Journal of Counseling Psychology Copyright 2000 by the American Psychological Association, Inc.

2000, Vol. 47, No. 2, 238-250 0022-0167AXVS5.00 DOI: 10.1037//0022-0167.47.2.238

Perfectionism, Attachment, and Adjustment

Kenneth G. Rice and Saied Ali Mirzadeh
Michigan State University

This research examined differences between types of perfectionists and whether perfectionism
related to attachment, academic integration, and depression. University students completed
the same attachment and perfectionism measures in two studies. In the 2nd study, measures of
academic integration and depression were also used. Replicated cluster analyses revealed 3
groups of perfectionists: adaptive, maladaptive, and nonperfectionists. Attachment predicted
type of perfectionist, with adaptive perfectionists reporting more secure attachments than did
maladaptive perfectionists. Adaptive perfectionists also had better academic integration than
maladaptive perfectionists. Maladaptive perfectionists, on average, reported depression in the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

clinically significant range. Results revealed academic and emotional benefits of adaptive
This document is copyrighted by the American Psychological Association or one of its allied publishers.

perfectionism, contrasted with the adverse emotional effects and no academic advantages of
maladaptive perfectionism.

Hamachek (1978) described two kinds of perfectionism. ism (Blatt, 1995; Pacht, 1984). Concerns about maladaptive
Normal or adaptive perfectionism involves adherence to perfectionism are well deserved. Maladaptive perfectionism
high self-standards, order, and organization but not at the has been correlated with numerous psychological problems
expense of self-esteem. Such perfectionists are free to be such as depression (Hewitt, Flett, & Ediger, 3996), anxiety
less precise and to experience positive feelings when a task (Alden, Bieling, & Wallace, 1994), and eating disorders
is completed. Neurotic or maladaptive perfectionism is (Cash & Szymanski, 1995; Minarik & Ahrens, 1996).
distinguished by high standards that never seem met or However, adaptive perfectionism, though less examined in
achievable. Completed projects are not enjoyed, and there is the literature, has been found to correlate positively with
considerable anxiety about imperfections. A number of efficacy (Frost, Marten, Lahart, & Rosenblate, 1990) and
independent programs of research with various measures of positive, but not negative, affect (Frost et al., 1993). Frost et
perfectionism have tended to support a two-dimensional, al. (1990) reported negative correlations between dimen-
higher order factor structure for perfectionism (Frost, Heim- sions of adaptive perfectionism and the frequency of procras-
berg, Holt, Mattia, & Neubauer, 1993; Rice, Ashby, & tination. Moreover, Rice et al. (1998) found a significant
Slaney, 1998; Slaney, Ashby, & Trippi, 1995; Terry-Short, though modest positive correlation between adaptive and
Owens, Slade, & Dewey, 1995). maladaptive perfectionism, although these constructs related
A multidimensional conceptualization of perfectionism in different ways with self-esteem and depression. Some
with adaptive and maladaptive aspects could appeal to studies have used empirical or conceptual classifications of
counseling psychology's historical emphasis on adjustment, perfectionists and nonperfectionists, thus permitting addi-
normal development, and individual strengths and re- tional analyses of group similarities and differences. For
sources. Unfortunately, the literature on perfectionism has example, Parker (1997) conducted a cluster-analytic study
tended to emphasize the adverse consequences or correlates of academically talented sixth graders and found that
of maladaptive perfectionism but has discounted or down- "dysfunctional" perfectionists tended to be anxious and
played the psychological advantages of adaptive perfection- disagreeable, whereas "healthy" perfectionists tended to be
conscientious, socially at ease, and achievement oriented (p.
555). Both groups of perfectionists had moderate to high
Kenneth G. Rice and Saied Ali Mirzadeh, Department of personal standards that, along with other characteristics,
Counseling, Educational Psychology, and Special Education, Michi- distinguished them from nonperfectionists. Thus, it may be
gan State University. the case that the dimensions of perfectionism are selective in
Preliminary results from this research were presented at the their prediction of certain psychological outcomes (e.g.,
105th Annual Convention of the American Psychological Associa- adaptive perfectionism may predict academic achievement
tion, Chicago, August 1997. This research was supported by a grant but not emotional adjustment). One purpose of the present
from the Spencer Foundation. We gratefully acknowledge the study was to examine such possibilities. Another purpose
assistance of Amanda Baden, Anna Crombach, Dave Derr, and
Ryan Smith in helping with several phases of the studies. We also was to explore empirically some presumed relational precur-
express appreciation to Jeffrey Ashby, Daniel Lapsley, and Robert sors of perfectionism.
Slaney for reading drafts of the article and providing us with
feedback. Several authors have suggested that perfectionism emerges
Correspondence concerning this article should be addressed to from differences in the quality of parent-child interactions
Kenneth G. Rice, Department of Counseling, Educational Psychol- (Hamachek, 1978; Sorotzkin, 1998). Parents of perfection-
ogy, and Special Education, 440 Erickson Hall, Michigan State ists have been characterized as harsh and critical, with
University, East Lansing, Michigan 48824-1034. Electronic mail unreasonably high standards for children to meet. According
may be sent to kgrl @ msu.edu. to Hamachek (1978), neurotic or maladaptive perfectionism

238
 PERFECTIONISM, ATTACHMENT, AND ADJUSTMENT 239
develops from children's need for acceptance from parents tion strategy seems to be 'Mom will stay with me if I don't
who hold high standards of accomplishment but are never raise any fuss' " (p. 289). Likewise, Harter (1998) summa-
satisfied with their children's striving toward these standards rized research in which insecurely attached children could
or are inconsistent with their approval. Adaptive or normal be described as having parents who were either underattuned
perfectionism presumably results from observing a parent (insensitive) or overattuned to their children's emotional
experiencing enjoyment and satisfaction from flexible adher- needs. Overattuned parents were described as intrusively
ence to high standards or from observing disorganized emphasizing what the child should feel as opposed to
parents. Though not specifically stated as such by Hama- understanding how the child does feel (emphasis in original;
chek, qualities of attachment to parents or parent-figures see also Stern, 1985). Thus, insecurely attached children
may set the stage for the development of different kinds of might describe themselves as being "perfect" and idealize
perfectionism. Indeed, the descriptive literature on attach- the self and others in order to "mask feelings of unworthi-
ment bonds and attachment styles is strikingly similar to the ness" (Harter, 1998, p. 566). They are likely to develop
literature on perfectionism. According to Bowlby (1969, negative (or false) views of themselves as unworthy and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1988) and others (e.g., Ainsworth, 1989; Sroufe, 1979), an either negative views of others (as uncaring) or positive
This document is copyrighted by the American Psychological Association or one of its allied publishers.

attachment denotes certain qualities of relationships in views of others (as perfect).

