|Home | About | Journals | Submit | Contact Us | Français|
The current study utilized a large undergraduate sample (n = 536), oversampled for DSM-IV-TR obsessive-compulsive personality disorder (OCPD) pathology, to compare eight self-report measures of OCPD. No prior study has compared more than three measures and the results indicated that the scales had only moderate convergent validity. We also went beyond the existing literature to compare these scales to two external reference points: Their relationships with a well established measure of the five-factor model of personality (FFM) and clinicians' ratings of their coverage of the DSM-IV-TR criterion set. When the FFM was used as a point of comparison the results suggested important differences among the measures with respect to their divergent representation of conscientiousness, neuroticism, and agreeableness. Additionally, an analysis of the construct coverage indicated that the measures also varied in terms of their representation of particular diagnostic criteria. For example, while some scales contained items distributed across the diagnostic criteria, others were concentrated more heavily on particular features of the DSM-IV-TR disorder.
The American Psychiatric Association (APA, 2000) defines Obsessive Compulsive Personality Disorder (OCPD) as “a pervasive pattern of preoccupation with orderliness, perfectionism, and mental and interpersonal control, at the expense of flexibility, openness, and efficiency” (p. 725). Torgersen's (2005) review of eight epidemiological studies found (perhaps surprisingly) that OCPD had the highest median prevalence rate of all the personality disorders (PDs), occurring in nearly 2% of the adult population. Subsequently, the National Epidemiological Survey on Alcohol and Related Conditions, which represents the largest national sample collected to date, indicated that OCPD is the most prevalent PD within the general population, occurring at a rate of 7.9% (Grant et al., 2004).
This personality disorder has been included in all previous editions of the DSM and traces its origins to Freud's (1908) description of the “anal” character traits of orderliness, parsimony, and obstinacy (Pfohl & Blum, 1995). Despite this long history, the construct described in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR; APA, 2000), has traveled a winding road that has seen substantial changes to what are considered its essential features (see Costa, Samuels, Bagby, Daffin, and Norton, 2005, for a more complete historical review). For example, the DSM-I (APA, 1952) description of what was termed compulsive personality featured overconcern “with adherence to standards of conscience or of conformity,” overinhibition, overconscientiousness, “an inordinate capacity for work,” rigidity, chronic tension, and a “lack [of] a normal capacity for relaxation” (APA, 1952, p. 37). There was little change from the first to the second edition of the diagnostic manual but a substantial shift in the third (Costa et al., 2005; Pfohl & Blum, 1995). Notably, the DSM-III criteria did not include overconcern with morality, overconscientiousness, difficulty relaxing, or chronic tension. DSM-III did retain the elements of perfectionism and workaholism but shifted the essential feature to include a “restricted ability to express warm and tender emotions” (p. 326). The diagnostic criteria underwent another substantial revision for DSM-III-R as four additional criteria were added “to better represent the original psychoanalytic constructs of parsimony and orderliness as well as obstinacy” (Widiger, Frances, Spitzer, & Williams, 1988, p. 791). As such, an essential feature became “a pervasive pattern of perfectionism and inflexibility” (APA, 1987, p. 354). However, the OCPD diagnosis was altered again with the publication of DSM-IV (APA, 1994) such that the restricted expression of affection that had been a defining feature within DSM-III was removed entirely from the diagnostic criterion set (along with indecisiveness) (Pfohl & Blum, 1995).
In short, the history of OCPD is characterized by significant alterations to its core features, additions and subtractions to its criterion sets, and indecisive shifts in its title (Costa et al., 2005; Pfohl & Blum, 1995). It might not be surprising, then, if extant measures conceptualize and assess OCPD in very different manners. Indeed, there are now a number of different self-report measures that include an OCPD scale (McDermutt & Zimmerman, 2005) and they have obtained questionable convergent validity (Widiger & Boyd, 2009).
