Outline: Adjectival Descriptors of Personality Disorders: A Convergent Validity Study of the MCMI-III
Robert J. Craig and Ronald E. Olson
A. Adjectival descriptors that correlate with specific Personality Disorders
A. Criterion Key Method
1. No randomization
2. Participants all received the same test
A. 202 African American man
B. Personality Disorders
1. Met DSM-IV
2. Lab Toxicology (urine sample)
A. Adjectival Checklist methodology
A. Used a biserial coefficient
B. Used a significant level at .20 instead of .01
C. Good convergent validity
A. Biased sample
B. Social desirability
Craig, R.J. & Olson, R.E. (2001). Adjectival Descriptions of Personality Disorders: A Convergent Validity Study of the MCMI-III. Journal of Personality Assessment, 77 (2) 259-271
The purpose of this study was to obtain adjectival descriptors that were most often associated with the diagnosis of a specific personality disorder. The theoretical influences for this study span from the limitations when using merely self-reports, or structured psychiatric interviews for diagnosing a personality disorder. Craig and Olson hypothesized that adjective checklist methodology as a measure of personality disorders may have some advantages over existing self-report measures of the DSM IV. Craig and Olsen found one limitation of other measures to be social desirability bias, where the subject wants to impress the experimenter or the psychiatrist. In this experiment they tried to limit that by using subjects who were from an outpatient drug abuse program. The two instruments were given as part of the general screening procedure, so the patients had no obvious reason to lie or exaggerate.
Craig and Olson wanted to find what adjectives best described each personality disorder as part of an effort to develop an adjective checklist scale for each personality disorder. They started with 900 adjectives for 14 personality disorders. The list was sent to 25 national experts in the field of personality disorders, and they were to reply if the adjectives were essentially or moderately related to the disorders. The 900 were reduced to 469 adjectives , which were then given to 202 males African American patients in treatment. The patients were asked to check an adjective if it described them, and to leave it blank if it did not.
The patients were also given the MCMI-III at the general screening procedure, and this test was used to correlate with the adjective checklist methodology. The test was found to have good convergent validity. Meaning the scores from the MCMI-III correlated strongly with scores from the adjective checklist methodology.
The results of this experiment were found through a correlation between adjectives and the MCMI-III. The correlation coefficient that was used the biseral correlation coefficient. This test was is a dichotomous test which, meant that the adjectives either described the participants completely or the adjectives did not pertain to the individual at all. The significant level for the biseral coefficient was at a .20 level instead of a .01 level. There were two reasons for this significant level being altered. The first is that there were too many adjectives that reached the .01. The second reason is that with a .20 level this makes the test more reliable.
The adjectives that were used to determine the results of the tests were then placed into different categories. These were known as scales descriptors. There were fourteen scale descriptors: schizoid, depressive, avoidant, dependent, histrionic, antisocial, narcissistic, aggressive- sadistic, compulsive, self- defeating, passive- aggressive, schizotypal borderline, and paranoid. These descriptors are applied to interpersonal domains and are specific ways to describe certain personality disorders. There are two extremes in the scale descriptors. The narcissistic scale only had two adjectives that correlated at a significance level of .20 or more. The two adjectives that correlated with this personality disorder are snobbish and smug. There are many reasons for there being a lack of convergent validity. The main reason is due to the fact that many of the adjectives could be representative of a healthy narcissistic personality disorder.
On the other hand, the anti- social scale descriptors had the most correlated items. The first reason is that the anti- social personality disorder has the most reliable personality disorder diagnosis. An example of this scale is:
Scale Descriptors r’
Self- Centered .46
After reviewing the data, the researchers found that it is not necessary to look at each domain of a specific disorder because the DSM-IV does not even go into that much detail. The DSM- IV is very reliable at assessing personality disorders. Overall, the results of this study show good convergent validity. This means that the adjective checklist methodology correlated with the DSM-IV at measuring the same construct which was adjectival descriptors of personality disorders. The researchers finish by commenting that further research is needed to refine the adjectives to make the assessing of personality disorders more accurate and reliable.
When reviewing a study done by Craig & Olsen, 2001. We found three interesting key points to contrast with three weaker or unclear points during our discussion.
1.) According to the experimenters, we found that the study had high convergent validity. The degree of correlation between the list of adjectives produced and the scores on the MCMI-3 was in fact high. The only scale that was not found to correlate highly was Narcissism.
2.) Previously, we had not heard of adjective checklist procedure that was used in this study. According to our sources (Craig & Olsen, 2001), this type of checklist is used in self-reported surveys, as shown in our study. We also found out that this type of testing procedure is also widely used in clinical work. Checklist methodology provides a basic understanding of a client’s personality, their mood, and it provides a method to measure changes that occur as a result of treatment. Also, specific needs of a special population can be assessed and accurately measured through this method (Craig & Olsen, 2001).
3.) Another aspect that we found upon analysis, was the degree of significance used. In most research, the significance for correlational procedures is set at .01. Our study set the significance at .20. All of the correlations between the describing adjectives and the associated personality disorders were found to be significant at a level greater than .01. With significance set at a higher level, the statistics have shown more reliability than lower correlations.
Three points that we found to be unclear as follows:
1.) The first subject that we found to be confusing or contradicting, was the idea of social desirability. In the begging of the article, voluntary participants were given the checklist upon admission into a drug treatment facility. From this fact, the experimenters assumed that these individuals would have no basis to lie. Since the men were already going to receive treatment, there would be no reason for social desirability measures. Then when discussing the results, social desirability was stated as the reason for high correlations of items in the dependent scale. One problem we found with these two statements, is that first off, you cannot assume that all participants will admit to not so favorable characteristics. And secondly, they tend to over lap and contradict each other. At first the experimenters were claiming to exclude social desirability, and then they actually stated it as a possibility for changing data.
2.) Another area that we found unclear or problematic, was the survey of characteristics. The way the self reported? Survey was set up was with nominal measures. The participant was given an adjective and asked if it was like them or not. The choices were limited, either the participant had that character trait i.e. angry or they did not.
The survey did not test for how much or how little of that quality was shown. There was not any indication of the degree. All or none type questions, limit choices of the participant and reduce the amount of each given trait.
3.) Lastly, we discussed the study’s external validity. First, each adjective may mean show something slightly different for each person. There was not a standardized definition given of each word. Different cultures can interpret the trait differently and can produce different responses. The participants, were all male, and from the same culture. The African American males were part of a biased sample that was already receiving treatment for addictions. Thus, a randomized sample of all different cultures was not taken.