Is There an Evidence Basis for
Attachment Therapy... As Its Proponents Claim?

by Jean Mercer, PhD
 Professor of Psychology
 Richard Stockton College
 Pomona, New Jersey


Family:Happy
For the last 10 to 15 years, the fields of medicine and clinical psychology have struggled to establish an evidence basis for practice — to determine what works and what does not. To meet this goal has required years of research as well as years of discussion about measurement of desired outcomes. Evidence-based medicine and psychotherapy come from the collection of data about thousands of patients, not just from anecdotes or subjective opinions. Research toward an evidence basis is carefully organized and monitored so the findings are not influenced by the researchers’ own hopes and wishes.


Advocates of Attachment Therapy (AT) claim that they too have established an evidence basis for their practices. Unfortunately, close examination shows that their research has failed to follow some elementary rules that could prevent the researchers’ bias from affecting their conclusions. An analysis of several research claims has already been made and published. (Mercer, 2002) None of the reports previously analyzed provides an acceptable evidentiary basis for the practice of AT.

Here are summaries of four more AT research claims — and some brief comments showing that this work, too, is not adequate evidence to give legitimate support to AT practice.

The RADQ

The Randolph Attachment Disorder Questionnaire (RADQ) is presented by AT proponents as if its development had followed the complex process needed to make a questionnaire both reliable and valid. However, a glance at the questionnaire items shows that this could not have been the case, for there is no attempt made to deal with the problem of response bias. Normally, when a questionnaire is properly constructed, each item appears in two forms, to prevent the answers from being influenced by a general tendency to agree (or, less often, to disagree). For example, if one item says, “The child always lies,” another must say “The child never lies.” To be consistent, a respondent who disagreed strongly with the first would also agree strongly with the second. Someone who simply agreed with everything would end up with a lower score than someone who was consistent in their opinions.

But the RADQ consists entirely of statements of one type, sentences attributing undesirable behaviors to the child. A parent’s tendency to agree creates an artificially-inflated score, suggesting that a serious emotional disturbance exists, whereas none would be diagnosed if both negative and positive statements were included.

The developers of the RADQ apparently shared the common erroneous belief that writing a questionnaire is simply a matter of putting down a list of questions or statements. Because the RADQ is flawed, its use in further AT research contaminates that research as well.

The Myeroff study.

The one published research report on the effect of AT (Myeroff, Mertlich, & Gross, 1999) is unfortunately seriously flawed in its design. Myeroff collected parents’ responses to questionnaires about their children in two situations, and compared the responses with each other.

All the parents had applied to bring their children for AT treatment at the Attachment Center at Evergreen, CO. About half the parents did bring the children to Evergreen; the others, for reasons of their own, did not do so. All of the parents in both groups replied to questionnaires about their children both before and after the first group had their treatment (the second group having no treatment at all).

Myeroff could then compare the amount of improvement from the first to the second questionnaire in the parents’ evaluations of the children who were treated, to the amount of improvement during the same period for the children who were not treated — improvement that could occur just because the children got older. Myeroff reported a significantly greater improvement in the group who were treated. This study has repeatedly been presented as evidence for the power of AT.

Myeroff is certainly to be commended for using a comparison group. (As we will see later, some AT researchers do not bother with this step.) However, her study suffers from confounded (confused) variables, so that it is impossible to know whether the difference was caused by AT or by some other factor.

For example, the group who were treated traveled to Colorado with their families, while the other children did not. Did the excitement of being away from home for several weeks cause a difference in children’s behavior or in parents’ attitudes?

Similarly, the treated group of children were separated from their parents for a two-week period. Did the separation itself affect either parents or children so that the parents rated the children as improved?

Again, the parents expected AT to help their children, or they would not have invested their time and money in this way. Were the parents influenced by their expectations so that they perceived treated children to have improved and untreated children to be the same or worse?

Finally, what were the reasons that kept the untreated comparison group from attending? Were these families suffering from difficulties — marital or financial problems, health issues, difficulties with other children — that prevented them from coming for treatment and also had a deleterious effect on the children they wanted to have treated?

Each of the factors mentioned above is a possible cause for the differences in improvement between the two groups. The effects of these factors are confounded or confused with the possible effect of AT, so that it is actually impossible to know what caused the difference, and it would be a mistake to claim that AT had been shown to be the cause.

The Becker-Weidman study

An unpublished study, reported by Arthur Becker-Weidman, used the RADQ (see discussion above) and a behavior checklist, administering the tests before and after treatment of 34 children. Becker-Weidman reported statistically significant improvements and attributed them to the treatment.

In the absence of a comparison group, this study suffers from confounded variables that make its interpretation impossible. Two such confounding factors can be stated simply:

First, as children get older, their behavior tends to become more acceptable to adults through a natural process of maturation. It cannot be known whether Becker-Weidman’s study shows the effects of maturation or of treatment, since the two were happening at the same time. It could even be that the treatment was harmful, but the positive effects of maturation overcame some of the damage.

Second, any person’s mood and behavior undergo natural variations, for better and for worse, over and above the effects of any experiences they may have. Parents are most likely to bring children for treatment when they are at their worst. Natural variation then goes in the direction of improvement for a while, whether treatment occurs or not. If treatment is given, people have a tendency to think that it has caused the change, but since variation and treatment are simultaneous it is impossible to know which factor is at work.

Becker-Weidman is thus in error when he claims this study as evidence that the treatment is effective.

Levy & Orlans’s study

Another unpublished study, this one by Terry Levy and Michael Orlans, has appeared only on the internet. The posted material makes it difficult to ascertain all the details of the study, but it is clear that the basic design is flawed in the same way as Becker-Weidman’s work. There is no comparison group.

Levy and Orlans gave a symptom checklist to the parents of 50 children, before and then after the children went through a two-week intensive treatment program. Some responses were collected as much as three years later, although the report does not show whether these were additional test administrations or not; Levy and Orlans refer to their study as “longitudinal”, which ordinarily means that there are repeated measurements over some period of time.

As was the case for Becker-Weidman’s report, Levy and Orlans have confounded the treatment variable with both maturation and natural variation. Treatment occurred at the same time as the two other causes of change. It is erroneous to conclude, as Levy and Orlans have done, that “the research proves that Corrective Attachment Therapy does help children with histories of maladjustment.” This poorly-designed study cannot contribute to any conclusions about the effectiveness of this type of treatment.

Conclusion

There continues to be no existing evidence basis for the claim that Attachment Therapy ameliorates childhood emotional disorders.

References