Widiger and Boyd (2009) compiled a list of 24 studies that have reported convergent correlations between at least two self-report OCPD inventories. The 38 correlations reported within these studies ranged from a low of −.50 (Zarrella, Schuerger, & Ritz, 1990) to a high of .70 (Hicklin & Widiger, 2000). The median convergent value across the studies was −.07; a value that was by far the lowest among the 10 PDs. In fact, the next lowest median value was for narcissistic PD, which obtained a median convergent value of .55. Clearly, this finding indicates that the existing measures assess OCPD in quite different ways.
A good number of studies have assessed the convergent validity of measures of OCPD, although they are typically confined to a comparison of just two instruments (Widiger & Boyd, 2009). None has considered more than three instruments at any one time, and none has compared multiple self-report inventories with respect to an external indicator. One potentially useful reference point by which to compare these alternative assessments of OCPD might be the five-factor model (FFM). The FFM was developed to provide a reasonably comprehensive description of general personality structure, consisting of the five broad bipolar domains of extraversion (vs. introversion), agreeableness (vs. antagonism), conscientiousness (vs. impulsivity), neuroticism (vs. emotional stability), and openness (vs. closedness to experience). The FFM has been recommended as a basis for comparing and integrating seemingly diverse personality constructs (Ozer & Reise, 1994). Goldberg (1993) has even analogized the domains of the FFM to the coordinates of latitude and longitude used to map the world.
A number of studies have employed the FFM as a metric with which to compare PD assessment instruments (e.g., Trobst, Ayearst, & Salekin, 2004). For example, Zuroff (1994) compared two alternative measures of dependency with respect to their relationship to the NEO Personality Inventory (NEO PI; Costa & McCrae, 1985). Costa and McCrae (1990) also used the NEO PI as a point of comparison for different versions of the Millon Clinical Multiaxial Inventory (MCMI; Millon, Millon & Davis, 1997). Hicklin and Widiger (2005) compared alternative measures of psychopathy and antisocial PD with respect to the FFM. Additionally, Miller and Campbell (2008) as well as Samuel and Widiger (2008) recently administered the NEO Personality Inventory – Revised (NEO PI-R; Costa & McCrae, 1992), along with several existing measures of narcissism and found that differences among these scales were readily understood with respect to their differential relationship with FFM concepts.
An additional method of understanding the differences among measures is to investigate how well they cover the OCPD construct, as rated by clinicians. This type of analysis would indicate whether the items contained within each instrument differ in terms of their representation of the individual diagnostic criteria for OCPD. Thus, the current study proposes to compare eight scales assessing OCPD with respect to their relationship with FFM personality traits, as well as their coverage of the diagnostic criteria included within DSM-IV-TR (APA, 2000). The OCPD self-report inventory scales included in this study were those most commonly used in prior research (Widiger & Boyd, 2009). It is particularly useful to make these comparisons within the same sample to provide the most direct, comprehensive, and unobstructed comparison of the respective measures. It should be noted that this comparison does not necessarily seek to identify the most valid measure but can ascertain meaningful differences in the traits and features emphasized by and/or excluded from the various instruments.
The study was approved by the appropriate institutional review board and the sample was drawn from the introductory psychology student participant pool at the University of Kentucky. In order to maximize the presence of DSM-IV–TR OCPD symptomatology, the OCPD scale from the Personality Diagnostic Questionnaire-4 (PDQ-4; Hyler, 1994) was included in a packet of pre-screening measures that were administered to over 1,000 potential participants. The PDQ-4 was chosen for this purpose because it is a commonly used PD measure (Widiger & Boyd, 2009) and its brevity allows for easy implementation as a screening measure. Individuals who endorsed at least five of the eight PDQ-4 items were formally invited (via email) to participate in the current study. After 150 from this group had participated to ensure the oversampling for OCPD pathology, the study was opened to the entire subject pool in order to expand the range. The study was conducted using an online survey tool (i.e., MRinterview). The scales were administered in the order they appear in the materials section, below. Participation took approximately 90-120 minutes and participants were invited at several intervals to take brief breaks if necessary. In total, 559 participants provided informed consent and completed selected scales from personality and personality disorder instruments over the course of approximately two hours. Of the total sample, twenty-three (4%) of the participants provided incomplete protocols and were dropped from the study, yielding a final sample of 536 participants, 155 (29%) of whom had been pre-screened for elevated OCPD symptomatology.
