Constructivist
Psychology
Network

www.constructivistpsych.org

Constructivist Chronicle

Research Spotlight: Understanding Dissociative Identity Disorder Using Hierarchical Classification Analysis
Vol. 8, Issue 1 (Spring 2004)

Q.  What got you started in this line of research?

A.  Rue Cromwell. Rue graduated from Ohio State in 1955 after having done his Masters Thesis with George Kelly (he did his doctoral dissertation with Julian Rotter). His 50 years in psychology have afforded him a vast breadth and depth of knowledge in and beyond psychology. The real delight of working with Rue is that he is open, honest, intensely curious, and the most intellectually stimulating conversationalist I have ever known. (This is not just an ode to Rue, I think you will find that many people involved in PCP have had a similar experience with him).
 
Q. What background led you to start on this?

A.  Psychology is my second career. Before I made the decision to go to graduate school, I spent 23 years working in New York theater. When Rue first discussed Kelly's constructive alternativism I immediately felt at home. After all, a collaborative form of constructive alternativism is at the heart of the rehearsal process, i.e., determining, from the myriad possibilities, what a character is saying, what it means, and how to say it to best communicate that meaning. I was not at all surprised to find out that one of Kelly's first jobs out of graduate school was teaching drama.

In general, psychology has afforded me the opportunity to ask many of the same fundamental questions about the human condition that I did in the theater, i.e., who is this person, what is s/he doing and why?  Most specifically, I am interested in how trauma experiences effect an individual's way of looking at the world (conceptual structure).     

In addition, I've always had a strong interest in math. HICLAS: hierarchical classification analysis, employs mathematical set theory to build a graphic representation of the structure of an individual's personality. To me, theater, math, analytic questions on the GRE, and hierarchical classification analysis are fun problem solving exercises.

My research follows the trail blazed by Dr. Ken Sewell.  When he was a grad student working with Rue, Ken did two studies in which he used hierarchical classification analysis on Kelly Role Construct Repertory Grids (rep grid) data, one with of individuals diagnosed with Post Traumatic Stress Disorder (PTSD) and a pilot study of 2 individuals diagnosed with Dissociative Identity Disorder (DID).  The hypotheses for the study  that Rue, Karl Yngvar Dale (University of Tromso, Norway), and I reported on at the XV International PCP conference at the University of Huddersfield, were derived from those studies.  We examined the conceptual structure of individuals who had been hospitalized with DID in Norway.  Simultaneously, I was gathering rep grid data from individuals diagnosed with Borderline Personality Disorder (BPD).  I will report on those findings at the NAPCN conference in Memphis in June 2004.
 
Q.  What problem were you hoping to address?

A.  One way to conceptualize the diagnoses of PTSD, BPD, and DID is on a continuum.  They appear to represent respectively increasing pathological reactions to traumatic experience.  These various reactions seem to be a function of the frequency and intensity of the experience, the genetic and psychological makeup of the individual, the age of occurrence, social support system, etc.  More important here is the idea that in all of these diagnostic categories, the individual's ability to construe future events (i.e., the individual's world view) is adversely affected.  We employ HICLAS to examine for structural patterns (that represent the traumatic experience, or the resultant impeding construct), in the conceptual structure of individuals with these diagnoses.  (If and) When we are successful, there will be many implications for assessment and treatment.
 
Q.  How did you get teamed up with Karl Yngvar Dale of The University of Tromso, Norway?

A.  Arne Holte, Karl's major professor, is an old friend of Rue's.  Both are old friends of DeBoeck, who created HICLAS.  Arne and Rue were discussing HICLAS and the work Rue had done with Ken Sewell about the time when Karl was looking for a dissertation topic.  Karl found it interesting.  He collected data from the 13 bona fide cases of individuals diagnosed with DID that he found in Norway.  As I was working on a similar project (with individuals diagnosed with BPD) and had spent considerable time learning how to operate and interpret HICLAS for my dissertation, I agreed to act as the psychometrician on Karl's study.
 
Q.  Can you explain more about what kind of analysis HICLAS is?
 
A.  The HICLAS algorithm (De Boeck, 1989; Rosenberg, DeBoeck, & Mechelen, 1996) presents a comprehensive representation of the binary matrix.  Using mathematical set theory, it identifies overlapping and separate patterns within the elements (objects) and constructs (attributes) respectively.  Instances of overlapping sets are used to order the matrix solution with the assumption that an overlap of two or more lower order (subordinate) sets implies an asymmetrical relationship with a higher order subsuming (superordinate) set.    An important feature of this model is that the structure can be rendered as a graphic representation that summarizes the membership of the elements into element classes and of the constructs into construct classes, the order of relations within each of these two sets of classes, and the way the element classes and the construct classes are associated to each other.
 
Principal components factor analysis seeks orthogonality among its final groupings (factors) and begins with only the symmetric index of relationships.  However, the HICLAS algorithm is an iterative procedure that seeks the best hierarchical structure already present in the matrix.
 
The HICLAS model is dependant upon its user choosing a “rank” for the matrix.  The rank of the matrix, in effect, sets the number of bottom classes that the hierarchical solution may have.  The rank is somewhat analogous to (eigenvalue in – or - the number of factors in) factor analysis in that it presents an arbitrary determination of number of groupings to be derived.  The rank size may vary from one to the total number of variables.  An ultimate choice of rank is usually determined on the basis of optimal utility and interpretability.
 
Set theoretical relationships implicit in two way matrices and their psychological significance were detected in early studies of person perception (Rosenberg, DeBoeck, & Mechelen, 1996).  Applications in other areas proceeded, particularly in the area of psychopathology, providing a new window to understanding psychopathological processes and the role that the patient’s view of self and others plays in these processes.
 
Q.  HICLAS is able to identify a divergence at the super-ordinate level of conceptual links. What is the significance of that?

A.  One would expect that the HICLAS conceptual hierarchy for both elements and constructs would be pyramidal in shape, i.e. dogs, cats, and tigers would converge into a classification of animals.  This higher order (superordinate) class of animals could then combine with a class of plants to yield yet a higher (more superordinate) class of “living things.”  We discovered that elements and constructs found in the base level of the conceptual hierarchy of individuals diagnosed with DID diverged into several higher level classifications.  This suggests the possibility that  (a) these individuals may be experiencing dissociation at “higher levels” of their conceptual hierarchy, and/or  (b) as you go up the conceptual hierarchy for these individuals, their worldview gets more complex and disorganized rather than simpler and more organized.