A Comparison of Passive Aggressive and Negativistic Personality Disorders

Abstract

Passive aggressive personality disorder (PAPD) has historically played an important role in clinical theorizing and was diagnosable prior to the DSM-IV, in which the construct was relabeled ‘negativistic’, expanded to include negative affective symptoms, and appendicized. In this study we tested the hypothesis that the expansion of PAPD to include content related to negative moods and non-specific personality pathology compromised its discriminant validity. In an undergraduate sample (N = 1215), a self-report measure of PAPD was only moderately related to NEGPD and showed less diagnostic overlap with other PDs than NEGPD. Furthermore, a conjoint factor analysis yielded a strong first factor (moodiness) which appeared less specific to passive aggressive behavior than three other factors (irresponsibility, inadequacy and contempt). We conclude that future research on this potentially important clinical construct should focus on core passive aggressive features and abandon the negativistic content that has been added to it in successive editions of the DSM.

‘Passive aggressive’ was first used clinically during World War II to describe soldiers who refused to comply with officers’ demands (Millon, 1981). However the concept has been central to clinical theorizing, for instance in the form of anal or masochistic character types (Abraham, 1924; Fenichel, 1945; Reich, 1949), for nearly a century and has been featured in personality disorder (PD) models from a variety of contemporary theoretical orientations (Beck, Freeman, Davis, & Associates, 2003; Benjamin, 1993; Fine, Overholser, & Berkoff, 1992; Lazarus, 1971; McCrae, 1994; Millon, 1981; Stone, 1993). Clinicians and clinical theorists seem to continue to value the concept. For instance, Benjamin (1993) recommends routine assessment of passive aggressive features because of the potential for passive aggressive behavior to undermine the successful treatment of other primary (e.g., ‘Axis I’) conditions.

In the first edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM; American Psychiatric Association [APA], 1952), passive-aggressive personality disorder (PAPD) was conceptualized as having three types. Passive-dependent patients were helpless, indecisive, and clingy. This type eventually became dependent PD. The passive-aggressive type was pouty, stubborn, inefficient, and prone to procrastination and obstruction. The aggressive type was irritable, destructive, and resentful, with an underlying dependency thought to differentiate these behaviors from antisocial personality. Features of the latter two types were merged in the DSM-II (APA, 1968), in which PAPD symptoms included obstructionism, pouting, procrastination, intentional inefficiency, and stubbornness, each of which were thought to reflect hostility with the individual was unable to express openly.

The retention of PAPD in the DSM-III (APA, 1980) was controversial because some work group members believed that passive aggressive symptoms reflected a specific behavioral response to particular situations rather than a broad personality syndrome (Millon, 1981; Wetzler & Morey, 1999). It was retained, but unlike other DSM-III PDs, clinicians could only diagnose PAPD if the patient did not meet criteria for any other PD. Symptoms included resistance to demands for adequate social or occupational performance in the form of procrastination, dawdling, stubbornness, intentional inefficiency, and apparent forgetfulness. In the DSM-III-R (APA, 1987), PAPD criteria were expanded to include more negative emotional features such as sulking, irritability, and argumentativeness in addition to passive aggressive behaviors, and the exclusion criterion was dropped.

In DSM-IV (APA, 1994), the diagnostic criteria were expanded further, the disorder was renamed ‘negativistic’ (NEGPD), and it was appendicized. Symptoms included passive resistance to routine social or occupational tasks, complaints of being misunderstood, sullen argumentativeness, criticism and scorn of authority, envy and resentment of the relatively fortunate, exaggerated complaints of personal misfortune, and alternation between hostility and contrition. Millon (1993), who championed the decision to expand and rename PAPD as a member of the DSM-IV PD work group, offered several rationales, including that a) the disorder hadn’t received sufficient acceptance in the clinical literature, b) its content was too narrow and behavioral, c) passive aggressive behavior was too situational to reflect a syndrome, d) the term implies a particular motivation in a time when PD criteria were descriptive, thus diagnosis would require clinical inference, and e) PAPD overlapped too much with other disorders.

