Psychopathy and Reactive Attachment Disorder

This is another SPSS post.

The idea for this survey was personal. I had an uncle who we retrospectively suspect was a psychopath, and two of his children, severely abused and messed up emotionally, ended up living with us. Both showed features of Reactive Attachment Disorder (RAD), and one also showed psychopathic traits. I'm trying to figure out what made him tick—did he act psychopathic because of his RAD, or did he inherit psychopathy as a separate trait? Is there any relationship between RAD and psychopathy?
To answer this, I wrote a survey assessing both RAD and psychopathic traits, and distributed it to a number of different attachment, ODD and CD-related forums and listservs. I wanted to get a sample of RAD kids, and compare them to kids with similar behavior problems who'd come from a much better environment.

I used three scales - the Relationship Problems Questionnaire (an assessment scale for RAD), the Antisocial Process Screening Device (an assessment scale for psychopathy) and the Strengths and Difficulties Questionnaire (Dadds et al designed a pooled scale of APSD and SPQ to assess psychopathy in children). I also included a number of questions assessing the child's history of trauma and attachment disruption, so I could directly assess the impact of life events on the children's behavior.

Sample Description

The sample consisted of 68 children, ranging in age from 3 to 23, with a mean of 10 years and a standard deviation of 4.8 years. There were 34 boys and 33 girls, and one child whose gender was not reported due to a technical error. The majority (55 kids) lived in the USA, with 5 from England, 3 from Canada, and 1 each from Australia, 'Europe', India and the Netherlands. The survey was filled out by 28 biological parents, 26 adoptive parents, 6 stepparents, 5 foster parents and 3 extended family members ([insert what kinds]).

As well as RAD (34 kids), ODD (24 kids) and CD (one child), the diagnoses mentioned for children (both official and possible) included ADHD (20 kids), 'anxiety' (8 kids), PTSD (6 kids), autism spectrum diagnoses (4 kids), depression (4 kids), OCD (3 kids), bipolar disorder (3 kids), panic disorder/panic attacks (3 kids), generalized anxiety disorder (2 kids) and FASD (2 kids). In addition, one child each had schizoaffective, borderline MR, 'disruptive behavior disorder', nonverbal learning disorder, agoraphobia, EDNOS, 'unspecified mental condition', learning disabilities, narcissistic personality disorder, intermittent explosive disorder, 'behavior problems', developmental dyspraxia, adjustment disorder, 'mood disorder with psychosis', CAPD, dysthymia, 'audio processing delay', sensory integration disorder and 'possible mood disorder'. Most kids had multiple diagnoses, with the average being 2 diagnoses per kid (standard deviation 1.6). Seven children had no diagnosis, and the child with the most diagnoses had 9.

To simplify the data, I grouped the diagnoses into internalizing, externalizing, trauma-related and neurological (a diagnosis with definite or very likely neurobiological basis, or reported possible neurological issues in the 'other trauma' text box), with internalizing and externalizing grouped together as 'neither', 'internalizing only', 'externalizing only' or 'both'. The resulting categories were:

• internalizing/externalizing: 23 neither, 7 internalizing only, 30 externalizing only and 10 both

• trauma-related: 35 yes (of whom 34 had a RAD diagnosis) and 33 no

• neurological: 15 yes (# only because they mentioned neurological issues under 'other trauma') and 53 no

The average scores on the behavior problems scales were (scales drawn from the studies linked above, although RPQ temperament excluded the single SPQ item loading on that scale):

• RPQ total score – 27.18+/-9.82

• RPQ temperament – 1.32+/-.72

• RPQ disinhibited – 1.35+/-1.02

• RPQ inhibited – 1.38+/-.61

• SDQ Conduct Problems – 7.12+/-1.97

• SDQ Emotional Problems – 5.15+/-2.68 (note that this scale remained unchanged when Dadds et al pooled the SDQ and APSD items together)

• SDQ Prosocial – 4.68+/-2.34

• SDQ Hyperactivity – 6.90+/-2.83

• SDQ Peer Problems – 5.16+/-2.40

• APSD total – 28.96+/-6.37

• APSD Narcissism – 10.20+/-2.88

• APSD Callous-Unemotional – 7.95+/-2.51

• APSD Impulsivity – 7.62+/-1.67

• SDQ/APSD Antisocial – 15.32+/-4.55

• SDQ/APSD Callous-Unemotional – 9.35+/-3.48

• SDQ/APSD Hyperactivity – 13.05+/-4.12

• SDQ/APSD Peer Problems – 6.58+/-2.82

Age and Gender Effects

Due to the wide age range, I decided to first compare the kids by age. I divided them into three roughly equal age groups - 3-7 years (26 kids), 7-13 years (20 kids) and 13-23 years (22 kids). My first analysis was to see if the items differed by age, by running an ANOVA.