The NEO PI-R (Costa & McCrae, 1992) is a measure of the five-factor model of personality and contains 240 items that are rated on a Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). This instrument is composed of five broad domain scales, which are each, in turn, assessed by six underlying facet scales. In the current sample, domain scales had alphas that ranged from .86 to .91 with a median value of .90. At the facet level, the alphas ranged from a low of .50 to a high of .81, with a median value of .74.
The MCMI-III (Millon et al., 1997) is a 175-item true/false self-report inventory, developed in accordance with the DSM-IV, which assesses 14 PDs as well as ten other clinical syndromes. The MCMI-III is among the most frequently used self-report inventories in clinical practice (Camara, Nathan, & Puente, 2000) and its 17-item OCPD scale was administered in this study.
The MMPI-2 (Butcher et al., 1989) is a 567-item true/false self-report inventory that provides scores on ten clinical scales as well as supplemental scales. Morey, Waugh, and Blashfield (1985) selected those items from the inventory that appeared to represent DSM-III (APA, 1980) OCPD and demonstrated good internal consistency. The resulting scale contained 13 items. Somwaru and Ben-Porath (1995) subsequently created their own OCPD scale from the MMPI-2 to identify DSM-IV OCPD utilizing 10 of the items from Morey et al. as well as including 10 additional items. Both scales were included in the current study.
The OMNI (Loranger, 2001) consists of 375 items designed to assess both normal and abnormal personality traits, including ten trait scales corresponding to the DSM-IV PDs. Items are scored on a scale ranging from 1 (definitely agree) to 7 (definitely disagree). The OCPD scale containing 18 items was administered in this study.
The WISPI-IV (Klein & Benjamin, 1996) consists of 204 items that are scored along a scale that ranges from 1 (“not at all, never applies to me”) to 10 (“extremely, always applies to me”). The WISPI-IV OCPD scale containing 18 items was administered in this study.
The SCID-II-PQ (First, Gibbon, Spitzer, Williams, & Benjamin, 1997) is a self-report, screening instrument for the SCID-II clinical interview, which assesses each of the DSM-IV PDs. It contains a total of 117 items that are answered as either true or false. The nine items corresponding to the diagnostic criteria for OCPD were administered.
The SNAP (Clark, 1993) is a 375-item true/false instrument that assesses a dimensional model of personality disorder containing 3 temperament and 12 primary trait scales, as well as the DSM-III-R (APA, 1987) PDs. The OCPD scale containing 23 items was administered in this study.
The PDQ-4 (Hyler, 1994) is a 99-item true/false self-report inventory that assesses 12 PDs according to the DSM-IV. The PDQ-4 is commonly used within clinical research (Bagby & Farvolden, 2004; Widiger & Boyd, 2009) and its OCPD scale has eight items, corresponding to each diagnostic criterion for the disorder. The entire PDQ-4 was administered in this study.
Six clinicians within the Lexington, Kentucky, community and familiar with the OCPD construct were asked to rate how well the individual items from the OCPD instruments assessed each of the eight DSM-IV-TR diagnostic criteria. They were provided with a list of all of the OCPD items from all eight instruments, and were asked to code each item with respect to all eight diagnostic criteria. These ratings were provided on a 0-4 Likert-type scale, where 0 = not representative of the given criteria, 1 = slightly representative, 2 = moderately representative, 3 = mostly representative, and 4 = fully representative of the given criteria. The clinicians were compensated $25 for their time and effort.
Consistent with the student population as a whole, the sample was largely female (62.7%) and predominantly Caucasian (91.0%). Four percent of the sample was African-American, 1.7 percent Asian-American, and additional 3.2 percent described themselves as “multiracial” or “other.” Two percent of the sample listed their ethnicity as Latino/a. The sample consisted primarily of first year students (68.4%) in their first semester of college, but also included 23.2 percent 2nd years, 6.0 percent 3rd years and 1.3 percent 4th years. The participants ranged in age from 18 to 27, with a mean of 18.8. Two hundred and thirteen (40.9%) of the participants reached the diagnostic threshold for OCPD on the PDQ-4 and 50.4% met criteria using the SCID-II PQ. Given the tendency of these screening instruments to diagnose at higher rates than structured interviews (Bagby & Farvolden, 2004), one should not conclude that 40% or 50% of the sample would or should be diagnosed with OCPD. Nonetheless, these results do suggest that the pre-screening was successful in obtaining an adequate range of OCPD symptomatology, as defined by these self-report measures.