The proposed solution was to focus upon ‘negativistic attitudes’ thought to underlie passive aggressive behavior. The general argument of the current paper is that the increasing focus on negativism that characterized the PAPD, and eventually NEGPD, diagnosis over time significantly altered and decreased the specificity and clinical utility of the construct. It is ironic that this move, which may have spared PAPD from the diagnostic graveyard in the DSM-III-R and DSM-IV, may have led to its ultimate demise, as represented by indifference to the construct in most contemporary research and the DSM-5. Currently the prevailing mood is to cut PD diagnoses, and constructs that are not commonly researched and do not offer demonstrably unique clinical value are the first on the chopping block (Skodol et al., 2011).

Indeed, NEGPD may have contributed to the problems described by Millon (1993) at least as much as it has helped to solve them. Regarding acceptance in the clinical literature, Wetzler and Morey (1999) noted that passive aggressive behavior was as acceptable among researchers and clinicians as most other PDs during the time of DSM-III and DSM-III-R, but it has been studied much less often than most other PDs since that time. Although passive aggressive behavior is narrow and behavioral, it is understood in clinical theory to be driven by certain relatively stable personality dynamics (e.g., Beck et al., 2003; Benjamin, 1996; Kernberg, 1985; Stone, 1993). Thus, the critique that DSM-III-R symptoms were too narrow and behavioral may say more about the a-theoretical approach of the DSM-III than passive aggressive personality per se. The notion that pre-DSM-IV passive aggressive personality is too situational to reflect a syndrome conflicts with the relatively high prevalence rates of DSM-III-R PAPD (Morey, 1988), given that symptoms must be persistent to achieve the diagnosis. Furthermore, recent evidence that PD symptoms are generally less stable than was previously thought (Lenzenweger, Johnson, & Willett, 2004; Morey et al., 2007) make stability a questionable criterion for the validity of PD constructs. That the level of inference required to diagnose PAPD was seen as problematic says more about the descriptive focus of the DSM-III and -IV than the construct itself, and we would suggest that this issue is more effectively framed as a measurement question: can PAPD be reliably measured? Seen this way, mixing in content that is not directly related to the theoretical construct would likely make reliable measurement more challenging, not less (see Bradley, Shedler, & Westen, 2006 for an alternative approach to measuring PAPD).

In this paper, we focus primarily on Millon’s fifth reason for expanding PAPD: its overlap with other disorders. Diagnostic overlap, a serious problem with all DSM-III and -IV PDs, is driven primarily by the tendency for many PDs to share certain features, and in particular a tendency to experience negative emotions and to be interpersonally antagonistic (Samuel & Widiger, 2008). Thus adding criteria involving moodiness, antagonism, and feeling misunderstood to any disorder would tend to decrease its discriminant validity. We therefore suspect that adding negativistic content to PAPD lessened discriminant validity and decreased its unique potential to describe patients.

Folding content related to non-specific personality pathology in to the PAPD diagnosis also may have decreased the likelihood of identifying its core features by creating ‘noise’ in the PAPD ‘signal’. Research on the structure of NEGPD is ambiguous, with some covariance analyses suggesting multi-dimensionality (Rotenstein et al., 2007) and others suggesting a single dimension (Hopwood et al., 2009). However, no study yet has conducted a conjoint analysis of PAPD and NEGPD features. Given that criteria were both expanded in content and trimmed in number in the transition from PAPD to NEGPD, a conjoint analysis of PAPD/NEGPD criteria will permit the identification of more dimensions than an analysis of the criteria of only one operationalization, and will facilitate distinctions between dimensions that appear to be closer or more distal to the theoretical core of passive aggression.

Identifying underlying dimensions will also permit an examination of the correlates of the various features of PAPD/NEGPD. Recent research suggests that the disorder is influenced by both genes (Czajkowski et al., 2008) and an abusive environment (Grover et al., 2007; Hopwood et al., 2009), predictive of clinical outcomes such as presence of anxiety disorder (Johnson, Cohen, Kasen, & Brook, 2006; Rotenstein et al., 2007), suicidal behavior (Joiner & Rudd, 2002), and dysfunction (Hopwood et al., 2009), and linked to poor response to treatment of other disorders (Fricke et al., 2006). In terms of personality features, research suggests that PAPD is related to concerns about autonomy (Morse, Robins, & Gittes-Fox, 2002), and is associated with higher levels of neuroticism, manipulativeness and aggression and lower levels of agreeableness and conscientiousness (Hopwood et al., 2009). However these correlations do not speak to the potential for different underlying components of the construct to have varying relations with criterion variables. Furthermore, this research is unable to evaluate the degree to which criterion relations are a function of non-specific features that PAPD/NEGPD share with many other PDs as opposed to their being explained by specific aspects of passive aggressive personality.