The RPQ total did not differ significantly by age (p = .078), however, two of the subscales reached a significance level of .05. Scores on the 'temperament' subscale (measuring general fearful/inhibited behaviour, such as clinginess) were significantly higher in 3-7 year olds than in the older children (p = .011), a result that fits with the developmental characteristics typical of preschoolers. In addition, scores on the RPQ 'disinhibited' scale (measuring overly friendly behavior to strangers) were significant at a p-value of .044. Although the Bonferroni and Scheffe tests failed to reach significance, the trend suggested that the 13-23 year olds were higher on disinhibited social behavior than the younger children. It's possible that in this age group, the disinhibited items could have been measuring promiscuity, which is correlated with delinquency (see this study and this study).

Within the SDQ subscales, Conduct Problems and Hyperactivity both were associated with age. Conduct Problems were significantly (p = .004) lower in 3-7 year olds than in older children, and Hyperactivity was significantly higher (p = .004) in the youngest age group. Both are consistent with the general age patterns that have been found, although it's surprising not to find any difference between 7-13 and 13-23 year olds in Conduct Problems. One possibility is that the SDQ items capture behavior that is likely to start earlier, such as fighting with other children, tantrums and disobediance (items which more closely resemble Oppositional Defiant Disorder than Conduct Disorder criteria). Another possibility is that my sample selected for earlier onset of conduct problems.

The APSD, in contrast, showed no association with age, nor did either CU scale. This is also consistent with research suggesting that psychopathic traits tend to remain stable with age.

I then ran a T-test to assess the association of gender and the various scales. The RPQ total, inhibited and disinhibited scales all differed by gender (p = .003, .038 and .006 respectively). Specifically, girls scored significantly higher on all three subscales. In addition, girls scored higher on SDQ Peer Problems. This most likely reflects the gender differences in frequency of RAD (which is either unaffected by gender, or more common in girls due to higher female victimization rates) and ODD/CD (which are more common in boys). The 34 boys included 12 with a trauma-related diagnosis, while 22 of the 33 girls had trauma-related diagnoses.

When I separated the sample based on RAD diagnosis (excluding the one child with a different trauma-related diagnosis) and re-ran the T-test, the association between gender and RPQ inhibited score disappeared. However, the RPQ total and disinihibited scale was still higher in girls for the RAD kids (p = .029 and .018 respectively) but not non-RAD kids (p = .237 and .369 respectively) and one other association was revealed that had been masked in the larger sample.

For children without RAD, the SDQ scale scores for Hyperactivity in boys had a significantly higher variance (p = .019) even though the mean scores did not differ (p = .190). It is difficult to know how to interpret this finding. There was no difference on Hyperactivity variances for the children with RAD diagnosis (p = .751), nor did they differ on mean score (p = .229).

Traumatic Experiences and Scale Scores

I asked a large number of trauma-related questions, assessing both abuse/neglect and caregiving disruption, and also included a question about 'other traumatic experiences'.

The first item asked how many times the child had gained/lost caregivers, with five options - 0, 1, 2, 3 or 4 or more times. In the sample, 17 children had never gained or lost a caregiver, 12 had once, 13 had twice, 8 had 3 times and 18 had 4 or more times. I also asked about whether any one person had been in the child's life continuously (representing a possible stable attachment figure) and found that 33 children had a caregiver present continuously, 4 had continuous contact with a sibling but no parent, 4 had visitation with biological parent on a regular basis but did not live with one continuous person, and 25 had no one who had always been in their lives (2 parents did not answer this question).

I also asked at what age range the children lost a caregiver (if they had) and found that 13 children were under the age of 1, 33 were 1-5 years old, 11 were 6-10 years old, 6 were 11-15 years old and one was 16-18 years old (note that many kids lost caregivers at multiple ages). Overall, 39 kids had lost at least one caregiver after the age of 1 year (past the usual 'safe' age for forming new attachments). In addition, 17 of the children had spent at least some time in an institutional setting (such as an orphanage or inpatient psychiatric ward). Of those, 8 had been in an institution before the age of five, 6 between 5-10 years old and 4 after the age of 10.