enduring emotional bonds, behaviors that serve to maintain
The conceptual literature on perfectionism has empha-
contact with significant other(s). Secure attachments emerge
sized the roles that parent-child relationships play in the
when parents have been emotionally accessible to, and
development of adaptive or maladaptive perfectionism.
nurturing of, their children. Secure attachments provide the
Therefore, in this article, the conceptual and operational
child with a sense of comfort and predictability, encouraging
understanding of attachment emphasizes the qualities of the
the child to approach rather than avoid new developmental
parent-child relationship, with measurement based largely
and interpersonal challenges. A secure bond has implications
on Bowlby's conceptualizations.
for affect regulation and subsequent interpersonal relation-
ships (Bowlby, 1969). Insecure attachments result from In a recent study. Rice, Ashby, and Preusser (1996)
unpredictable, harsh, or unsupportive parent-child interac- compared maladaptive (neurotic) and adaptive (normal) late
tions. Children with insecure attachment bonds have difficul- adolescent perfectionists who were classified into their
ties managing developmental challenges and experience a respective perfectionism group on the basis of median split
variety of personal and interpersonal adjustment problems cutoffs determined from two perfectionism subscales. They
(Bowlby, 1973). found that maladaptive perfectionists, compared with adap-
tive perfectionists, reported their parents to be significantly
Sorotzkin (1998) described parents of adaptive perfection- more critical and demanding. However, there were no
ists (children who strive for excellence) as positive, support- differences between groups on retrospective accounts of
ive, and encouraging; these children are likely to develop parental bonds (recalling parents as caring or overprotec-
realistic views of themselves as worthy and confident and tive). Other studies have found some concordance between
views of others as trustworthy and responsible. Harter perfectionism of parents and their children (Frost, Lahart, &
(1998) echoed similar sentiments in her review of the Rosenblate, 1991), and several investigations have found
literature on children's self-presentations. She observed that self-criticism to be associated with poor relationships with
children with secure attachment relationships "can accept parents (e.g., Brewin, Firth-Cozens, Furnham, & McManus,
both their desirable and undesirable features" and are able to 1992), but we could only locate one study (Rice et al., 1996)
describe themselves in positive terms but also can admit to that forthrightly examined the quality of late adolescent
being "less than perfect" (p. 566; see also Cassidy, 1988; parent-child bonds (albeit recollections of those bonds) and
Crittenden, 1994). their association with perfectionism. Minimal attention to
Using language similar to that used in describing attach- aspects of attachment bonds other than parental criticism is
ment, Sorotzkin (1998, pp. 87-89) noted that parents of surprising given the prominent role parent-child relation-
(maladaptive) perfectionists tend to be nonapproving or ships play in conceptualizations of perfectionism. One
inconsistently approving. They may display interest in their purpose of the present studies was to address that limitation
children as "objects" who are expected to perform as in the literature by focusing on current perceptions of
opposed to displaying interest in, and sensitivity to, the parent-child attachment relationships and their capacity to
subjective experiencing of the child (cf. Miller, 1996). predict perfectionism.
Children of such parents may, in turn, learn to emphasize the If attachment figures are available, responsive, predict-
importance of their performance over and above their able, and nurturing but not smothering, then the child is
emotional needs. Being perfect and discounting emotional likely to experience consistent standards for behavior, and
needs offer an opportunity to receive or maintain recognition meeting such standards is not a prerequisite for parental
from parents or to escape or avoid critical or punitive approval. It seems plausible, then, that secure parent-child
responses from parents. Cassidy (1994), in her description of attachment bonds should predict adaptive perfectionism.
insecurely attached children, noted an emergence of emo- Maladaptive perfectionism, on the other hand, may result in
tional regulation such that, when distressed, insecurely response to insecure parent-child attachment bonds signal-
attached children suppress negative emotions in order to ing an absence of parental support, inconsistency in respond-
maintain connection with the caregiver. As summarized by ing to the child's need for acceptance or approval, or the
Saarni, Mumme, and Campos (1998), the "emotional regula- parent's adherence to unreasonably rigid and critically
 240 RICE AND MERZADEH

imposed expectations for the child (Barrow & Moore, 1983; mistakes with failure; Personal Standards (seven items) taps high
Hamachek, 1978). Indeed, Blatt (1995) concluded that standards, goals, and high personal expectations for performance.
"highly perfectionistic, self-critical individuals . . . struggle Parental Expectations (five items) measures the belief that parents
had high expectations and set high standards for the respondent.
to meet harsh judgmental parental standards and are identi-
Parental Criticism (four items) taps the belief that parents were
fied with these attitudes of their parents, attitudes that they overly critical. Doubts About Actions (four items) measures the
now direct toward themselves so that whatever they accom- degree to which respondents doubt their ability to complete tasks
plish is never fully sufficient" (p. 1012). Hence, insecure effectively. The Organization subscale (six items) measures the
parent-child attachment bonds are likely to be present importance one places on order and organization.
among maladaptive perfectionists, who, in turn, experience Research using the MPS has demonstrated that, despite some of
significant psychological maladjustment the subscales consisting of only a few items, reliability estimates
We hypothesized that adaptive perfectionists would evi- (Cronbach's coefficient alphas) have been adequate and have
dence secure attachment bonds with parents, whereas mal- ranged from .77 to .93 (Frost et al, 1990). Concurrent validity for
adaptive perfectionists would evidence insecure attach- the MPS has been demonstrated in that the measure relates in
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ments. We also considered the possibility of curvilinear expected directions with other measures of perfectionism such as
This document is copyrighted by the American Psychological Association or one of its allied publishers.

associations between attachment and perfectionism to ad- Hewitt, Flett, Turnbull-Donovan, and Mikail's (1991) MPS and
Slaney et al.'s (1995) Almost Perfect Scale (Frost et al.. 1993; Rice
dress the issues of overattunement or underattunement in
et al., 1998). Criterion-related validity is evidenced by correlations
parent-child relationships described by Harter (1998). Thus, between MPS subscales and measures of psychological symptoms
it may be that too much (i.e., smothering) or too little (i.e., (e.g., Brief Symptom Inventory; Derogatis & Melisaratos, 1983)
insensitivity) of the qualities of secure attachment bonds and adjustments such as compulsiveness, self-esteem, procrastina-
may predict the different types of perfectionism. We ex- tion, and depression (Frost & Marten, 1990; Frost et al., 1990,
pected adaptive perfectionists would have high academic 1993). Of particular note for the purposes of this study has been the
integration and less depression, compared with maladaptive finding that Personal Standards and Organization correlated with
perfectionists who were expected to report emotional and positive affect but not negative affect and that factor analyses
academic difficulties. We examined these questions in two revealed these two subscales loaded onto an Adaptive Perfection-
separate studies to replicate and cross-validate our classifica- ism factor (labeled positive striving), whereas the remaining
tion procedure and our findings with regard to attachment. subscales appear to tap maladaptive perfectionism (labeled maladap-
tive evaluation concerns, Frost et al., 1993). Further evidence for
We also used a more diverse group of participants in the
the validity of the MPS is derived from a cluster analysis in which
second study to extend the generalizability of the results. three theoretically expected homogenous groups of perfectionists
were reliably identified as "healthy" perfectionists, "dysfunc-
Study 1 tional" perfectionists, and "nonperfectionists" (Parker, 1997).
Attachment. Attachment was assessed with the revised version
Method of the Inventory of Parent and Peer Attachment (IPPA; Armsden &
Greenberg, 1987), which measures the overall quality or security in
Participants the attachment bond with mother and with father. The revised IPPA
is called the Relationships Questionnaire (RQ) and contains two,
A total of 49 men and 129 women (A' ~ 179,1 with missing data 25-item, self-report subscales yielding separate scores for attach-
for gender) were recruited from elective undergraduate personal ment to mother and attachment to father (the original IPPA
adjustment courses at a large, public midwestem university. The contained 28 items measuring attachment to "parents" and did not
average age for this sample was 21.13 years (SD = 1.56). Ninety- specify mother or father). Participants are asked to consider their
four percent of the sample identified themselves as White, Euro- mother or father or the person who has acted most in the role of
pean American. Sixty-three percent indicated that their parents mother or father for them. (In the current study, 96% of the
were married and living together, 14% reported their parents had participants indicated that their responses on the RQ pertained to
divorced and not remarried, and 16% indicated one or both of their their relationship with their biological or adoptive mother, and 92%
parents had remarried. Fifty percent of the students indicated that responded by considering their relationship with their biological or
their home was less than 300 miles from campus. While attending adoptive father.) Participants respond to items along a 5-point
the university, the students tended Co live in sorority or fraternity Likert continuum ranging from 1 (almost true or always true) to 5
houses (44%) or off-campus dwellings (40%). Most of the partici- (almost never or never true). The IPPA and its revision, the RQ,
pants were seeking liberal arts degrees (58%). The frequencies of were designed to assess positive and negative perceptions of
other majors revealed that 13% of the participants were pursuing attachment relations. Positive aspects of attachment are operation-
degrees in consumer and family studies, 10% were in business, 7% alized by items that measure trust in the accessibility of parents and
were in education, 4% were in technology, 3% were in pharmacy, responsiveness of parents. Negative aspects include items that
and 3% were in science. measure anger or hopelessness resulting from unresponsive par-
ents. The RQ is scored such that higher scores represent more
Measures secure attachment.
Adequate reliability and validity for the IPPA have been reported
Perfectionism. Frost et al.'s (1990) Multidimensional Perfec- by Armsden and Greenberg (1987), who found Cronbach's coeffi-
tionism Scale (MPS) was used to assess adaptive and maladaptive cient alphas to range from .72 to .91. Test-retest correlations
perfectionism. The MPS consists of 35 items responded to with a ranged from .86 to .93 over a 3-week period (Armsden, McCauley,
Likert scale of 1 {disagree strongly) to 5 (agree strongly), with Greenberg, Burke, & Mitchell, 1990). The IPPA has correlated in
higher scores indicating greater perfectionism. The MPS contains expected directions with measures of family cohesion and parent
six subscales. The Concern Over Mistakes subscale (nine items) utilization (Armsden & Greenberg, 1987). Reliability and validity
assesses negative reactions to mistakes and a tendency to equate data for the RQ are not published, although some support exists for
 PERFECTIONISM, ATTACHMENT, AND ADJUSTMENT 241
the factor structure of the measure (M. T, Greenberg, personal (Hair et al., 1995, p. 442). Instead, it is recommended that
communication, May 30, 1993). Moreover, several authors have different emergent cluster sets be examined in light of
used the revised version of the measure (i.e., 25 items, scales for
"practical judgment, common sense, or theoretical founda-
mother and father) although they identified it as the IPPA. For
example, in other studies of college students, Schultheiss and tions*' (p. 443). However, there are some guidelines. For
BlusteJn (1994) and Benson, Harris, and Rogers (1992) reported example, a hierarchical cluster analysis produces an agglom-
psychometric and descriptive results for the IPPA comparable with eration schedule, and agglomeration coefficients (or within-
those found with the RQ in our study. cluster sum of squares) can be examined to suggest the
degree of change or heterogeneity in moving from one
Procedure cluster solution to another. Said differently, agglomeration
coefficients are analogous to eigenvalues examined in a
Participants completed a brief demographic form and then scree test in factor analysis; large increases in the coeffi-
completed the measures of perfectionism and attachment. These cients suggest that clusters have been identified that are very
measures were imbedded in a larger survey of instruments that was different from one another (more internally homogenous).
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