In summary, despite the historical relevance of passive aggression to a number of influential clinical theories, empirical research on the construct is limited (Blashfield & Intoccia, 2000; Boschen & Warner, 2009) and the diagnosis has been abandoned by the DSM. We hypothesize that the saturation of the PAPD diagnosis with non-specific distress and interpersonal dysfunction that is common across PDs but not particular to passive aggressive behavior contributed to this trajectory. We also suspect that passive aggression is a unique and clinically important construct despite historical problems in operationalizing it, and believe that PAPD is worthy of further investigation. Thus we aim to begin the process of building an empirical basis for future decisions about the nosological status of passive aggressive personality by articulating what is and what is not uniquely PAPD. Specifically, our goal is to evaluate the impact of the negativistic content that has been added to it in the last two editions of the DSM, and to begin the process of separating this negativistic content from the PAPD.

We first correlated DSM-III-R PAPD and DSM-IV NEGPD with one another and with other DSM-IV PDs. We hypothesized that the emphasis of NEGPD on symptoms involving a general propensity to experience negative emotions reduced the syndrome’s discriminant validity, such that correlations with other PDs would be uniformly higher for NEGPD than for PAPD. We next evaluated the structure of passive aggressive personality pathology via conjoint factor analysis of PAPD and NEGPD symptoms. We then related the dimensions revealed by this analysis to a number of clinically relevant outcomes, including PDs, mood, attachment style, and interpersonal problems, to evaluate the differential relations among passive aggressive features. This conjoint structure and criterion associations were exploratory given the absence of comparable analyses in previous research. Nevertheless, we generally expected some dimensions to be more closely linked to theoretical conceptions of passive aggressive behavior, whereas others we expected to reflect non-specific distress and dysfunction (i.e., ‘negativism’).

Method

Participants

We sampled 1453 undergraduates who were compensated with course credit for participating in this IRB approved study. Of these, 1215 responded to at least 95% of all study items and had scores that were less than 2.5 standard deviations higher than a community mean on a scale measuring random responding (Morey’s [1991] Personality Assessment Inventory Infrequency scale), and were thus retained for further analyses. The average age was 19.02 (S.D. = 1.68); 644 (53.0%) were female, and 1042 (85.8%) were non-Hispanic Caucasian.

Measures

We administered the Personality Diagnostic Questionnaire-4+ (PDQ-4+; Hyler, 1994), a self-report questionnaire with a true-false item response format and items that correspond directly to DSM content, to measure NEGPD (α = .58, average corrected inter-item r = .30) and the PDQ-R (Hyler et al., 1990) to measure DSM-III-R PAPD (α = .60, inter-item r = .27). On average respondents reported having 1.80 (S.D. = 1.59) of 7 possible NEGPD symptoms and 2.26 (S.D. = 1.85) of 9 possible PAPD symptoms. The PDQ-4+ was also used to assess other DSM-IV PDs, which were treated as criterion variables. (Mdn α = .57, Mdn of average corrected inter-item correlations = .29). These somewhat low values may be due to factors such as the multidimensionality of some PDs, binary response format, and limitations of the DSM criteria in terms of content coverage. Importantly for the current study, low internal consistencies did not appear to be due to the potential for range restriction in student samples, as on average students had around 4 PAPD/NEGPD symptoms, variabilities were substantial, and skew tended to be modest (Mdn = .73). This is consistent with previous research that shows relatively higher rates of PD endorsement on questionnaire relative to interview assessments (Hyler et al., 1990) as well as heightened rates of PD psychopathology among young adults (Grant et al., 2004). It is also reassuring that internal consistency appears to have limited impacts on criterion-validity estimates (Schmitt, 1996; McCrae, Kurtz, Yamagata, & Terraciano, 2010).