Regarding abuse experiences, 10 children had been sexually abused (5 by peers, 3 by older children, 4 by caregivers and 2 by other adults, with many experiencing abuse from multiple sources) and 19 had been physically abused (3 by peers, 2 by older children, 15 by caregivers and 4 by other adults). In addition, 24 children had gone without basic physical caregiving needs (of which 7 were felt to have been neglected despite parental ability to provide care and 16 were felt to have been neglected due to parental inability, with one leaving this question blank). In addition, 23 children were often left unsupervised for long periods of time below the age of 8 years old.

In terms of other trauma, 6 children had witnessed violence (3 parent-perpetrated physical abuse, one found her mother dead from suicide, one witnessed physical and sexual abuse perpetrated by parents, and one saw her father resist arrest and get pepper-sprayed), 5 had suffered serious health issues, 2 were kept in confinement for long periods (eg left in a car seat all day), 4 frequently moved or lived in unstable housing and 3 had lost a sibling. In addition, 3 parents wrote in this section that their information about the child's history was extremely limited.

I then calculated scale scores for the children's total traumatic experiences. I computed three scores - one, (total trauma) summed up all the adverse life experiences the child had gone through, and the other two reflected Criterion A for the DSM-IV criteria for PTSD (PTSD risk) and Criterion C for RAD (RAD risk).
The RAD risk scale was formed by summing number of lost caregivers (0-4), whether there was a constant caregiver (2 = no, 1 = visitation or sibling, 0 = yes), losing a caregiver as a preschooler (yes = 1, no = 0), losing a caregiver after infancy (yes = 1, no = 0), institutionalization in early (yes = 2, no = 0), middle (yes = 1, no = 0) or late childhood (yes = 1, no = 0), parent-perpetrated sexual (yes = 1, no = 0) or physical abuse (yes = 1, no = 0), physical deprivation (deliberate = 2, situational = 1, no = 0), lack of supervision (yes = 1, no = 0), witnessing violence (yes = 1 if perpetrated by parent and/or resulting in parental death), physical confinement (yes = 1) and moving frequently/unstable home (yes = 1). The score could range from 0-20. Note that some items are counted more strongly, these reflect items I felt were especially likely to disrupt attachment.

The PTSD risk scale was formed by summing sexual abuse by peer (yes = 1, no = 0), older child (yes = 1, no = 0), parent (yes = 1, no = 0), or other adult (yes = 1, no = 0), physical abuse by the same four sources (yes = 1, no = 0 for each), witnessing violence (yes = 1), health issues (yes = 1) and confinement (yes = 1). The score could range from 0-11. Note that four items - parent-perpetrated physical abuse, parent-perpetrated sexual abuse, witnessing parent-related violence and confinement - are counted for both risk scales. These items all reflected parent-perpetrated abuse (or in one case a parent's death), and therefore reflect both a PTSD-style traumatic event and a disruption in caregiving.

The total score included the items for both RAD and PTSD scales, as well as losing a sibling (yes = 1) which I felt reflected a distinct type of experience (bereavement). The total score could therefore range from 0-38.

Note that the three children who were stated to have extremely limited information available nonetheless achieved total and RAD risk scores of 8, 11 and 5, although their PTSD risk scores were all 0. Therefore, I decided to exclude them from the PTSD risk analysis, but include them for RAD and total analysis. As a result, the PTSD risk scale has a sample size of 65 rather than 68.

The average scores were 6.03+/-4.455 for total trauma (with 11 kids scoring 0, and one child scoring 14), 5.56+/-4.101 for RAD risk (12 kids scored 0, and one scored 13), and .74+/-1.035 for PTSD risk (36 scored 0, and 2 scored 4). The higher RAD risk scores reflect both that I asked about that more, and that my sample included many more kids with RAD than PTSD diagnoses. Unsurprisingly, kids with trauma-related diagnoses had significantly higher scores on all three scales (total and RAD p = <.001, PTSD p = .001), suggesting that those diagnoses accurately reflected the children's traumatic history.