organized in random order to control sequencing effects. Measures The percentage of change in moving from one cluster
This document is copyrighted by the American Psychological Association or one of its allied publishers.

were completed within the first 2 months of the semester. Partici-

solution to another (based on the difference between agglom-
pants were volunteers who received research or extra course credit
for their participation. eration coefficients) can serve as an index of relative
increases in cluster homogeneity (Hair et al., 1995, p. 448).
Results Parker's (1997) strategy and other recommendations for
conducting cluster analyses (Hair et al., 1995) were followed
Preliminary Analyses in this study. The MPS subscale scores were standardized
Means, standard deviations, scale ranges, and reliability and initially submitted to a hierarchical cluster analysis.
estimates are summarized in Table 1. For comparative Ward's linkage method and the squared Euclidian distance
purposes, psychometric results for the second sample (Study measure were used. Three participants had missing data on
2) also appear in Table 1. Because limited psychometric data one or more of the variables and therefore were deleted from
exist for the RQ, our first step was to examine the summary the analyses. Changes in the agglomeration coefficient
or composite attachment scores* reliability estimates. Cron- suggested support for a five- or three-cluster solution. The
bach's coefficient alphas ranged from .74 to .96, with the percentage of change in agglomeration coefficients indicated
highest internal consistency estimates evidenced for the RQ a 16% change when moving to the three-cluster solution and
attachment indices (.95 to .96). Thus, it appeared that we a 27% change when moving to the five-cluster solution.
were measuring an internally coherent dimension of attach- However, in the five-cluster solution, two of the clusters had
ment by combining the 25-item RQ scale into single very small cell sizes compared with the others (/is = 10 and
attachment composites (Attachment-Mother and Attach- 16, respectively). The three-cluster solution yielded more
ment-Father). Listwise deletion of data resulted in a sample interpretable results and larger frequency distributions than
size of 173 with complete data on the perfectionism and the five-cluster solution. The three-cluster solution also was
attachment measures in Study 1. adopted because of a priori theoretical expectations for a
three-cluster grouping. Moreover, findings from other re-
searchers using the same measures and analyses have
Perfectionism Groups
yielded three clusters (Parker, 1997). Centroids (standard-
Cluster analysis was used to identify natural groupings of ized subscale means) from the three clusters were used as
perfectionists in the study sample. Cluster analysis is a initial seed or starting points in a subsequent fc-means
multivariate data analytic technique that is useful for identi- (nonhierarchical) cluster analysis. In this procedure, the final
fying "homogenous subtypes within a complex data s e t . . . cluster solution yielded 70 participants in the first cluster
to identify groups within a data set" (Borgen & Barnett, (40%, 21 men, 49 women), 76 participants in the second
1987, p. 460). Cluster analysis was selected over factor cluster (42%, 16 men, 60 women), and 30 participants in the
analysis because, although the techniques are somewhat third cluster (17%, 10 men, 20 women). There was not a
conceptually similar, cluster analysis yields groupings of significant difference in the gender distribution across these
objects or people, whereas factor analysis yields groupings clusters, x2(2, N = 176) = 2.22, p < .313. Average MPS
of variables (Borgen & Barnett, 1987; Hair, Anderson, subscale (nonstandardized) scores for these three clusters
Tathan, & Black, 1995). Clustering algorithms essentially appear in Table 2.
examine the proximity of various indicators or variable
scores to one another to determine groupings. The determi- One-way analyses of variance revealed significant be-
nation of adequate and relevant numbers of clusters is based tween-cluster differences on each of the MPS subscales (all
on changes in agglomeration coefficients and a priori ps < .0001). Results from Tukey pairwise comparisons (see
theoretical expectations. For instance, in this study, we Table 2) facilitated the labeling of the clusters. For example,
expected to find three groups or clusters of perfectionists participants in the first two clusters were not significantly
similar to those found in Parker's (1997) study using the different on concerns about making mistakes, parental
MPS. There is "no standard, objective, selection procedure" expectations and criticism, or doubts about their actions.
to determine the number of relevant clusters within a data set However, the second cluster revealed significantly higher
 242 RICE AND MIRZADEH

Table 1
Descriptive Statistics and Measurement Reliabilities
Measure Possible (actual) scale ranges Study 1 Study 2
Attachment
RQ Attachment-Mother 25-125(48-125)
M 97.62 97.09
SD 17.99 17.87
a .95 .96
RQ Attachment-Father 25-125 (28-125)
M 86.06 87.59
SD 21.22 20.28
a .96 .96
Perfectionism and Adjustment
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

MPS Concern Over Mistakes 9-45 (9-43)

M 22.58 22.39
SD 7.16 7.38
a .89 .89
MPS Personal Standards 7_35 (14-35)
M 24.54 25.54
SD 5.10 4.66
a .82 .79
MPS Parental Expectations 5-25 (5-25)
M 14.79 15.48
SD 4.10 4.15
a .80 .78
MPS Parental Criticism 4-20 (4-19)
M 8.46 8.71
SD 3.63 3.60
a .85 .82
MPS Doubts About Actions 4-20 (4-18)
M 10.20 10.25
SD 3.09 3.12
a .70 .70
MPS Organization 6-30(8-30)
M 23.87 24.58
SD 4.76 4.87
a .91 .93
Academic Integration Items 3-15(5-15)
M — 11.10
SD — 2.70
a — .74
CES-D 0-60 (0-54)
M — 15.63
SD — 10.28
a — .87
Note. RQ = Relationships Questionnaire; MPS = Multidimensional Perfectionism Scale;
CES-D - Center for Epidemiological Studies-Depression Scale. Participants in Study 1 did not
complete the academic integration items or the CES-D.

average scores for personal standards and organization than Attachment and Perfectionism
those of the first cluster. Consequently, descriptors of the
second cluster appeared to be theoretically consistent with Logistic regression was used to determine whether attach-
adaptive perfectionism, whereas the first cluster appeared to ment to parents predicted perfectionism. Attachment secu-
be composed of nonperfectionists. Participants in the third rity was expected to be evident among adaptive perfection-
cluster were significantly different from the other clusters on ists, whereas insecure attachment was expected among
each MPS subscale, and those differences revealed more maladaptive perfectionists. Curvilinear associations be-
concerns about making mistakes, higher personal standards, tween attachment and perfectionism also were tested to
greater experiences with parental expectations and criticism, examine the possibility that extremes of attachment (enmesh-
more doubts about their actions, and strong preferences for ment, disengagement) might also be important consider-
organization (higher than nonperfectionists but not as strong ations in identifying adaptive and maladaptive perfection-
as adaptive perfectionists). Thus, participants in the third ists. Because there were no hypotheses regarding attachment
cluster seemed best described as maladaptive perfectionists. and nonperfectionists that group was excluded from these
 PERFECTIONISM, ATTACHMENT, AND ADJUSTMENT 243
Table 2
Multidimensional Perfectionism Scale (MPS; Frost et ai, 1990) Means and Standard
Deviations by Subscale and Cluster Group (Study 1)
Adaptive Maladaptive
Nonperfectionists perfectionists perfectionists
(n = 70) ( 6 (n = 30)
MPS subscale M SD M SD M SD F(2,173)
Concern Over Mistakes 21.23. 5.77 2O.02a 5.14 31.90b 7.06 48.86
Personal Standards 20.33a 3.19 26.67b 4.02 28.97C 4.20 77.50
Parental Expectations 13.49a 2.91 13.62a 3.43 20.67b 2.62 65.54
Parental Criticism 7.90a 2.91 6.92a 2.36 13.70b 3.23 68.03
Doubts About Actions 9.90a 2.74 9.45a 3.13 12.80,, 2.44 15.32
Organization 20.278 3.92 26.95b 2.69 24.37C 4.75 62.56
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Note. All univariate F tests were significant at p < .0001. Values with different subscripts indicate
significant within-row differences between the clusters (Tukey post hoc comparisons, p < .05). For
example, on Concern Over Mistakes, adaptive and nonperfectionists were not significantly different
from one another, but both groups were significantly lower than maladaptive perfectionists.