Table 1 presents additional descriptive statistics for the eight OCPD scales. The internal consistency values of these scales ranged from .44 to .90. Because the scales varied tremendously in length, the average corrected item-total correlations were also calculated as a second indicator of internal consistency. These values were small to medium (e.g., Cohen, 1992) suggesting that the OCPD scales are quite heterogeneous in content, assessing a collection of disparate traits rather than a single, unidimensional construct.
Table 1 also presents the skewness and kurtosis of the distribution of scores for these scales. The PDQ-4 and WISPI evidenced significant positive skew, indicating that the right half of their distributions were longer than would be expected in a normal distribution. Additionally, the SNAP, MCMI-III, and the two MMPI-2 scales exhibited negative kurtosis values, indicating these distributions were significantly platykurtic (i.e., flatter, such that the “peak” of the distribution is wider and the tails thinner than a normal distribution).
Table 2 provides the correlations among the eight OCPD scales. The median value was .49, suggesting significant convergence, albeit not as strong as one might prefer among measures purportedly assessing the same construct. It appeared that, consistent with the findings of Widiger & Boyd (2009), the MCMI-III stood apart from the other measures. However, there were other weak findings as well, such as the convergence of the MMPI-2 scales with the WISPI-IV.
Table 3 presents the discriminant validity correlations, which were calculated by correlating each OCPD scale with the PDQ-4 scale for each of the other nine DSM-IV-TR personality disorders. Cronbach's alpha values for the 9 other PDQ-4 scales ranged from a low of .50 (schizoid) to a high of .69 (avoidant), with a median of .58. The patterns of the discriminant correlations are quite similar across measures, with the OCPD scales most differentiated from antisocial PD and moderately correlated with each of the other PDs. The MCMI-III was again a clear exception as it correlated negatively with every other PD and had a mean discriminant correlation of −.23. In contrast, the other OCPD scales all obtained mean discriminant correlations that were significantly positive. The MMPI-2 scales obtained the worst discriminant validity, with the Somwaru and Ben-Porath scale correlating as high as .54 with the PDQ-4 avoidant PD scale (a value higher than its correlation with the PDQ-4 OCPD scale). The remaining OCPD scales did appear to evidence sufficient discriminant validity, with mean correlations at or below .30.
Table 4 presents the correlations of each OCPD scale with the domains and facets of the NEO PI-R. Because this table contains such a large number of correlations (i.e., 280), we utilized a Bonferroni correction to minimize the chance of Type I error. A Bonferroni correction for this number of correlations (i.e., .05/280) is .00018, thus we set p < .0002 (two-tailed). The results down each column of Table 4 represent the unique FFM profile of each OCPD scale. When looking across a row one can compare the eight OCPD scales with respect to a particular domain or facet of the FFM.
It is clear from these correlations that the measures converged in terms of their relationships with certain FFM domains and facets. For example, nearly all the measures obtained strong positive relationships with the domain and facets of neuroticism, suggesting that a tendency toward negative affect is a core part of the conceptualization of OCPD offered by these instruments. Another commonality among many of the measures was a positive relationship with conscientiousness, as five of the measures achieved a significant correlation with this domain. In addition, the MCMI-III correlated significantly with all six facets, while the SNAP correlated with five, and the OMNI, SCID-II PQ, and WISPI-IV each correlated with three facets of conscientiousness.
However, it was also clear that the OCPD instruments were not solely confined to conscientiousness and neuroticism. Seven of the eight instruments correlated negatively with openness to actions, suggesting that closedness to doing or trying new things is a fundamental part of each instrument's assessment of OCPD. Additionally five correlated negatively with the agreeableness facet of trust and three with the extraversion facets of gregariousness and positive emotions.