Trauma Impact

Next, I analyzed how trauma scale scores and trauma-related diagnoses were related to symptom scales. Total and RAD risk scores were correlated with RPQ total (r = .537 and .571), disinhibited (r = .363 and .384) and inhibited (r = .540 and .570) scores; SDQ Conduct Problems (r = .401 and .410), Prosocial (r = -.290 and -.287), Hyperactivity (r = .260 and .256), and Peer Problems (r = .281 and .245) scores; APSD total (r = .362 and .346), Narcissism (r = .329 and .314) and Callous-Unemotional (r = .349 and .328) scores; and the combined SDQ/APSD Antisocial (r = .500 and .493), Callous-Unemotional (r = .331 and .327) and Peer Problems (r = .298 and .261) scores. Meanwhile, PTSD risk was only associated with the two Peer Problems scales (SDQ r = .311, SDQ/APSD r = .320).

I thought the PTSD risk - Peer Problems association may be related to the peer-perpetrated abuse items, since peer problems could contribute to the risk of peer-perpetrated abuse (children with social impairments are more often bullied), however, when I calculated an alternate PTSD risk scale that excluded peer-perpetrated abuse, not only did the Peer Problems correlations remain (SDQ r = .360 and SDQ/APSD r = .359), but new correlations appeared between PTSD risk and APSD total (r = .351) and Callous-Unemotional (r = .334) scales. However, since most kids had a higher RAD risk score, and RAD risk score was significantly correlated with PTSD risk score both with (r = .489) and without (r = .558) peer abuse, it could be that these correlations are explained by attachment disruption rather than PTSD-style traumatic events.

When I ran a t-test on trauma-related diagnosis and scale scores, I found that RPQ total (p = .001), disinhibited (p = .042) and inhibited (p < .001) scores were all higher in children with trauma-related diagnoses, supporting the use of that scale to assess RAD (note that the RPQ temperament subscale had no correlation, p = .117, suggesting that the items on this scale are not related to RAD). In addition, trauma-related diagnosis was related to higher SDQ Conduct Problems (p = .049) and Peer Problems (p = .047) and lower Prosocial (p = .045) scores; higher APSD total (p = .013), Narcissism (p = .020) and Callous-Unemotional (p = .015) scores; and higher SDQ/APSD Antisocial (p = .002), Callous-Unemotional (p = .015) and Peer Problems (p = .044) scores.

Since psychopathy has shown some gender differences, I reran the analyses separating boys and girls. For boys, most of the correlations with total trauma and RAD risk scales were confirmed, with the exception of RPQ disinhibited (r = .257 and .307, neither significant), SDQ Hyperactivity (r = .311 and .297, neither significant) and APSD Narcissism (r = .290 and .280, neither significant). In addition, RAD risk was no longer correlated to Peer Problems (SDQ r = .336, SDQ/APSD r = .340, neither significant), though total trauma still was (SDQ r = .364, SDQ/APSD r = .369).

The boys also had new correlations between PTSD risk and outcome, with RPQ inhibited (r = .363) and APSD Impulsive (r = .353) correlated to PTSD risk. Interestingly, the correlation between PTSD risk and Peer Problems disappeared (SDQ r = .309, SDQ/APSD r = .304, neither significant).

Boys with trauma-related diagnoses differed from boys without such diagnoses on APSD Callous-Unemotional traits (p = .028, higher in boys with trauma diagnoses) and in the variance for SDQ Hyperactivity (p = .033, higher variation in boys without trauma diagnoses). However, the sample sizes were low (11 boys with trauma diagnoses, 21 without) so the absence of other correlations could be due to reduced power.

For girls, RPQ disinhibited (r = .314 and .330, neither significant) and SDQ Hyperactivity (r = .279 and .292, neither significant) were also no longer correlated with total trauma and RAD risk, however, many other correlations disappeared as well, with only RPQ inhibited (r = .373 and .435), SDQ Conduct Problems (r = .446 and .423), APSD Narcissism (r = .403 and .374) and combined SDQ/APSD Antisocial (r = .497 and .468) retaining their correlations for both scales, with RPQ total also correlating with RAD risk (r = .365) but not total trauma (r = .316, not significant).

Unlike the boys, the girls did not show any correlations between PTSD risk and any symptom scales. However, on the t-test by diagnosis, girls with trauma diagnoses also had higher APSD Callous-Unemotional traits (p = .028) and less variance on SDQ Hyperactivity (p = .033), just like the boys.