analyses. The curvilinear attachment terms were calculated overall accuracy in classification (75%) and much better
by squaring the standardized attachment scores (Pedhazur, sensitivity than specificity. All of the adaptive perfectionists
1982). The linear and curvilinear variables were entered in (75 of 75) were correctly classified based on their Attach-
separate logistic regression analyses predicting the dummy- ment-Father scores. Only 7% of the maladaptive perfection-
coded adaptive (1) and maladaptive (0) perfectionism groups. ists (2 of 28) were correctly classified. The average Attach-
The overall multivariate model involving both the linear ment-Father score for adaptive perfectionists was 88.89
and curvilinear Attachment-Mother scores was significant, (SD = 20.09), and the average score for maladaptive perfec-
X2(2,tf = 105)= 19.17 1 p<.001.Thecurvilineartermwas tionists was 78.72 (SD = 24.41), d = 0.42.
not a significant predictor of perfectionism after accounting A third set of logistic regression analyses were conducted
for the linear attachment term, Ax 2 (l,W = 105) = 0.13,p > in which the linear Attachment-Father and Attachment-
.05. Without the curvilinear term in the model, coefficients Mother scores were entered in equations to examine the
and odds ratios (ORs) revealed that students with higher differential effects of attachment to one parent while control-
Attachment-Mother scores were more likely to be adaptive ling the effects of the other parent (order of entry was
perfectionists; the odds of being an adaptive perfectionist alternated for these tests, and the curvilinear terms were
were 2.62 times greater for a student with secure attachment excluded from the models). Attachment-Father did not
to mother than for a student with insecure attachment (95% account for significant effects in the model over and above
confidence interval [CI] for the OR was 1.62-4.25). Overall, the information gleaned from Attachment-Mother, Ax 2 (l.
Attachment-Mother scores correctly classified approxi- # = 103) = 0.33, p > .05. However, the converse was not
mately 78% of the perfectionists. However, the model true; Attachment-Mother added significantly to the predic-
evidenced better sensitivity than specificity. (Sensitivity tion model after controlling for Attachment-Father, Ax2(l»
refers to the rate of true positives or those who were N= 103) = 12.28, p < .001. The ORs for Attachment-
correctly predicted to be adaptive perfectionists; specificity Father and Attachment-Mother in the full model were 1.17
is the rate of true negatives or the rate of correctly predicting (95% CI = 0.69-1.98) and 2.40 (95% CI = 1.41-4.02),
maladaptive perfectionists.) Approximately 96% (73 of 76) respectively. Overall correct classification was 79%, and the
of the adaptive perfectionists were correctly classified when sensitivity (96%) was much better than the specificity (32%)
their Attachment-Mother scores were known. Only 31% (9 of the predicted classifications.
of 29) of the maladaptive perfectionists were correctly
classified in the prediction model. The average Attachment-
Mother score for adaptive perfectionists was 102.73 Discussion
(SD = 15.18), whereas the average score for maladaptive Cluster analysis procedures revealed groups of perfection-
perfectionists was 84.86 (SD = 21.31), d = 0.84. ists that were remarkably similar to those found by Parker
The overall model of Attachment-Father predicting perfec- (1997), who used similar procedures and measures but a
tionism also was significant, \ 2 (2, N = 103) = 6.06, p < much younger sample than the one used here. In both
.05. The curvilinear term did not add significantly to the investigations, adaptive ("healthy" according to Parker)
model, Ax2O> N= 103) = 1.52, p > .05. Without the perfectionists emerged who had high personal standards and
curvilinear term, the model revealed that greater security in preferences for organization. Likewise, a group of maladap-
the attachment relationship with father increased the likeli- tive ( "dysfunctional") perfectionists also was identified that
hood of being an adaptive perfectionist (OR = 1.60, 95% evidenced high standards and preferences for organization.
CI - 1.03-2.50). As with the analyses of attachment to However, maladaptive perfectionists also were character-
mother, this model yielded a generally good percentage of ized by excessive concerns about making mistakes and
 244 RICE AND NORZADEH

self-doubt, and they reported their parents to be exceedingly negative evaluative expectations helps some maladaptive
critical and to have quite high expectations of them. Indeed, perfectionists feel closer to their parents (cf. Saarni et al.,
there were important differences between the perfectionists 1998) or possibly contributes to defensive distortions of the
on a measure of general quality of attachment to parents. parent-child bond. In this way, consistent with their internal-
Maladaptive perfectionists reported insecure relationships, ized self-criticism, they perceive inadequacies within them-
whereas adaptive perfectionists reported secure attachments selves but not with their parents, or at least not with the
to parents. quality of relationships they have with their parents.
The relative importance of the mother-child bond in Although the preceding interpretations seem plausible, it
attachment-perfectionism associations was alluded to in an also should be recalled that an earlier study found no
earlier study of parents and college-age daughters conducted differences between perfectionists on retrospective accounts
by Frost et al. (1991). They found that mothers' but not of parental bonds (Rice et al., 1996). However, Rice et a),
fathers1 perfectionism was significantly associated with did not use the cluster analysis procedure described here and
perfectionism among daughters. Harshness on the part of may have identified groups of perfectionists that were not
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

both parents was predictive of daughters' perfectionism. comparable with those observed in this study. Alternatively,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Frost et al. speculated that daughters may be more affected it may be that differences between perfectionists on mea-
by their mother because of more opportunities to interact sures of attachment may be more discernible when examin-
with their mother than with their father, same-sex modeling ing concurrent perceptions of relationships rather than
effects, and other reasons. Frost et al. also pointed out that, retrospective perceptions. Likewise, although there were
as in the current study, correlational data limit causal differences on attachment between perfectionists in this
inferences regarding the impacts of relationships over time. study, there were only 30 maladaptive perfectionists in-
In contrast to the Frost et al. (1991) findings, the present volved in the analyses, revealing obvious concerns about
study found that attachment bonds reported by adaptive generalizability of these results. Finally, there were no
perfectionists were conspicuous by their high security or independent psychological adjustment indicators used in
quality, and security of attachment was a much better this study that might help clarify the relationship between
predictor of adaptive perfectionism than maladaptive perfec- perfectionism and attachment insofar as they have implica-
tionism. Adaptive perfectionism may emerge from unique tions for other aspects of functioning. Therefore, a second
qualities in attachment relationships that encourage the study was designed to (a) replicate the cluster analysis
development of adaptive perfectionism without detrimental groupings of perfectionists with a larger and more diverse
relational consequences for children when standards are not sample, (b) replicate the findings with regard to perfection-
met. Of course, the correlational design and self-report ism and attachment, and (c) extend the research by examin-
methods used in the present study severely restrict confi- ing the effects of perfectionism on depression and academic
dence in the longitudinal accuracy of the inferences regard- integration.
ing previous parent-child bonds and the development of
perfectionism. Nevertheless, the data do provide support that
perceptions of attachment to parents, especially attachment Study 2
to mothers whether accurate or not, discriminate perfection- Study 2 served as a replication and extension of Study 1.
ism groups. Replication and longitudinal extension of this One purpose of the second study was to demonstrate the
study will need to be undertaken to more completely explore generalizability of the findings from the first study, and
and support these inferences. thereby cross-validate those results by replicating the study
An implication of our findings, limitations and other with another sample. The initial aims for the second study
possibilities notwithstanding, is that the development of were the same as the first study: We expected to be able to
maladaptive perfectionism may emerge in a complex, parent- identify types of perfectionists (adaptive, maladaptive, and
child attachment relationship matrix. Perhaps, as Sorotzkin nonperfectionists) by clustering their scores on the MPS.
(1998), Miller (1996), and others have observed, parents of Adaptive perfectionists then were expected to evidence
maladaptively perfectionistic children may be less emotion- secure attachment bonds with parents, and maladaptive
ally attuned to their children and may be more concerned perfectionists were expected to have insecure attachments,
with the performance or accomplishments of their children. although our confidence in the latter expectation was
Moreover, these parents may impose harsh, critical sanc- diminished by results from Study 1. Linear and curvilinear
tions on their children for not measuring up to expectations, associations were examined again; on the basis of the first
although they may ignore accomplishments simply because study, curvilinear associations between attachment to par-
they are expected and therefore not worthy of special ents and perfectionism were not expected to emerge. In
acknowledgment. The child concurrently or, as suggested by addition, this study examined the association between
our results, subsequently internalizes those expectations perfectionism and two indicators of adjustment: academic
from others as expectations for the self, resulting in a integration and depression. These indicators were selected to
self-critical stance that could likely lead to psychological represent academic and emotional areas of functioning that,
disturbance at worse and limited personal growth at best. on the basis of previous research, would seem likely to be
However, this story seems incomplete given our relatively affected by perfectionism among a sample of university
poorer ability to predict maladaptive perfectionism with students (cf. Parker, 1997; Rice et al., 1996,1998). Adaptive
attachment. It may be that the intemalization of critical and perfectionists were expected to evidence better academic
 PERFECTIONISM, ATTACHMENT, AND ADJUSTMENT 245
integration than maladaptive perfectionists, and maladapti ve have performed academically as well as I anticipated I would at
perfectionists were expected to be significantly more de- [name of institution]," "I am satisfied with my course curriculum
pressed than adaptive perfectionists. at [institution],1' and "I am satisfied with my academic experi-
ence." Higher scores on these items indicated greater academic
integration. Although reliability coefficients have not been re-
Method ported, Cabrera et al. (1993) found the items produced adequate
factor loadings (.54 to .67) on an Academic Integration factor in
Participants their measurement model. This factor, in turn, related moderately to
the Grade Point Average factor and indirectly to the College
This sample consisted of students from the same midwestern Persistence factor.
university as the sample in Study 1. A total of 218 students (54 men Depression. Depression was measured with the Center for
and 164 women) were recruited from an initial random listing of Epidemiological Studies-Depression Scale (CES-D; Radloff, 1977).
students across the university. The average age of this sample was The CES-D is a 20-item, self-report measure designed to tap
21.05 years (SD = 3.87), with ages ranging from 18 to 44 years. depression in the general population. Respondents report the
Minority students were oversampled to ensure increased represen-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