Nonetheless, there were also notable differences among the measures in their relationships with the NEO PI-R. For example, the disaggregation of the OCPD scales in terms of the FFM helps to understand the relatively weak convergent validity of the MCMI-III with the other six OCPD scales. The MCMI-III obtained, by far, the highest correlation with conscientiousness (r = .71). The MCMI-III OCPD scale also correlated negatively with neuroticism and positively with agreeableness, whereas all of the others correlated positively with neuroticism and two correlated negatively with agreeableness.
The MCMI-III, however, did not provide the only disparate description of OCPD, with respect to FFM personality traits. Table 4 indicates that the MMPI-2 assessments of OCPD were confined largely to facets of neuroticism. These scales obtained the highest correlation with neuroticism and failed to correlate significantly with conscientiousness (even correlating negatively with the facet of self-discipline). Thus, it might not be surprising that the MCMI-III and the MMPI-2 Somwaru and Ben-Porath (1995) scales actually correlate negatively. The FFM profile obtained by the PDQ-4 was similar to that of the MMPI-2 scales, in that it correlated primarily with neuroticism, although not nearly as strongly. However, it was not significantly correlated with any other FFM domain. Thus, with respect to the FFM traits, it appears as though the PDQ-4, like the two MMPI-2 scales, conceptualizes OCPD primarily in terms of neuroticism.
We also sought to identify how well each instrument covered the DSM-IV-TR OCPD construct using the clinicians' ratings of each item from all measures in terms of the eight diagnostic criteria for OCPD within DSM-IV-TR (APA, 2000). These ratings were then averaged across the clinicians such that every item had a mean rating for each diagnostic criterion. These mean item ratings were then examined for each item within each instrument to determine how strongly the instrument represented a given diagnostic criterion1. Those items that received a mean rating of at least 2.5 were considered as indicative, as this was the minimal score necessary to indicate that an item was at least “moderately” to “mostly” representative of a respective criterion. Aggregated interrater reliability was calculated using intraclass correlation coefficients for the composite of six clinicians, with items treated as cases and raters as variables. The values ranged from a low of .54 (criterion 8, “rigidity and stubbornness) to a high of .85 (criterion 5, “unable to discard worthless objects”), with a median of .70.
It is notable that only the PDQ-4, SCID-II-PQ, and OMNI scales include at least one item for each of the eight diagnostic criteria, consistent with their having been constructed explicitly to represent the diagnostic criterion sets. The WISPI and the SNAP both included an item for six criteria, but failed to include any items assessing criteria 7 (“miserly spending style”) and 8 (“rigidity and stubbornness”). The MCMI-III included items assessing five of the diagnostic criteria, but did not include items for criterion 2 (“perfectionism”), 5 (i.e., “unable to discard worthless objects”), or 6 (“reluctant to delegate”).
The MMPI-2 scales appear to be somewhat distinct as they did not include items representing five of the eight diagnostic criteria, including 1 (“preoccupied with details), 2 (“perfectionism”), 3 (“workaholism”), 7 (“miserly spending style”), and 8 (“rigidity and stubbornness”). In fact, fully 18 of the items from the MMPI-2 scales did not represent any OCPD diagnostic criteria. In examining these items more closely, they appear to be more representative of Axis I anxiety disorders than of OCPD (e.g., item 166 “worry about sex”) or a more general ruminative anxiousness (e.g., item 328 “unimportant thoughts run through my mind for days”).2 The inclusion of items that did not assess any criteria though was not unique to the MMPI-2. In fact, 10 of the 17 items from the MCMI-III and 10 of the 23 items from the SNAP were not related to a single criterion. However, in the case of the SNAP, 7 of these 10 items were actually written to assess DSM-III-R criteria which were not retained (“indecisiveness” and “restricted affection”) or were significantly altered (“not generous with time, money or gifts”) for DSM-IV-TR.