Overall, a clear picture emerges - the kids with attachment disruption are across-the-board more disturbed than kids without attachment disruption, with only a few items (eg APSD Impulsivity and SDQ Emotional Problems) failing to correlate with a history of attachment disruptions or a RAD diagnosis. Most importantly for my study purpose, a clear link emerges between attachment disruption and Callous-Unemotional traits (less clear in girls). This suggests that RAD kids are, as I suspected, more likely to show psychopathic traits. Indeed, if we define psychopathy as APSD scores above 29, 30.3% of the non-RAD (10 kids, 6 boys and 4 girls) and 52.9% of the RAD kids (18 kids, 7 boys and 11 girls) would be considered psychopaths.

If we instead divided them based on total trauma score, clumping 0-1 as 'no or single trauma' and 2 or more as 'multiple or severe trauma', we find that 12.5% of the no/single trauma group (only 2 out of 13 children) and 50.0% of the multiple/severe trauma group were above cutoff on the APSD. Despite the small number of children in the no/single trauma group, this difference was significant (p = .022 for APSD total score).

I also looked at whether the two CU scales (which showed no age effect for the full sample) showed any correlation with age within groups divided by attachment status. The pooled SDQ/APSD CU scale showed no correlation with age in either category, regardless of whether I analyzed by RAD diagnosis or trauma status. However, the APSD CU scale was correlated with age only for the RAD/traumatized kids, with p = .016 when dividing by RAD diagnosis and .019 when dividing by total trauma score. RAD children showed a pattern of increasing APSD CU scores with age.

Developmental Trauma Disorder

In DSM-5, a new category has been proposed, known as Developmental Trauma Disorder. The rationale for this category is based on research suggesting that kids with both serious attachment disruption and at least one traumatic event show a cluster of symptoms that tend to get them a mixed-bag of DSM-IV diagnoses, while at the same time these kids often do not meet criteria for PTSD.

To assess the DTD phenomenon in my study, I first checked how many kids fell into each category, and found that 12 children had experienced neither attachment disruption (RAD risk < 2) nor trauma (PTSD risk = 0), 5 had experienced trauma but not attachment disruption (PTSD risk > 0 and RAD risk < 2), 24 had attachment disruption only (RAD risk > 1 and PTSD risk = 0) and 24 had experienced both (PTSD risk > 0 and RAD risk > 1). I lumped the two groups without attachment disruption together to improve sample size.

I then ran an ANOVA comparing the three trauma groups on the symptom subscales, and found significant correlations for RPQ total (p = .001), disinhibited (p = .009) and inhibited (p = .008) scores; SDQ Conduct Problems (p = .005) score; APSD Callous-Unemotional (p = .039) scores and combined SDQ/APSD Antisocial (p = .006) and Peer Problems (p = .047) scores. In all cases, the 'both' group scored significantly higher than the 'no attachment disruption' group, with the 'attachment disruption only' group not significantly differing from either. This supports that children who meet Criterion A for DTD are more impaired than children with less severe traumatic history, and there are interesting parallels between many of the symptom criteria for DTD and the elevated scales. In particular, the SDQ and SDQ/APSD Peer Problems, APSD Callous-Unemotional and RPQ items resemble Criterion D (self and relational dysregulation), C1 (preoccupation or impaired capacity to read threat) and C4 (self-harm); and the SDQ Conduct Problems and combined SDQ/APSD Antisocial items seem related to Criterion B1 (inability to modulate affect states including prolonged tantrums) and C2 (impaired capacity for self-protection, including extreme risk-taking or thrill-seeking).

Although I did not ask the right questions to determine if any kids in my sample meet full criteria for DTD, it appears likely that many would. It's unclear if a RAD diagnosis would exclude DTD, or DTD exclude RAD, or if a child could receive both diagnoses. (Children can have both RAD and PTSD, as many of my sample demonstrate, so it's not unheard of to give children multiple trauma-related diagnoses.) More research should be done, as DSM-5 comes out, on the overlap between RAD and DTD.

Conclusion

Out of 56 kids whose APSD total could be calculated, 28 met cutoff for childhood psychopathy, the vast majority of whom had experienced significant attachment disruption. Approaching the problem from several angles, I consistently found a relationship between attachment disruption and psychopathic traits. What does that mean? Are those 28 kids psychopaths? How do I reconcile this with the finding that psychopathy has a strong genetic component?

I can see three possibles theories to explain this correlation.