frequency with which they experienced scale items during the

tation of minority students. Oversampling was necessary because
This document is copyrighted by the American Psychological Association or one of its allied publishers.

previous week, using a Likert-type scale, ranging from 0 (rarely or

the total enrollment of ethnic minorities at this institution was only none of the time) to 3 (most or all of the time). A total depression
10% of the student body at the time of this study. In the final score is derived from the measure that is based on summing all
sample, 63% of the participants identified themselves as White, items, with higher scores indicating greater depression. The
European American, 20% were Asian or Asian American, 8% were CES-D is not designed to diagnose depression and Santor and
Black or African American, 5% were Latino or Latina, 2% were Coyne (1997) specifically (and empirically) cautioned against the
Native American or American Indian, and 2% identified themselves use of arbitrary cutoff scores to identify cases of depression with
as multicultural mixed race or ethnicity. As expected, these the CES-D. For descriptive purposes, however, it may be useful to
proportions were higher than proportions within the general student note a number of studies that have indicated a cutoff score of 16 as
body (90% White, 4% Asian, 4% Black, 2% Latino or Latina, and optimal for distinguishing individuals who have significant
< 1 % Native American). Approximately 54% of the sample had depression from those who do not (Breslau, 1985; Weissman,
completed four or fewer semesters at the university. Most of the Sholomskas, Pottenger, Prusoff, & Locke, 1977).
students majored in liberal arts (20%) or engineering (20%), with
There is considerable psychometric support for the CES-D.
the remaining distribution of majors evidencing broad representa-
Single factor and second-order factor models for the CES-D have
tion of the university community (13% science; 9% health-related
been supported in exploratory and confirmatory factor analyses
professions; 7% each in management, child and family studies, and
(Radloff, 1977; Sheehan, Fifield, Reisine, & Tennen, 1995).
technology; and 5% each in pharmacy, agriculture, and education).
Cronbach's coefficient alpha for the total depression scale has
Approximately 75% of the sample indicated that their parents were
ranged from .89 to .90 in several studies with various populations
married and living together, 4% reported that their parents had
(Breslau, 1985; Radloff, 1977). The CES-D correlates in expected
divorced and not remarried, whereas 16% indicated that one or
directions with other measures and clinical interviews of depres-
both of their parents had remarried. Approximately 80% of the
sion and emotional adjustment (Weissman et al., 1977).
students indicated that their home was 300 miles or less from
campus. These students tended to live in on-campus residence halls
(58%) or off-campus apartments or houses (29%). Socioeconomic
status of the sample was determined with parent education and
Procedure
occupational prestige (Stevens & Hoisington, 1987). Approxi- The registrar provided two listings of random students at the
mately 49% percent of the mothers but only 2% of the fathers had university. Students in the first list were sent information about the
less than a high school education. Approximately 18% of the study and a consent form. Students who were interested in
mothers and 56% of the fathers had completed an undergraduate participating were scheduled to complete a packet of question-
degree or additional graduate or professional training. Sixty-six naires in small group settings. Those declining to participate were
percent of the mothers and 86% of the fathers worked full time replaced with another student selected at random from the
outside of the home. The average occupational prestige score for second list. Data were collected during the second and third months
mothers was 44.69 (SD = 15.63) and for fathers was 48.58 of the fall semester. Each participant was paid $5 for his or her
(SD = 17.53). These scores indicated that parents worked, on participation.
average, in technical, sales, administrative support, and managerial
occupations.
Results
Measures Preliminary Analyses
Attachment. Attachment was measured with the RQ as de- Means, standard deviations, scale ranges, and reliability
scribed in Study 1. estimates are summarized in Table 1. Cronbach's coefficient
Perfectionism. Frost et al.'s (1990) MPS, as described in Study alphas for this sample ranged from .70 to .96. Scale
1, was used to measure perfectionism. psychometric properties for the Study 2 sample were
Academic integration. Three items used by Cabrera, Nora, and comparable with those observed for the same measures in
Castafieda (1993) were used to measure dimensions of academic
integration and adjustment. These items were based on the work of Study 1. Nevertheless, data were collected at different times
researchers who investigated college student adjustment and reten- during the academic year for the two samples, and some
tion (Bean, 1982; Pascarella & Terenzini, 1979,1980). Participants demographic characteristics of the sample were different
responded to the items using a Likert-type scale ranging from 1 (i.e., racial composition) or not assessed with both samples
(not at alt like me) to 5 (very much like me). The items were: "I (i.e., socioeconomic status). Therefore, a preliminary analy-
 246 RICE AND MIRZADEH

sis was conducted to determine sample differences on the about making mistakes and fewer experiences with parental
two scales completed by both samples (MPS and RQ). criticism than evidenced by the third cluster. The average
Separate multivariate analyses of variance were conducted scores within clusters and the comparisons between clusters
for the two constructs (perfectionism and attachment), with suggested that the first cluster was composed of nonperfec-
sample as the independent factor and the multiple subscales tionists, the second cluster was adaptive perfectionists, and
from the MPS and RQ as the dependent variables. There the third cluster was maladaptive perfectionists.
were no significant differences between samples on the MPS
subscales, Wilks's X = 0.98, F(6, 387) = 1.43, p < .202. Attachment and Perfectionism
There also were no differences between samples on the RQ
Attachment-Mother and Attachment-Father subscales, The same logistic regression strategy used in Study 1 was
Wilks's \ = 0.99, F(6,384) = 0.48, p < .622. These results adopted for Study 2. As before, nonperfectionists were
suggested that despite some sample variations, the two excluded from these analyses. Attachment subscale scores
groups were comparable on the main constructs of interest. were standardized before calculating the curvilinear terms.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The perfectionism group was dummy coded such that 1 =

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Perfectionism Groups adaptive and 0 = maladaptive perfectionism.