We also administered three non-PD criterion measures. The Revised Experiences in Close Relationships scale (ECR-R; Fraley, Waller, & Brennan, 2000) is a 36-item measure of the anxiety (α = .75, inter-item r = .49) and avoidance (α = .76, inter-item r = .50) dimensions of attachment. The Inventory of Interpersonal Problems-Short Circumplex, (Soldz et al., 1995), a 32-item measure of interpersonal problems with a total score reflecting overall interpersonal distress (α = .88, inter-item r = .60) and indicators of difficulties related to agency and communion based on linear combinations of eight circumplex octant scores (Mdn α = .79). The trait version of the 20-item Positive Affect Negative Affect Scales (PANAS; Watson, Clark, & Tellegen, 1988) was used to measure dimensions reflecting propensities for negative (e.g., anger, sadness, shame; α = .88, inter-item r = .60) and positive (e.g., happiness, excitement, joy; α = .88, inter-item r = .62) emotions.

Analyses

We first correlated PAPD and NEGPD to one another and to other DSM-IV PDs to test their overlap and discriminant validity. We tested differences between dependent correlations using Steiger’s (1980) z-test at Type I error probabilities of .01 and .001. We next conducted a conjoint factor analysis on the tetrachoric correlation matrix of PAPD and NEGPD items in Mplus. Weighted least squares (mean and variance adjusted) estimation was applied to account for the non-normality of dichotomous items. We requested an oblique geomin rotation in order to achieve a desirable balance between factor complexity and interpretability (Sass & Schmitt, 2010). Scales were constructed based on the results of this analysis. These scales were then related to outcome criteria via bivariate correlations and multiple regression models to investigate the nomological net of passive aggressive features. Given the large number of analyses and relatively large sample size, we used Type I error rates of .001 and focused interpretively on effect sizes in this family of analyses.

Results

Correlations between PDs are listed in Table 1. The correlation between PAPD and NEGPD was somewhat modest (.52), suggesting that the change from PAPD to NEGPD was substantial. In fact, this value is significantly lower than the correlation between NEGPD and borderline PD (z = 3.86, p < .001), and is equal to correlations between NEGPD and paranoid and narcissistic PDs.

Table 1.

Correlations of DSM-III-R Passive Aggressive, DSM-IV Negativistic, and other DSM-IV Personality Disorders.

DSM-IV Criterion Disorder	DSM-III-R Passive Aggressive	DSM-IV Negativistic	Steiger’s Z
Negativistic	.52**	--
Paranoid	.38**	.52**	5.77**
Schizoid	.30**	.30**	0.00
Schizotypal	.38**	.45**	2.81*
Histrionic	.33**	.45**	4.75**
Narcissistic	.37**	.52**	6.16**
Borderline	.44**	.61**	7.51**
Antisocial	.37**	.41**	1.58
Avoidant	.36**	.43**	2.77*
Dependent	.40**	.46**	2.43*
Obsessive-Compulsive	.28**	.40**	4.62**

^*p < .01

^**p < .001

Our first hypothesis was that PAPD would demonstrate superior discriminant validity to NEGPD. Overall, 9 out of 10 correlations were larger for NEGPD than PAPD (binomial probability < .01), with eight of these differences being statistically significant at p < .01. On average, the NEGPD correlations were greater than the PAPD correlations by .09. To provide context for interpreting these correlations, we note that the average of the inter-correlations between all of the personality disorders was .41, and the standard deviation of these inter-correlations was .09. A majority of the correlates for PAPD were within one standard deviation of this mean, whereas we found that three correlations for NEGPD (with paranoid, narcissistic, and borderline PDs) were more than a standard deviation higher than the mean. Each of these coefficients is also more than .10 larger than the corresponding correlation for PAPD, a difference which is in each case statistically significant at p < .001.

We next conducted a conjoint factor analysis of PAPD and NEGPD symptoms. A four factor solution was selected as the first four factors appeared to be interpretable and had eigenvalues > 1. The fit of this model was excellent (RMSEA = .03, 90% CI = .025–.039; CFI = .98). Although the fifth factor also had an eigenvalue > 1 (1.04), one of its factors was almost completely influenced by a single item (NEGPD: “Others consider me moody or ‘hot-tempered’”) that also showed a sizeable loading (.40) on another factor. We consequently retained a four-factor solution.