Theory 1 - Phenocopy:

One possibility may be that RAD kids who score above cutoff on psychopathic traits are no more psychopaths than those who meet criteria for autism are autistic. Several studies (examples here, here and here) have found that a subset of RAD kids meet criteria for autism. However, these kids do not appear to be genuinely autistic, mainly because they respond a lot better to treatment than non-RAD autistic kids and often spontaneously improve when their environment is improved (particularly by transitioning from institution to foster or adoptive home). This has sometimes been termed 'institutional autism' or 'situational autism'.

If psychopathy in RAD is also a phenocopy, it is probably less prone to spontaneous remission than RAD-related autism. Indeed, my survey suggests that it may in fact worsen with age. However, it remains to be seen if psychopathic RAD kids are easier to treat than non-RAD psychopaths. In addition, perhaps the two groups may respond to different treatments, which would more strongly support the phenocopy hypothesis. For example, maybe RAD psychopaths are more likely to respond to attachment oriented therapies focusing on the development of an emotional bond between the child and caregivers, whereas non-RAD psychopaths might do better with reward-heavy behavior modification programs.

In addition, many studies have documented neuropsychological characteristics of psychopathy. In particular, psychopaths show reduced autonomic responsivity, reduced moral-conventional distinction, and reward dominance. Reward dominance has not, as far as I could find, been assessed in relation to either RAD or risk factors for RAD, but both autonomic responsivity and moral/conventional distinction have been assessed among children in foster care. Autonomic responsivity was only assessed in the context of the Strange Situation (a behavioral assessment of attachment), a very different setting from the studies into psychopathy, so it's unclear if RAD kids' autonomic responsivity resembles that of psychopathic kids.

However, Mullins and Tisak (2006)'s study of moral, conventional and personal domains in foster youth is more informative. They found both age and gender effects—boys aged 9-13 years showed a poorer moral/conventional distinction than girls of any age or boys aged 14-17 years. In addition, African American foster youth showed a poorer moral/conventional distinction than Caucasian youth. However, both African-American youth and younger boys showed inconsistent correlations, showing a poorer distinction on some measures but not others. It would be interesting to perform a similar study while more carefully controlling for both attachment disruption and psychopathic traits.
If psychopathy in RAD is merely a phenocopy, this would make distinguishing RAD from genetic psychopathy crucial. The researchers who proposed DTD did so on the basis that children with trauma and attachment disruption, who meet criteria for DSM diagnoses such as ADHD, ODD, depression and so forth, do not respond to treatment in the same way as children genetically predisposed to those conditions. In the case of psychopathy, this could result in children being 'written off' as untreatable on the basis of psychopathic personality traits, when in fact those traits could respond to the therapies that have been validated in the treatment of RAD and other trauma-based conditions. This further emphasizes the point made by the DTD researchers—ascertaining attachment and trauma history is vital in a child with behavioral and/or emotional problems.

Theory 2 - Multiple Pathways

Secondly, maybe there are multiple pathways to psychopathy. Perhaps psychopathy reflects a failure to form attachments, which can result either from a genetic inability to attach or from a lack of opportunity to. This is supported by Minnis et al (2007)'s finding of a substantial genetic component to RPQ scores (especially in boys). Interestingly, both this study and the one into psychopathy genetics used the Twins Early Development Study sample, a large sample of twins from the general population. Most of these twins have likely not experienced significant attachment disruption.

If a general population sample (such as the Twins Early Development Study sample) confirmed the correlation between RPQ and psychopathic traits, this would support this theory. Furthermore, evidence against the phenocopy hypothesis (described above) would support this hypothesis instead (or the third one, described below). In addition, if research found a stronger genetic link for both psychopathy and RPQ score in children without attachment disruption than in children with it, this would support the hypothesis that attachment disruption and inborn deficits form alternate pathways to psychopathy.

This would also predict something regarding treatment. Psychopathic kids with and without RAD should show similar response to treatment. If psychopathy is due to failure to attach, even genetic psychopaths should benefit from therapies designed to improve parent-child attachment. In particular, if we can identify the specific barrier to attachment in these children, ameliorate it, and then provide them with therapy to repair missed attachment, this should be an effective treatment.

This theory would also not contradict the findings, in samples with low rates of attachment disruption, which indicate that parenting style has minimal influence on psychopathy. In those samples, any problems in parenting would presumably be below the threshold needed to cause severe attachment problems.