The results from the current study indicate that although eight self-report scales for the assessment of OCPD evidenced moderate convergence (median = .49) with one another and reasonable divergence from other PD constructs (median = .26). This convergent value was substantially higher than the median value reported by Widiger and Boyd (2009), but was still perhaps lower than might be expected for measures of the same construct3. Considering the fluctuations and alterations that have characterized the history of the disorder (Costa et al., 2005), it might not be surprising that extant measures would vary in terms of their operationalizations. A closer examination of the results suggests that a great deal of the variability among measures is attributable to somewhat discrepant assessments offered by the scales from the MMPI-2 and MCMI-III. Although there are important differences among the other five scales, their convergence was more robust, with a median correlation of .61 among them.
Beyond their convergence, important differences were also noted among the measures with respect to their relationships with the traits of the FFM and their representation of the DSM-IV-TR (APA, 2000) diagnostic criteria. Although there was a great deal of consistency in how most of the scales related to the FFM (i.e., high conscientiousness, high neuroticism, low openness to actions, and, to a lesser extent, introversion), the magnitude of these relationships varied. Additionally, in the cases of the MMPI-2 and MCMI-III scales, the differences in their relationships with the FFM variables can help to explain their somewhat lower convergence with other measures.
For example, both the Morey et al. (1985) DSM-III and the Somwaru and Ben-Porath (1995) DSM-IV MMPI-2 assessments of OCPD were confined largely to the domain and facets of neuroticism, with no representation of maladaptively high conscientiousness or other FFM domains (perhaps contributing to its relatively worse discriminant validity). The MMPI-2 scales' strong relationships with neuroticism are not necessarily problematic. In fact, inability to relax (APA, 1968) and chronic tension (APA, 1952) were included within the earliest formulations of this PD, but disappeared with DSM-III (APA, 1980). Clinicians also consider traits concerning anxiousness, tension, and self-consciousness to be characteristic of prototypic cases of OCPD (Samuel & Widiger, 2004). Nonetheless, the fact that, according to the clinicians' ratings, the two MMPI scales fail to include any items assessing five of the DSM-IV-TR diagnostic criteria and actually contain several items unrelated to any OCPD diagnostic criteria does raise concerns about the MMPI-2 scales that warrant future research.
These findings for the MMPI-2 are perhaps explained by the unique method with which its OCPD scales were constructed. Whereas the other measures are comprised of items written to provide an adequate to optimal assessment of OCPD, the MMPI-2 scales were confined to items that were already included within this pre-existing self-report inventory. Thus, to the extent that the item pools of the MMPI or MMPI-2 lack an adequate representation of OCPD pathology they would be unable to represent these aspects in the ensuing scale.
In fact, previous research has suggested that the MMPI and MMPI-2 do not contain items assessing conscientiousness. A principal components analysis of the MMPI item pool indicated that only neuroticism was well represented from the FFM (Costa, Zonderman, McCrae, & Williams, 1985), with the “absence of any items or scales corresponding to…Conscientiousness” (p. 932). This finding was replicated by Costa, Busch, Zonderman, and McCrae (1986) and by Trull, Useda, Costa, and McCrae (1995) with the MMPI-2. Thus, while Morey et al. (1985) and Somwaru and Ben-Porath (1995) may have indeed selected the best available OCPD items from the MMPI and MMPI-2, their selections were limited by the instruments' existing content. As a result, the MMPI-2 OCPD scales do not appear to contain FFM conscientiousness or some specific aspects of the OCPD construct (e.g., workaholism, perfectionism, and preoccupation with details).
Aspects of conscientiousness are evident among most of the APA (2000) diagnostic criteria, including preoccupation with details, rules, lists, and order (i.e., the FFM facet of order), perfectionism (i.e., an excessive emphasis on competence), devotion to work and productivity (i.e., excessive achievement-striving), and even a criterion that refers explicitly to conscientiousness. In fact, a preoccupation with orderliness and perfectionism are two of the three defining features of DSM-IV-TR OCPD (APA, 2000, p. 725). Consistent with this expectation, the SNAP, SCID-II-PQ, OMNI, MCMI-III, and WISPI-IV scales all obtained significant correlations with conscientiousness.