Eigenvalues, geomin-rotated factor pattern coefficients, and factor correlations for this model are given in Table 2. The first factor had pattern coefficients greater than .40 on four NEGPD and three PAPD criteria involving negative emotions, resentment, and moodiness. Three PAPD and one NEGPD criteria that reflect irresponsibility, such as forgetting or procrastinating undesirable responsibilities and working slowly or poorly comprised the second factor. The third factor had sizeable loadings on three PAPD criteria which, when reverse scored, appear to reflect work-related insecurity, credit-seeking, and inadequacy. The fourth factor included two NEGPD and three PAPD criteria referring to feeling underappreciated, authority conflicts or need for approval at work, and contempt.

Table 2.

Results from Conjoint Factor Analysis of passive aggressive and negativistic personality disorder symptoms.

	1	2	3	4
Eigenvalues	4.88	2.16	1.44	1.11
Pattern Coefficients
Others have complained that I do not fulfill my obligations. (N)	.01	.62	.11	.05
I am not understood or appreciated. (N)	.27	.17	−.05	.36
People think I am moody and have a temper. (N)	.56	−.04	.21	.11
I do not like supervisors telling me how I should do my job. (N)	.14	.14	.03	.56
I resent people who are luckier than me. (N)	.50	.03	−.07	.25
I complain about my difficulties. (N)	.68	−.07	.08	.00
I am sometimes so nasty to others that I have to apologize later. (N)	.60	−.02	−.06	−.11
I procrastinate too much. (P)	−.03	.79	.05	.05
I get irritable with people rather than openly refusing to do something they want. (P)	.58	.21	.05	−.02
I screw things up for people whom I resent. (P)	.43	.44	−.05	−.01
I don’t like it when people want me to do things that are not my job. (P)	.54	.14	.03	.09
I am forgetful about things I do not like doing. (P)	.28	.54	−.03	−.06
People appreciate my hard work. (P) (R)	−.04	.03	.45	.40
I am grateful for suggestions about how I could improve my work performance. (P) (R)	.00	−.05	.56	.52
People see me as a worker who does my fair share (P) (R)	.02	.05	1.15	−.04
I could do a better job than my supervisors. (P)	.00	−.07	−.05	.53
Factor Correlations
2	.41
3	−.01	.18
4	.32	.22	.12	--

Note. All items are paraphrased from the PDQ, which is copywritten by Steven Hyler, and do not reflect actual item content. (N) indicates DSM-IV negativistic criteria, (P) indicates DSM-III-R passive aggressive criteria. (R) indicates item was reverse scored prior to analysis. The fifth eigenvalue = .99. The correlation matrix on which this analysis was conducted is available upon request. Note that it is possible with certain rotations, including Geomin, to observe pattern coefficients > 1, as is the case here for one coefficient on the third factor.

Scores were computed for each of these factors by summing corresponding items (i.e., those in bold in Table 2). Alphas for the factors were as follows: moodiness = .67 (corrected average inter-item r = .38), irresponsibility = .61 (inter-item r = .39), inadequacy = .57 (inter-item r = .39), contempt = .47 (inter-item r = .26). Some of these values are lower than might be desirable, likely because of problems related to the DSM and PDQ as discussed above as well as the limited saturation of these apparent sub-dimensions of passive aggressive personality pathology in this measure and corresponding low number of items for each factor. Again, it is noteworthy that internal consistency values in this range have a limited effect on criterion validity (Schmitt, 1996). Table 3 shows bivariate correlations between these factors and criterion variables, as well as R² values and β weights from multiple regression models in which the four conjoint PAPD/NEGPD scale scores predicted each criterion variable.

Table 3.

Bivariate and partial associations of passive aggressive/negativistic features with personality disorders, mood, attachment styles, and interpersonal problems.

