & Kjell, O. N. E. (2025). The LEADING guideline: Reporting standards for expert panel, best-estimate diagnosis, and longitudinal expert all data (LEAD) methods. Comprehensive Psychiatry, 141, Article 152603. https://doi.org/10.1016/j.comppsych.2025.152603

Fekken, G. C., Holden, R. R., & Cotton, D. H. G. (1999). Person reliability of psychiatric patients’ responses to a psychopathology inventory. Journal of Clinical Psychology, 55(10), 1299–1306. https://doi.org/10.1002/(SICI)1097-4679(199910)55:10<1299::AID-JCLP10>3.0.CO;2-Z

Fekken, G. C., & Jackson, D. N. (1988). Predicting consistent psychological test item responses: A comparison of models. Personality and Individual Differences, 9(5), 873–882. https://doi.org/10.1016/0191-8869(88)90005-0

Gioia, G. A., Isquith, P. K., Guy, S. C., & Kenworthy, L. (2015). Behavior Rating Inventory of Executive Function (2nd ed.). Psychological Assessment Resources.

Gotham, K., Pickles, A., & Lord, C. (2009). Standardizing ADOS scores for a measure of severity in autism spectrum disorders. Journal of Autism and Developmental Disorders, 39(5), 693–705. https://doi.org/10.1007/s10803-008-0674-3

Harold, G. T., & Sellers, R. (2018). Annual research review: Interparental conflict and youth psychopathology: An evidence review and practice focused update. Journal of Child Psychology and Psychiatry, 59(4), 374–402. https://doi.org/10.1111/jcpp.12893

Henry, B., Moffitt, T. E., Caspi, A., Langley, J., & Silva, P. A. (1994). On the “remembrance of things past”: A longitudinal evaluation of the retrospective method. Psychological Assessment, 6(2), 92–101. https://doi.org/10.1037/1040-3590.6.2.92

Hus, V., Bishop, S., Gotham, K., Huerta, M., & Lord, C. (2013). Factors influencing scores on the Social Responsiveness Scale. Journal of Child Psychology and Psychiatry, 54(2), 216–224. https://doi.org/10.1111/j.1469-7610.2012.02589.x

Hus, V., Taylor, A., & Lord, C. (2011). Telescoping of caregiver report on the Autism Diagnostic Interview-Revised. Journal of Child Psychology and Psychiatry, 52(7), 753–760. https://doi.org/10.1111/j.1469-7610.2011.02398.x

Juvonen, J., & Graham, S. (2014). Bullying in schools: The power of bullies and the plight of victims. Annual Review of Psychology, 65, 159–185. https://doi.org/10.1146/annurev-psych-010213-115030

Kraemer, H. C., Measelle, J. R., Ablow, J. C., Essex, M. J., Boyce, W. T., & Kupfer, D. J. (2003). A new approach to integrating data from multiple informants in psychiatric assessment and research: Mixing and matching contexts and perspectives. American Journal of Psychiatry, 160(9), 1566–1577. https://doi.org/10.1176/appi.ajp.160.9.1566

Lord, C., Rutter, M., DiLavore, P. C., Risi, S., Gotham, K., & Bishop, S. L. (2012). Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) manual (Part I): Modules 1–4. Western Psychological Services.

Meade, A. W., & Craig, S. B. (2012). Identifying careless responses in survey data. Psychological Methods, 17(3), 437–455. https://doi.org/10.1037/a0028085

Monteiro, M. J., & Stegall, S. (2018). Monteiro Interview Guidelines for Diagnosing the Autism Spectrum (2nd ed.). Western Psychological Services.

Morey, L. C. (2007). Personality Assessment Inventory-Adolescent. Psychological Assessment Resources.

NCS Pearson. (2020). Wechsler Individual Achievement Test (4th ed.).

Reuben, A., Moffitt, T. E., Caspi, A., Belsky, D. W., Harrington, H., Schroeder, F., Hogan, S., Ramrakha, S., Poulton, R., & Danese, A. (2016). Lest we forget: Comparing retrospective and prospective assessments of adverse childhood experiences in the prediction of adult health. Journal of Child Psychology and Psychiatry, 57(10), 1103–1112. https://doi.org/10.1111/jcpp.12621

Reynolds, C. R., & Kamphaus, R. W. (2015). Behavior Assessment System for Children (3rd ed.). Pearson.

Sherman, E. M. S., & Brooks, B. L. (2015a). Child and Adolescent Memory Profile. Psychological Assessment Resources.

Sherman, E. M. S., & Brooks, B. L. (2015b). Memory Validity Profile. Psychological Assessment Resources.

Sherman, E. M. S., Slick, D. J., & Iverson, G. L. (2020). Multidimensional malingering criteria for neuropsychological assessment: A 20-year update of the malingered neuropsychological dysfunction criteria. Archives of Clinical Neuropsychology, 35(6), 735–764. https://doi.org/10.1093/arclin/acaa019

Slick, D. J., Sherman, E. M. S., & Iverson, G. L. (1999). Diagnostic criteria for malingered neurocognitive dysfunction: Proposed standards for clinical practice and research. The Clinical Neuropsychologist, 13(4), 545–561. https://doi.org/10.1076/1385-4046(199911)13:04;1-Y;FT545

Spitzer, R. L. (1983). Psychiatric diagnosis: Are clinicians still necessary? Comprehensive Psychiatry, 24(5), 399–411. https://doi.org/10.1016/0010-440X(83)90032-9

Sweet, J. J., Heilbronner, R. L., Morgan, J. E., Larrabee, G. J., Rohling, M. L., Boone, K. B., Kirkwood, M. W., Schroeder, R. W., Suhr, J. A., & Conference Participants. (2021). American Academy of Clinical Neuropsychology (AACN) 2021 consensus statement on validity assessment: Update of the 2009 AACN consensus conference statement on neuropsychological assessment of effort, response bias, and malingering. The Clinical Neuropsychologist, 35(6), 1053–1106. https://doi.org/10.1080/13854046.2021.1896036

Toplak, M. E., West, R. F., & Stanovich, K. E. (2013). Practitioner review: Do performance-based measures and ratings of executive function assess the same construct? Journal of Child Psychology and Psychiatry, 54(2), 131–143. https://doi.org/10.1111/jcpp.12001

Vrij, A. (2005). Criteria-based content analysis: A qualitative review of the first 37 studies. Psychology, Public Policy, and Law, 11(1), 3–41. https://doi.org/10.1037/1076-8971.11.1.3

Wechsler, D. (2014). Wechsler Intelligence Scale for Children (5th ed.). Pearson.

Wood, D., Harms, P. D., Lowman, G. H., & DeSimone, J. A. (2017). Response speed and response consistency as mutually validating indicators of data quality in online samples. Social Psychological and Personality Science, 8(4), 454–464. https://doi.org/10.1177/1948550617703168

Zickar, M. J., & Keith, M. G. (2023). Innovations in sampling: Improving the appropriateness and quality of samples in organizational research. Annual Review of Organizational Psychology and Organizational Behavior, 10, 315–337. https://doi.org/10.1146/annurev-orgpsych-120920-052946

You Might Also Like