If one considers OCPD to be simply a disorder of excessive conscientiousness, the MCMI-III might be thought of as providing one of the best assessments, as it correlated substantially with this domain (r = .71). However, potentially problematic to the validity of the MCMI-III is its very low convergent correlations with other OCPD scales. It appears, though, that this low convergence is not due to its heavy emphasis on conscientiousness, but rather because the MCMI-III obtains correlations with neuroticism and agreeableness that are opposite in direction from the other measures. Whereas the other measures consider persons with OCPD to be high in neuroticism, the MCMI-III conceptualizes OCPD as low in neuroticism and was designed not to assess state anxiety (Millon et al., 1996). In this sense, the MCMI-III conceptualization is actually more consistent with the existing criterion set that excludes symptoms of anxiousness or neuroticism (Widiger et al., 2002). This could reflect the fact that the MCMI-III was designed to distinguish the PDs from one another within a clinical sample (most of the other personality disorders are characterized by high neuroticism), whereas the other measures were perhaps more designed to distinguish the presence versus absence of each PD. As such, the findings for the MCMI-III might indicate that OCPD is lower in neuroticism and agreeableness, relative to the other PDs.
The clinicians' ratings of the individual items for each criterion are again useful complements to the FFM ratings in helping to understand the unique nature of the MCMI-III scale. While there is no infallible indicator of OCPD, the diagnostic criteria do provide a common, if imperfect, definition of the construct4. In addition, the instruments are typically used to provide assessments of DSM-IV OCPD and in most cases were explicitly constructed to do so. Using this metric, the MCMI-III includes items that were rated as assessing criterion 1 (“preoccupied with details, lists, orders”), criterion 3 (“workaholism”), and criterion 4 (“overconscientious about morals”). All of these appear to be heavily tied to the trait of conscientiousness and one could perhaps argue that these particular criteria have some potential adaptive consequences. Indeed, a positive response to MCMI-III item #137 (“I finish my work before taking time for leisure”), while clearly in the realm of workaholism, might prove adaptive within an occupational context. These findings may also help to explain previous research suggesting that the MCMI-III OCPD scale is associated with successful functioning rather than impairment (Craig, 1999).
The very strong relationship between the MCMI-III and conscientiousness is certainly consistent with previous research, yet it is important to note that this relationship is not unique to the MCMI-III. Saulsman and Page (2004, 2005) conducted a meta-analysis and reported that the mean weighted effect size for all OCPD measures with conscientiousness was .23. However, they further noted that the instrument significantly moderated the relationship. When the effect size was recalculated excluding MCMI-III studies, the relationship dropped to .03, leading them to suggest that the relationship between OCPD and conscientiousness could reflect an idiosyncrasy of the MCMI-III.
This interpretation by Saulsman and Page (2004) is understandable given the very weak convergence between the MCMI-III and other measures of OCPD found currently and in previous studies (Widiger & Boyd, 2009). However, it should be noted that none of the studies considered by Saulsman and Page (2004) included the SNAP, SCID-II-PQ, WISPI-IV, or OMNI; all of which correlated significantly with conscientiousness in the current study. If these measures had been included in prior FFM PD research, the results of the Saulsman and Page meta-analysis might have been different.
Additionally, of the 24 studies compiled by Widiger and Boyd (2009), only two included the SNAP, only one included the WISPI-IV, and none administered the OMNI (Widiger & Boyd did not include the SCID-II-PQ in their review). The findings of the current study suggest that it is precisely these instruments that correlate most highly with conscientiousness. As noted earlier, two of the three DSM-IV-TR defining features of OCPD concern facets of conscientiousness (i.e., orderliness and perfectionism). Thus, it appears that, perhaps ironically, the instruments that might best assess conscientiousness within OCPD are being used the least frequently. Minimally, the findings suggest that the SNAP, WISPI-IV, SCID-II-PQ, and OMNI should be more heavily considered as measures of OCPD in future research.
Nonetheless, even the relationships for these measures with conscientiousness (ranging from .17 to .31) are perhaps lower than one might expect given the theoretical connection between OCPD and conscientiousness (e.g., Widiger et al., 2002). The OCPD measures can be said to concern a maladaptive conscientiousness whereas the NEO PI-R is confined largely to an adaptive conscientiousness (although it does include one item concerning workaholism). Haigler and Widiger (2001) demonstrated that the correlation of NEO PI-R conscientiousness with OCPD increases substantially when its items are revised to assess maladaptive variants of conscientiousness.