		Moodiness		Irresponsibility		Inadequacy		Contempt
	R2	r	β	r	β	r	β	r	β
Personality Disorders
Paranoid	.31*	.49*	.42*	.27*	−.03	.11*	−.15*	.37*	.34*
Schizoid	.13*	.23*	.07	.24*	.15*	.16*	−.05	.31*	.28*
Schizotypal	.27*	.37*	.16*	.34*	.18*	.12*	−.20*	.40*	.43*
Antisocial	.19*	.34*	.16*	.38*	.26*	.12*	−.03	.26*	.16*
Borderline	.37*	.58*	.48*	.38*	.07	.13*	−.07	.34*	.22*
Histrionic	.21*	.45*	.39*	.29*	.06	.07	−.04	.20*	.10
Narcissistic	.26*	.45*	.35*	.29*	.04	.09	−.16*	.34*	.32*
Avoidant	.22*	.44*	.35*	.30*	.08	.12*	−.04	.27*	.17*
Dependent	.25*	.48*	.38*	.37*	.15*	.14*	.05	.23*	.05
Obsessive-Compulsive	.22*	.42*	.37*	.23*	−.01	−.04	−.24*	.21*	.25*
Mood
Negative Affect	.22*	.43*	.36*	.29*	.07	.19*	.08	.28*	.11
Positive Affect	.16*	−.24*	−.07	−.33*	−.28*	−.24*	−.20*	−.21*	.00
Attachment Styles
Avoidance	.07*	.10*	−.07	.23*	.23*	.16*	.06	.18*	.11
Anxiety	.14*	.37*	.36*	.18*	−.03	.10*	.01	.20*	.09
Interpersonal Problems
Total Problems	.23*	.45*	.36*	.32*	.08	.13*	−.02	.28*	.16*
Agentic Problems	.03*	.15*	.20*	.01	−.11	.05	.02	.08	.03
Communal Problems	.05*	−.01	.06	−.09	−.07	−.19*	−.10	−.19*	−.13*

Note. R² and β values are from regression models in which the four passive aggressive/negativistic scales predicted the criterion scale in the left-most column.

^*p < .001

Of particular relevance to study hypotheses, moodiness demonstrates the strongest overlap with other PDs. Moodiness also showed the strongest correlations to negative affectivity, anxious attachment, and generalized interpersonal problems. This pattern is consistent with the hypothesis that the broadening of the passive aggressive construct to include negativistic symptoms worsened discriminant validity by saturating PAPD with criteria involving a general tendency to experience distress and dysfunction. As such, the best clues as to what is unique about passive aggressive personality may lie in the composition and correlates of the other three dimensions.

Contempt and irresponsibility also showed consistent overlap with other PDs, although this overlap was of a lesser magnitude than that of moodiness, particularly for irresponsibility. Contempt was most strongly related to paranoid, schizotypal and narcissistic PDs. It was also associated with negative affectivity, coldness, avoidant attachment, and interpersonal distress. This pattern suggests that it shares with moodiness a propensity for negative affectivity, but it is differentiated based on interpersonal characteristics. Whereas moodiness connotes more generalized interpersonal problems, contempt is more specific to interpersonal coldness, insecurity, and mistrust. This pattern suggests that mistrustful, idiosyncratic, and self-focused rumination about others’ unjust or abusive behavior and consequent contempt may be a core aspect of passive aggressive personality.

Irresponsibility showed the strongest correlations with antisocial PD, and also had unique associations with low positive affect and attachment avoidance. This factor seems to involve an immature aspect of personality in which negative emotions or interpersonal disruptions characteristically lead to irresponsible behavior. The management of distress through ineffective interpersonal behavior and neglect of responsibilities is a central aspect of most theories of passive aggression.

Inadequacy had the weakest correlations with PDs, and in fact regression coefficients were mostly negative, suggesting perhaps a suppression effect in the regression models that indicates a very weak relation between inadequacy and generalized personality pathology. It also correlated negatively with positive affectivity. Overall, this pattern suggests that inadequacy connotes a depressive and self-defeating tendency to be somewhat disorganized and affectively dull. One might question the centrality of this factor to passive aggressive personality. Indeed, there are psychometric reasons to be skeptical. First, this pattern of weak correlations in general and negative correlations with PDs may simply reflect that the existence of the inadequacy factor is mostly due to the reverse scored items on the PDQ. Second, this factor is heavily influenced by a single item involving the degree to which others view the respondent as being a good worker. Thus it is possible that it relates to actual work performance or concern with impressions, independent of the potentially passive aggressive influences on that performance or concern.