The multivariate model involving linear and curvilinear
A statistical procedure similar to the one explained in Attachment-Mother scores was significant, x2(2,N = 155) =
Study 1 was used to identify clusters of perfectionists. 33.00, p < .001. The curvilinear term did not add signifi-
Rather than conduct another hierarchical cluster analysis, cantly to the model and was dropped from subsequent
the centroids from the nonhierarchical analysis used in the consideration, Ax 2 (l, N= 155) = 0.50, p > .05. Model
first study were used as seed points in a nonhierarchical statistics again revealed that students with higher Attachment-
cluster analysis for Study 2's MPS data, with the a priori Mother scores were more likely to be adaptive perfectionists
expectation that three clusters would emerge. The MPS (OR = 2.75, 95% CI - 1.84-4.12). Approximately 70% of
subscales were standardized prior to conducting this analy- the perfectionists were correctly classified by Attachment-
sis. Convergence on the final clusters was achieved after six Mother scores, with more comparable levels of specificity
iterations and yielded 63 participants in the first cluster and sensitivity evidenced in this sample than in Study 1.
(29%, 20 men, 43 women), 81 participants in the second About 78% (63 of 78) of the adaptive perfectionists and 62%
cluster (37%, 14 men, 67 women), and 74 participants in the (46 of 74) of the maladaptive perfectionists were correctly
third cluster (34%, 20 men, 54 women). There was not a classified in the model. The average Attachment-Mother
significant difference in the gender distribution across these score for adaptive perfectionists was 103.83 (SD = 14.86).
clusters, X2(2, N = 218) = 4.39, p < .111. Average MPS The average Attachment-Mother score for the maladaptive
subscale scores for the clusters appear in Table 3. perfectionists was 87.18 (SD = 19.71), d = 0.84. These
Significant differences between clusters emerged on each descriptive results were comparable with Study 1, and the
of the MPS subscales (all ps < .0001). Tukey pairwise effect size of the differences in Study 2 was identical to that
comparisons (see Table 3) revealed several similarities and observed in Study 1. In both studies, maladaptive perfection-
differences between clusters on the MPS subscales. For ists appeared to have more variable attachment scores.
example, the second and third clusters were comparable in The full model of linear and curvilinear Attachment-
their (high) scores on personal standards and organization, Father scores predicting perfectionist group was significant,
but the third cluster revealed significantly higher, on aver- Xz(2, # = 152) = 21.53, p < .001. In this analysis, the
age, concerns about making mistakes, parental expectations curvilinear term added significantly to the model after
and criticism, and doubts concerning behavior. The first and controlling for the linear term, Ax 2 d, N = 152) = 6.55, p <
second clusters were similar in their relatively low concerns .01. Model coefficients with both terms in the model

Table 3
Multidimensional Perfectionism Scale (MPS; Frost et al, 1990) Means and Standard
Deviations by Subscale and Cluster Group (Study 2)
Adaptive Maladaptive
Nonperfectionists perfectionists perfectionists
(n = 63) (n = 81) (n - 74)
MPS subscale M SD M SD M SD F(% 215)
Concern Over Mistakes 18.30a 4.42 19.20a 4.75 29.36* 6.76 92.79
Personal Standards 20.85a 3.45 27.5 l b 3.61 27.36,, 3.69 75.53
Parental Expectations 12.81a 2.90 14.48b 3.62 18.85C 3.30 62.42
Parental Criticism 7.32a 2.53 6.59a 2.02 12.20b 3.11 104.41
Doubts About Actions 10.06a 2.49 8.52b 2.46 12.31C 3.07 38.65
Organization 21.23, 3.98 27.59b 3.03 24.14C 5.19 42.44
Note. All uni variate F tests were significant at p < .0001. Values with different subscripts indicate
significant within-row differences between clusters (T\jkey post hoc comparisons, p < .05).
 PERFECTIONISM, ATTACHMENT, AND ADJUSTMENT 247
revealed that greater security in the attachment relationship than maladaptive perfectionists (M = 10.09, SD = 2.51),
with the father increased the probability of being an adaptive r(153) = -5.69, p < .001, d = 0.89. On the CES-D Scale,
perfectionist (OR = 2.44, 95% CI = 1.61-3.71). Overall maladaptive perfectionists (M = 20.54, SD = 10.08) had
classification was modest (63%), with comparable specific- significantly higher scores than adaptive perfectionists
ity and sensitivity emerging: Attachment-Father revealed a (M = 11.90, SD = 8.96), /(153) = 5.65,p < .001, d = 0.86.
68% correct classification of maladaptive perfectionists and It should be recalled that a typical cutoff score indicating
a 59% correct classification of adaptive perfectionists. The clinically significant depression is 16 on the CES-D. The
average Attachment-Father score for adaptive perfectionists average CES-D score for maladaptive perfectionists was
was 93.71 (SD = 20.55), and the average score for maladap- approximately one half a standard deviation above that
tive perfectionists was 80.97 (SD = 18.95), d = 0.62. cutoff and over three fourths of a standard deviation above
As noted, the curvilinear Attachment-Father term also the score for adaptive perfectionists.
was significant (OR = 1.44, 95% CI = 1.08-1.90), al-
though this effect was not significant in the first study.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Therefore, we cautiously explored this finding by examining Exploratory Analyses

This document is copyrighted by the American Psychological Association or one of its allied publishers.

the frequency distributions of perfectionists within different

groupings of Attachment-Father scores. One group con- A number of studies have observed that mood states,
sisted of students who were in the top third of the sample particularly negative mood states, can affect perceptions of
distribution of Attachment-Father scores (n = 67), the sec- relationships and individual functioning. For example, Lewin-
ond group was in the middle third (n — 68), and the third sohn and Rosenbaum (1987) found that in a longitudinal
group consisted of students whose Attachment-Father scores study of depression, "negative memories of parental behav-
were in the bottom third of the distribution (n = 70). The ior are characteristic only of currently depressed persons"
frequencies of adaptive (A) and maladaptive (M) perfection- (p. 617) and not of those who had been depressed in the past
ists in each of these groups were, respectively, 40 (A) and 10 (remitted) or who had never been depressed (for an excep-
(M) in the high Attachment-Father group, 17 (A) and 27 (M) tion, see Gotlib, Mount, Cordy, & Whiffen, 1988). There-
in the middle group, and 20 (A) and 32 (M) in the bottom fore, we designed exploratory logistic regression analyses to
group. Perfectionists were not proportionately distributed determine whether, after controlling for depressed mood,
across the three groups, x2(2, N = 146) = 23.92, p < .001. attachment would predict perfectionism. In a linear multiple
There was a disproportionate number of adaptive perfection- regression analysis, we also examined the prediction of
ists in the high Attachment-Father group (80%) but nearly academic integration by perfectionism after partialling the
identical proportions of adaptive and maladaptive perfection- effects of depression. In each analysis, the standard score for
ists in the other two groups (approximately 39% and 61%, depression was entered first, followed by the additional
respectively). These frequencies suggest that the modest predictor(s). In the logistic regression analysis, these addi-
curvilinear effect may be the result of a sharp increase in the tional predictors consisted of the attachment subscales (and
proportion of adaptive perfectionists in the highly secure the curvilinear terms) entered as a block predicting dummy-
Attachment-Father group. coded perfectionism groups. In the linear multiple regres-
We conducted additional logistic regression analyses in sion analysis, the additional predictor was the dummy-coded
which we entered both Attachment-Mother and Attachment- perfectionism score predicting academic integration. As in
Father scores (alternating order in separate equations) to previous analyses, the nonperfectionists were excluded from
examine the differential effects of attachment to one parent these analyses and the perfectionism groups were coded 1 =
while controlling the effects of attachment to the other adaptive and 0 = maladaptive.
parent. These analyses revealed that Attachment-Father did After controlling for depression, the attachment scores
not add significantly to the model after controlling for added significantly to the prediction of perfectionism, Ax2(4,
Attachment-Mother, Ax 2 (l, N = 152) = 2.77, p < .09. In N = 152) = 18.84, p < .001. The OR in this full model
contrast, Attachment-Mother added significantly to the indicated that being an adaptive perfectionist was 1.92 times
model after controlling for the effects of Attachment-Father, greater for those with higher security of attachment with
AxKhN= 152) = 19.04, p < .001. The ORs (and 95% their mothers and 1.55 times greater for those with higher
CIs) in the full model were 1.40 (0.94-2.09) for Attachment- security of attachment with their fathers. The curvilinear
Father and 2.38 (1.55-3.67) for Attachment-Mother. Over- terms were not significant in this model.
all correct classification was 70%, and the sensitivity (75%) In the linear multiple regression analysis, after partialling
and specificity (64%) of the predicted classifications were effects for depression, perfectionism accounted for addi-
comparable. tional and significant variation in the prediction of academic
integration (AR2 = .07), F(l, 152) = 13.75, p < .001. The
direction of effects indicated adaptive perfectionists had
Perfectionism, Academic Integration, and Depression better academic integration, or, alternatively, that maladap-
We conducted t tests to examine the mean differences tive perfectionists reported more academic integration diffi-
between adaptive and maladaptive perfectionists on the culties (B = 1.52, SE B = .41, p = .28, p < .001). There-
measures of academic integration and depression. Adaptive fore, with the exception of one previous curvilinear finding,
perfectionists exhibited, on average, significantly higher the effects of attachment in the prediction of perfectionism
scores on academic integration (M = 12.32, SD = 2.36) and the effects of perfectionism in the prediction of aca-
 248 RICE AND MIRZADEH