It makes sense how psychopathy would be adaptive in a child with attachment disruption. Psychopathic children can manipulate parents to get what they need, while being somewhat shielded from the impact of losing a caregiver or being abused by one. If a secure attachment bond is not an option, it may be most adaptive to forgo forming attachments altogether. And in an environment where others do not meet the child's needs, selfishness may be necessary. In addition, both psychopathy and RAD are linked with disinhibited social behavior towards strangers—psychopathy is linked with a fearless temperament, characterized by reduced fear of novel stimuli, which includes strangers. Many researchers have suggested that disinhibited RAD is an adaptive strategy to gain care from anyone who may be willing to provide it, however transient they may be in the child's life. However, it's important to note that disinhibited social behavior also appears in children with Williams Syndrome, a genetic condition that appears to have no relationship to psychopathy or attachment security.

Theory 3 - Psychopathic Parents

A third possibility exists. The link between RAD and psychopathy could be related to a third factor—having a psychopathic parent.

Psychopathic traits have an obvious relevance to parenting behavior. A lack of empathy for others and chronic irresponsibility could easily result in neglect, abuse or abandonment of a vulnerable child. Therefore, a child born to a psychopathic parent (especially a psychopathic mother) is likely, due to the parent's personality traits, to be exposed to more attachment disruption and trauma.

Unfortunately, I was not able to find any empirical research assessing the impact of psychopathy on parenting. However, the impact of antisocial behavior on parenting has been documented, particularly for antisocial fathers. Florsheim et al (1999) found that among adolescent fathers, antisocial behavior predicts poorer parenting ability. In addition, several studies have found a complex interaction between father presence, father antisocial behavior and child antisocial behavior—if fathers are not antisocial, their absence puts children at greater risk of antisocial behavior; whereas antisocial fathers' presence increases risk of antisocial behavior in their children. Most of these fathers are not psychopaths, though, and the child behavior outcome that was measured was not psychopathy but merely a behavior correlated with psychopathy, so it's unclear what implications this has for parent-child transmission of psychopathy.

According to this hypothesis, among parents, psychopathy should predict severe deficits in parenting, resulting in attachment disruption in the child. This connection should be more pronounced than the already established link between parental antisocial behavior and poor parenting. Most importantly, if you control for parental psychopathic traits, history of attachment disruption should no longer be linked to psychopathic traits. In addition, if you assess twins who have experienced attachment disruption, zygosity should predict the degree of concordance in psychopathic traits better than shared environment does. In other words, identical twins should be more similar on psychopathy than fraternal twins, but twins raised together should be no more similar than twins raised apart (eg sent to different foster homes).

In addition, I couldn't find any studies into the rate of psychopathic traits in early-adopted children (adopted under a year old, with no other caregiving changes). These children are at minimal risk of attachment disorder, because the change in caregivers occurs before attachment is solidified (or in some cases present at all). However, children of psychopathic parents may be more likely to be put up for adoption, because psychopathic parents may be less inclined to keep the child or may have the child taken away at an early age. Theoretically, therefore, early-adopted children should have a somewhat higher rate of psychopathic traits.

It seems likely that attachment-disrupted kids would more often have psychopathic parents than early-adopted kids, but this has not been studied. Whatever the frequency, if samples are formed where biological parents are matched for psychopathic traits, this theory would predict no difference in the children's psychopathic traits between those two groups. Any difference in the rate of psychopathy between early-adopted and attachment-disrupted kids would be solely due to differences in biological parents' psychopathic traits.

Theory 4 - Combination of the Above

It may be that multiple theories explain part of the findings. For example, maybe the link between psychopathy and RAD operates on different mechanisms in girls as opposed to boys (indeed, at least one study suggested that callous-unemotional traits are more environmentally based in girls and genetic in boys) or in some races as opposed to others. Or maybe multiple mechanisms interact—for example, maybe psychopathy is influenced by genes, but the genetic threshold for showing psychopathic traits is lower if the child has also experienced attachment disruption. Or maybe some attachment-disrupted kids have genetically based psychopathic traits, while others have a phenocopy or an environmentally based psychopathy. Research into complex systems often doesn't leave you with a neat simple explanation.

As for me, I'm left still not knowing exactly what was going on with my cousin. All I know is that he's not alone. An estimated 50% of kids with significant attachment disruption show marked psychopathic tendencies. It remains to be seen what mechanisms contribute to this, but at the very least, it confirms the commonly held sentiment that psychopathic traits are associated with a bad childhood.