Discussion

This study was designed to evaluate the overlap of DSM-III-R PAPD and DSM-IV NEGPD in order to determine the effects of broadening diagnostic criteria and towards identifying core dimensions of passive aggressive personality. Overall, our results suggest a) modest correspondence between PAPD and NEGPD, b) that the broadening of the construct to include more criteria involving generalized negative affectivity and dysfunction may have been detrimental in terms of discriminant validity, and c) that core dimensions of passive aggressive personality, at least in terms of DSM criteria, involve immature, irresponsible behavior, inner feelings of inadequacy and need for acknowledgment, and ruminative resentment about and contempt for authority figures.

Even though the DSM-III-R PAPD criteria included some ‘negativistic’ features, the modest correspondence between NEGPD and PAPD suggests substantial construct redefinition from DSM-III-R to DSM-IV. Of course this was intentional, and may have even served in the short term to save the diagnosis. However, it is also possible that this redefinition and the decision to place NEGPD in the DSM-IV appendix may have affected the lack of research on PAPD since the publication of DSM-IV despite the importance of passive aggressive behavior in several clinical theories. Indeed, although there has been some research on the construct as described above, most PAPD research has been the result of broader studies which assessed all PDs and in which PAPD was not the central construct of interest. There has been very little, if any, programmatic research on the construct. To the extent that passive aggressive personality features are clinically important and risk being lost altogether in future editions of the diagnostic manual, it may be important for researchers to focus more on historical conceptions of passive aggressive behavior than negativism.

This is particularly true because the over-sampling of symptoms related to negative affectivity and generalized interpersonal dysfunction in the criteria appears to have compromised the discriminant validity of the disorder. NEGPD had stronger correlations with PDs than PAPD across the board. The first factor from our conjoint analysis, moodiness, was comprised of NEGPD and PAPD items with negativistic content, and showed strong and consistent correlations with other PDs. This factor also had the strongest average correlation with other conjoint factors and correlated with generalized distress in the form of attachment anxiety, interpersonal problems, and negative affectivity. As such, it appears that the inclusion of these items significantly compromises the discriminant validity of passive aggressive personality, as the items do not clearly reflect passive aggression but instead indicate non-specific personality dysfunction. The fact that this was the first factor extracted and that more criteria showed strong pattern coefficients with it than any other factor demonstrate the extent to which PAPD was saturated with negativistic content, beginning in the DSM-III-R and continuing to a greater extent in DSM-IV. It is striking that 57% of NEGPD symptoms primarily involve moodiness. Future researchers interested in potentially unique aspects of passive aggressive personality should conceptualize the construct independent of this non-specific factor, and in general should focus on feelings, thoughts, and behaviors that are theoretically specific to passive aggression.

Based on this study, three such features may be worthy of further consideration. First, passive aggressive individuals may tend to engage in irresponsible or unproductive behavior as a way of expressing negative emotions or interpersonal grievances. Second, this behavior may relate to an inner insecurity regarding one’s value or worth, particularly with respect to authority figures. Third, passive aggressive people may ruminate about how others treat them unfairly or disrespect them, and develop a deep but perhaps unexpressed sense of resentment and contempt.

However, we caution that the degree to which the nature of passive aggressive personality can be understood based on the results of this study is constrained by the use of DSM criteria to operationalize the construct. We used DSM criteria because our primary purpose was to evaluate the impact of the change in conceptualization over the course of successive issues of the DSM, not to derive a canonical structure of PAPD moving forward. The DSM criteria are limited in a number of ways in terms of identifying the core features of passive aggressive personality. For instance, the lack of motivational content in the a-theoretical DSM renders pathological behaviors context-free. Simply having contempt or feeling inadequate or being irresponsible may or may not be sufficient to be labeled ‘passive aggressive’. Irresponsibility is only passive aggressive if it is meant to communicate a negative interpersonal message or expresses latent anger–irresponsibility can also be a function of attention problems, motivation, or disorganization associated with other forms of pathology. Similarly, many forms of personality pathology involve feelings of contempt (e.g., paranoid, antisocial) and inadequacy (e.g., dependent, avoidant). However the cognitive context of these feelings likely depends on the underlying personality. For instance, whereas the dependent person may feel inadequate due to the belief that they might lose the affection of close others if they do not do something to appease them, the passive aggressive person may feel inadequate due to their history and expectations of being humiliated. Thus the motivations to either maintain attachment (in the case of dependency) or retain autonomy and respect (in the case of passive aggression) contextualize the symptom, and without this context the meaning of the symptom is obfuscated. Future research should use measures of passive aggressive personality that incorporate motivational dynamics and the contents of cognitive attributions as a way of distinguishing the meaning of behaviors that are descriptive of but not uniquely passive aggressive. Notably, the DSM-III criteria were both behaviorally specific and implied a specific dynamic: PAPD was diagnosed if the person showed resistance to occupational and social performance through procrastination, dawdling, stubbornness, intentional inefficiency, or ‘forgetfulness’ (APA, 1980, p. 329). Perhaps the DSM-III would be a good place to restart.