demic integration did not appear to be significantly altered tive perfectionist group. Future studies may want to examine
by depressed mood. more closely the possibility of perceptual distortions and
other defenses activated within the self-systems of maladap-
tive perfectionists.
General Discussion Attachment to mother may play a different and perhaps
The results of these studies are meaningful on conceptual, more prominent role, when compared with attachment to
empirical, and clinical fronts in the study of perfectionism. father, in the development of perfectionism. Unfortunately,
We were able to discern three theoretically meaningful our assessment of attachment was restricted to a general
clusters of perfectionists on the basis of subscale score dimension of perceived security in relationships; therefore,
groupings of Frost et al.'s (1990) MPS. Similar to Parker's finer grained analyses of the differential impact of mother
(1997) study of sixth graders, the university students who and father on the development of perfectionism were not
participated in this study could be meaningfully grouped possible. Future research may consider parenting roles,
into adaptive, maladaptive, and nonperfectionist clusters. degree and type of contact between parents and perfectionis-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Adaptive perfectionists, compared with the other groups, tic children, and the development of affective as well as
This document is copyrighted by the American Psychological Association or one of its allied publishers.

evidenced high preferences for order and organization. They cognitive aspects of perfectionism for further clues regard-
tended to endorse setting high personal standards at about ing the attachment-to-parent differences observed in this
the same degree as maladaptive perfectionists, and both of study. Moreover, the design of the current studies and the
these perfectionist groups endorsed the importance of high reliance on students* self-report data did not permit a more
standards more so than nonperfectionists. Maladaptive per- forthright examination of causal or longitudinal associations
fectionists were distinguished from the other groups by their or consideration of different relational perspectives, so
excessive concerns about making mistakes, their strong inferences must be restricted to concurrent consideration of
self-doubts, and their experiences of parents as being critical perfectionism-attachment associations. Future research may
and expecting much of them. want to incorporate relationship perspectives from parents,
Relational correlates with perfectionism have been mini- in addition to students' reports of perfectionism, to examine
mally examined in the literature. Replicated findings in the the relative impact of self- versus other-perceptions in the
current studies revealed that perceptions of secure attach- prediction of perfectionism.
ment to either parent increased the likelihood of being an Maladaptive perfectionists were revealed to be less aca-
adaptive perfectionist, although the effects of attachment to demically integrated and more depressed than adaptive
mother appeared stronger than those of attachment to father. perfectionists were. Perhaps most alarming was that the
However, more comparable odds of being an adaptive average degree of depression among maladaptive perfection-
perfectionist were observed after partialling the effects that ists exceeded a typical cutoff for clinically significant
depression might have on perceptions of attachment. Also, a depression. The strong and consistent association between
curvilinear association between attachment to father and maladaptive perfectionism and depression supports Blatt's
perfectionism decreased to nonsignificance when control- (1995) and Pacht's (1984) emphasis on the "destructiveness
ling the effects of depression (and this curvilinear effect was of perfectionism." Taken with the results regarding aca-
not replicated). Associations between attachment and perfec- demic integration, it seems that maladaptive perfectionists,
tionism must be cautiously interpreted given the correla- despite high standards for excellence, fare poorly on aca-
tional and self-report nature of these data, and with some demic and emotional indicators at the university. Adaptive
results (i.e., curvilinear findings, specificity in predictions) perfectionists, who also endorse high personal standards,
not emerging consistently across studies. It may be that report satisfaction with their courses and academic tasks,
maladaptive perfectionism develops in response to parents and confidence in their academic performance. Such results
who expect much from their children, are excessively suggest that adaptive perfectionism may play a role in
critical of their children, and who may provide little facilitating the academic adjustment and integration of
emotional support and inconsistent responsiveness as their college students, without the emotional or depressogenic
children attempt to meet such demands. It may be the case costs apparent among maladaptive perfectionists. Thus,
that the frequently found association between insecure adaptive perfectionists appear decidedly advantaged in their
attachment and maladajustment in this population (cf. academic pursuits, whereas maladaptive perfectionists ap-
Kenny & Rice, 1995) is in part the result of maladaptive pear decidedly disadvantage^ What is not clear from these
perfectionism. For example, problematic attachment to studies is the degree to which the effects of perfectionism are
parents may set the stage for internalizing harsh, self- domain-specific or more generalized. For example, we
defeating expectations of self and others that place the child observed effects of perfectionism on academic integration
at risk for developing later emotional as well as academic but not on other aspects of college adjustment (e.g., grade
difficulties. Indeed, this general possibility has been consid- point average, social activities, and interpersonal relation-
ered in several theoretical models of counseling and develop- ships). Likewise, we found a relationship between perfection-
ment (Beck, Rush, Shaw, & Emery, 1979; Bowiby, 1973; ism and depression but were not able to examine various
Kohut & Wolf, 1978). However, as noted throughout our subtypes of depression or different components of depres-
two studies, the quality of attachment is an imperfect sion. Future research could explore these within-construct
predictor of maladaptive perfectionism, with some quite variations.
securely attached students categorized within the maladap- If the patterns of findings here are consistent in other
 PERFECTIONISM, ATTACHMENT, AND ADJUSTMENT 249
research with client samples, then counselors working with of perfectionism, or does it help clients learn how to manage
clients who present with academic problems or depression relatively stable perfectionistic tendencies?). The search for
would be advised to carefully assess the typology of the origins of adaptive perfectionism could be broadened,
perfectionism rather than consider perfectionism from a and such a search may find useful means for encouraging the
unidimensional perspective. Care should be taken with development of adaptive perfectionism and discouraging
interventions directed at perfectionists. For example, from a maladaptive perfectionism.
psychoeducational perspective, it seems reasonable to help
clients learn about perfectionism and its roles in various
adjustment outcomes. It also seems reasonable to help References
clients strengthen the adaptiveness of perfectionism (e.g., Ainsworth, M. D. S. (1989). Attachments beyond infancy. Ameri-
standards, organization), while challenging the usefulness of can Psychologist, 44, 709-716.
maladaptive perfectionism (e.g., self-doubt, self-criticism). Alden, L. E., Bieling, P. J., & Wallace, S. T. (1994). Perfectionism
Theoretical and some empirical precedents suggest that in an interpersonal context: A self-regulation analysis of dyspho-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

maladaptive perfectionism may be more intractable than ria and social anxiety. Cognitive Therapy and Research, 18,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

adaptive perfectionism, but future research should examine 297-316.