There is reason to believe that the maladaptive traits of OCPD may be studied effectively within an undergraduate population. Traits such as workaholism and perfectionism may not be terribly uncommon within a collegiate setting. Grant and colleagues (2004) found that rates of OCPD were significantly higher for persons with at least some college education. Torgersen, Kringlen, and Cramer (2001) reported that within a large community sample, OCPD was the only PD that obtained a significant, positive relationship with education level. Blanco et al. (2008) reported that OCPD was the single most prevalent PD within the college population (8%). In addition, over 1,000 participants were screened to obtain persons with clinically significant OCPD symptomatology. Nevertheless, the relevance of the OCPD scales is typically understood in reference to psychiatric populations. This is particularly true for the MCMI-III, as its authors do not recommend using the instrument within non-clinical samples as they suggest it may measure only qualitative aspects of the pathology, rather than their severity (Millon et al., 1997). Thus, it would be of interest to determine whether comparable findings would be obtained within outpatient clinical samples where persons diagnosed with OCPD are being treated.
The instruments compared in the current study were exclusively self-report. Semi-structured interviews are often the preferred method for the assessment of PDs (McDermutt & Zimmerman, 2005; Rogers, 2001) and the lack of an interview in this study perhaps then limits the conclusions that can be drawn regarding validity. For instance, some might contend that self-report inventory scales cannot differentiate adaptive from maladaptive versions of personality traits. Future research that compares these self-report instruments to an interview assessment of OCPD would be helpful in extending the findings. On the other hand, it is also possible that any assessment instrument, even a semi-structured interview, could be prone to the same variations in conceptualization and assessment that were obtained for the self-report inventories. As such, it might be of interest for future studies to consider comparing alternative semi-structured interview assessments of OCPD with respect to their representation of traits of neuroticism, conscientiousness, and extraversion.
While there is no gold standard by which to judge the validity of self-report measures of OCPD, it might be useful for future research to compare the eight scales to additional external validators. The NEO PI-R does include two scales that relate closely to two of the three defining features of DSM-IV-TR OCPD (i.e., orderliness and perfectionism) but the NEO PI-R assertiveness facet would not serve as a strong marker for the third defining feature, mental and interpersonal control.
Finally, this study utilized specific scales, removed from larger inventories. We are unaware of any evidence to suggest responses are strongly context dependent (i.e., the response to a given item are affected by which item comes before and after it) and, in fact, computer adaptive tests such as the Graduate Record Examination regularly administer items in different orders depending on an individual's ability estimate. In addition, the PDQ-4 items are organized with respect to each personality disorder scale. Nonetheless, this is ultimately an empirical question and it is possible that when these items may perform differently when they are removed from their standard ordering. Additionally, while the responses were filtered for incomplete protocols, we did not remove any participates based on standardized validity scales.
This research was made possible, in part, by a fellowship (F31MH074245-02) awarded to the first author from the National Institute of Mental Health. Writing of this manuscript was supported by the Office of Academic Affiliations, Advanced Fellowship Program in Mental Illness Research and Treatment, Department of Veterans Affairs.
1These mean ratings for all of the items are available by contacting the first author
2In light of copyright issues, all items presented within the text are paraphrased using content from the item, rather than producing it verbatim.
3It should also be noted that the relatively low internal consistency of these scales might cap their relationship with one another. However, the instrument with the lowest internal consistency (the PDQ-4) did relate well with most of the other instruments. The PDQ-4 is one of the instruments that assesses all of the DSM-IV-TR diagnostic criteria. We suspect its low internal consistently is due in part to the heterogeneity of the diagnostic criterion set and the relatively fewer number of items.
4It should also be noted also that any potential departure from the DSM-IV-TR definition does not imply invalidity. In fact, such departures might also reflect choices on the part of an instrument's author to deviate from those parts of the definition which he/she does not agree with theoretically.
Douglas B. Samuel, Department of Psychiatry Yale University School of Medicine Department of Psychology University of Kentucky.
Thomas A. Widiger, Department of Psychology University of Kentucky.