Another issue with item content involves the importance of work. The term passive aggressive was conceived in an occupational (military) context, and many behavioral examples of and DSM criteria for passive aggression involve the failure to comply with job-related expectations. Two of the factors from this study, inadequacy and contempt, appear to be mostly specific to occupational situations. However, Millon (1993) and others have argued that, in order to be a PD, associated behaviors should generalize beyond a particular setting. For example, obsessive-compulsive personality is strongly associated with workaholism, perfectionism, and other occupation-relevant behaviors, but it is also associated with a particular interpersonal style (Pinto et al., in press) that is observable across settings. Future research should explore the degree to which passive aggressive behavior depends on the context, and in particular the extent to which it generalizes from work to other situations.

This study was limited by our use of questionnaires among undergraduates. This convenient method permitted the collection of data on a number of variables in a fairly large sample. Although there was sufficient variability in PD symptoms for covariance analyses and the large sample contributed to small confidence intervals, results may not generalize to other assessment methods or other kinds of samples. Replication in clinical samples would be informative. Although the PDQ corresponds to the DSM in terms of criterion content and number, it is possible that meanings shift slightly from the DSM to PDQ and thus a different structure and pattern of correlates would emerge from diagnostic interviews based precisely on the DSM. Additionally, self-reports and interviews of PD often show somewhat modest agreement. Future research should employ multiple methods across diverse samples to replicate and extend the current results.

There were additional issues with our measurement strategy. For instance, there may have been limited content coverage for some of the dimensions that underlie personality pathology, such as those identified here in the conjoint structure of PAPD and NEGPD. This likely contributed to, among other things, low scale internal consistencies. Future research should employ methods that go beyond DSM criteria to more adequately examine such dimensions. Another issue was the presence of a factor (3) with exclusively reverse-scored items. It is not clear how this factor might differ substantively from the others given this method artifact, particularly since some of its items have fairly direct positively-worded analogues that load on other factors. Finally, although the use of a number of external validating variables was a strength of this study, it would be beneficial for future research on the underlying structure of PAPD to include a wider array of criterion variables, particularly including other psychopathology constructs and measures of general personality functioning.

In conclusion, in the current study we evaluated the impact of expanding DSM PAPD criteria to include more negativistic content. Overall, this expansion appears to have compromised discriminant validity, perhaps contributing to the loss of potentially unique and clinically important aspects of passive aggressive behavior in the DSM-5. We acknowledge that studying a construct such as PAPD is against the current stream of diagnostic cutting in the DSM. We also recognize that the DSM-5 trait-specified proposal may permit an articulation of PAPD for interested clinicians and researchers, although we do not expect the dynamics of passive aggressive behavior to ever fully be captured by traits (see Wright et al., 2011). More specifically, we view the two traits suggested for depicting PAPD – depressivity and hostility (www.dsm5.org), to primarily indicate negativism as opposed to passive aggressive behavior. Evaluating the ability of proposed DSM-5 traits to account for passive aggressive and other forms of pathological behavior is an important area of ongoing research (Hopwood et al., in review).

Although there is insufficient research to recommend PAPD for inclusion in the diagnostic manual at this time, we believe that there is sufficient theoretical work and research to suggest that the construct is worthy of further investigation. A first step in this research involves identifying and focusing upon content that is unique to the diagnosis, a process we have initiated in this study. The capriciousness of the diagnostic manual does not bear upon the validity of any particular diagnostic construct. Thus PAPD’s absence from the DSM-5 should not impede research – however given the history of the DSM’s impact on PD research, it probably will. This is all the more reason to continue studying passive aggressive behavior, and perhaps the vacuum left by its abandonment in the DSM could spur novel conceptualizations.