whether perfectionists can actually change as a result of Annsden, G. C , & Greenberg, M. T. (1987). The inventory of
therapy or other interventions. Indeed, attachment theory parent and peer attachment: Individual differences and then-
may prove to be an important component of the conceptual relationship to psychological well-being in adolescence. Journal
and strategic arsenal the therapist uses in working with of Youth and Adolescence, 16, 427-453.
Armsden, G. C , McCauley, E., Greenberg, M. T., Burke, P. M.? &
perfectionistic clients (Bowlby, 1988; Pistole, 1989). One
Mitchell, J. R. (1990). Parent and peer attachment in early
reason maladaptive perfectionism has been rather resistant adolescent depression. Journal of Abnormal Child Psychology,
to change (Blatt, Zuroff, Quinlan, & Pilkonis, 1996) may be 18, 683-697.
because changing perfectionism requires fundamental Barrow, J. C , & Moore, C. A. (1983). Group interventions with
changes in self- and other-perceptions that, in turn, require perfectionistic thinking. Personnel and Guidance Journal, 61,
fundamental changes in relational experiences. Unfortu- 612-615.
nately, our measure of attachment to parents was imprecise Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles
regarding insecurity of attachment. Thus, it may be that among young adults: A test of a four-category model. Journal of
some types of insecure attachment are more relevant to the Personality and Social Psychology, 61, 226-244.
prediction of perfectionism and to intervening with perfec- Bean, J. P. (1982). Student attrition, intentions, and confidence:
tionistic clients than other types. Future research could Interaction effects in a path model. Research in Higher Educa-
tion, 17, 291-319.
attend to other conceptualizations of attachment, such as
Beck, A., Rush, A., Shaw, B., & Emery, G. (1979). Cognitive
adult attachment styles (Bartholomew & Horowitz, 1991), therapy of depression. New York: Guilford Press.
in efforts to assess more finely the distinctions between types Benson, M. J., Harris, P. B., & Rogers, C. S. (1992). Identity
of perfectionists and types of insecure attachment styles, and consequences of attachment to mothers and fathers among late
how those attachment-perfectionism typologies may play a adolescents. Journal of Research on Adolescence, 2, 187-204.
role in therapeutic processes and outcomes. Blatt, S. J. (1995). The destructiveness of perfectionism: Implica-
tions for the treatment of depression. American Psychologist, 50,
To summarize, the literature on perfectionism historically 1003-1020.
has focused on maladaptive or problematic perfectionism. Blatt, S. J., Zuroff, D. C , Quinlan, D. M., & Pilkonis, P. A. (1996).
Indeed, this type of perfectionism appears to place individu- Interpersonal factors in brief treatment of depression: Further
als at considerable risk for a host of adjustment difficulties. analyses of the National Institute of Mental Health Treatment of
Ourfindingsrepresent a growing literature aimed at articulat- Depression Collaborative Research Program. Journal of Consult-
ing empirically the relational correlates of perfectionism and ing and Clinical Psychology, 64, 162-171.
the multidimensional nature of perfectionism. We were Borgen, F. H., & Barnett, D. C (1987). Applying cluster analysis in
particularly interested in the adaptiveness of perfectionism. counseling psychology research. Journal of Counseling Psychol-
ogy, 34, 456-468.
Our results are limited by the use of self-report measures, Bowlby, J. (1969). Attachment and loss, Vol. 1: Attachment. New
samples composed of university students in the Midwest, York: Basic Books.
and imprecise indicators of insecure attachment. These Bowlby, J. (1973). Attachment and loss, Vol. 2: Separation.
limitations notwithstanding, the results indicate that adap- London: Hogarth Press.
tive perfectionists (distinguished mostly by high personal Bowlby, J. (1988). A secure base. New York: Basic Books.
standards and organization but minimal anguish about Breslau, N. (1985). Depressive symptoms, major depression, and
meeting those standards) appear to have particularly advan- generalized anxiety: A comparison of self-reports on CES-D and
tageous internal resources for managing college-related results from diagnostic interviews. Psychiatry Research, 15,
challenges. Adaptive perfectionists have good academic 219-229.
integration and, compared with maladaptive perfectionists, Brewin, C. R., Firth-Cozens, J., Furnham, A., & McManus, C.
very little emotional distress. Adaptive perfectionism may (1992). Self-criticism in adulthood and recalled childhood
experiences. Journal of Abnormal Psychology, 101, 561-566.
emerge in the context of secure attachment relations students
Cabrera, A. F., Nora, A., & Castafieda, M. B. (1993). Structural
have with parent figures. Future research could address equations modeling test of an integrated model of student
additional counseling implications of the present investiga- retention. Journal of Higher Education, 64, 123-139.
tion (e.g., does counseling alter perfectionism, certain kinds Cash, T. F., & Szymanski, M. L. (1995). The development and
 250 RICE AND MIRZADEH

validation of the Body-Image Ideals Questionnaire. Journal of Pacht, A. R. (1984). Reflections on perfection. American Psycholo-
Personality Assessment, 64, 466-477. gist, 39, 386-390.
Cassidy, J. (1988). Child-mother attachment and the self at age 6. Parker, W. D. (1997). An empirical typology of perfectionism in
Child Development, 57, 331-337. academically talented children. American Educational Research
Cassidy, J. (1994). Emotion regulation: Influences of attachment Journal, 34, 545-562.
relationships. In N. Fox (Ed.), Biological and behavioral founda- Pascarella, E. T., & Terenzini, P. T. (1979). Interaction effects in
tions of emotion regulation. Monographs of the Society for Spady's and Tinto's conceptual models of college dropout.
Research in Child Development, 59 (2/3, Serial No. 240). Sociology of Education, 52t 197-210.
Crittenden, P. M. (1994). Peering into the black box: An explor- Pascarella, E. T., & Terenzini, P. T. (1980). Predicting freshman
atory treatise on the development of self in young children. In D. persistence and voluntary dropout decisions from a theoretical
Cicchetti & S. L. Toth (Eds.), Rochester Symposium on Develop- model. Journal of Higher Education, 51, 60-75.
mental Psychopathology: Disorders and dysfunctions of the self Pedhazur, E. J. (1982). Multiple regression in behavioral research
(%L 5, pp. 79-148). Rochester, NY: University of Rochester New York: CBS College Publishing.
Press. Pistole, M. C. (1989). Attachment: Implications for counselors.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Derogatis, L. R., & Melisaratos, N. (1983). The Brief Symptom Journal of Counseling & Development, 68, 190-193.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Inventory: An introductory report. Psychological Medicine, 13, Radloff, L. S. (1977). The CES-D Scale: A self-report depression
595-605. scale for research in the general population. Applied Psychologi-
Frost, R. O., Heimberg, R. G., Holt, C. S., Mattia, J. I., & Neubauer, cal Measurement, I, 385-401.
A. L. (1993). A comparison of two measures of perfectionism. Rice, K. G., Ashby, J. S., & Preusser, K. J. (1996). Perfectionism,
Personality and Individual Differences, 14, 119-126. relationships with parents, and self-esteem. Individual Psychol-
Frost, R. O., Lahart, C. M., & Rosenblate, R. (1991). The ogy, 52, 246-260.
development of perfectionism: A study of daughters and their Rice, K. G., Ashby, J. S., & Slaney, R. (1998). Self-esteem as a
parents. Cognitive Therapy and Research, 15, 469-489. mediator between perfectionism and depression: A structural
Frost, R. O., & Marten, P. A. (1990). Perfectionism and evaluative equations analysis. Journal of Counseling Psychology, 45,
threat. Cognitive Therapy and Research, 14, 559-572. 304-314.
Frost, R. O., Marten, P., Lahart, C , & Rosenblate, R. (1990). The Saarni, C , Mumme, D. L., & Campos, J. J. (1998). Emotional
dimensions of perfectionism. Cognitive Therapy and Research, development: Action, communication, and understanding. In W.
I4t 449-468. Damon & N. Eisenberg (Eds.), Handbook of child psychology:
GoUib, I. H., Mount, J. H., Cordy, N. I., & Whiffen, V. E. (1988). Vol. 3. (pp. 237-309). New York: Wiley.
Depression and perceptions of early parenting: A longitudinal Santor, D. A., & Coyne, J. C. (1997). Shortening the CES-D to
investigation. British Journal of Psychiatry, 152, TA-ll. improve its ability to detect cases of depression. Psychological
Hair, J. F., Anderson, R. E., Tkthan, R. L., & Black, W. C. (1995). Assessment, 9, 233-243.
Multivariate data analysis with readings (4th ed.). Englewood Schultheiss, D. E. P., & Blustein, D. L. (1994). Role of adolescent-
Cliffs, NJ: Prentice Hall. parent relationships in college student development and adjust-
Hamachek, D. E. (1978). Psychodynamics of normal and neurotic ment. Journal of Counseling Psychology, 41, 248-255.
perfectionism. Psychology, 15, 27-33. Sheehan, T. J., Fifield, J., Reisine, S., & Tennen, H. (1995). The
Hatter, S. (1998). The development of self-representations. In W. measurement structure of the Center for Epidemiologic Studies
Damon & N. Eisenberg (Eds.), Handbook of child psychology: Depression Scale. Journal of Personality Assessment, 64, 507-
Vol. 3. (pp. 553-617). New York: Wiley. 521.
Hewitt, P. L., Flett, G. L., & Ediger, E. (1996). Perfectionism and Slaney, R. B., Ashby, J. S., & Trippi, J. (1995). Perfectionism: Its
depression: Longitudinal assessment of a specific vulnerability measurement and career relevance. Journal of Career Assess-
hypothesis. Journal ofAbnormal Psychology, 105, 276-280. ment, 3, 279-297.
Hewitt, P. L., Flett, G. L., Tumbull-Donovan, W, & Mikail, S. F. Sorotzkin, B. (1998). Understanding and treating perfectionism in
(1991). The Multidimensional Perfectionism Scale: Reliability, religious adolescents. Psychotherapy, 35, 87-95.
validity, and psychometric properties in psychiatric samples. Sroufe, L. A. (1979). The coherence of individual development:
Psychological Assessment, 3, 464-468. Early care, attachment, and subsequent developmental issues.
Kenny, M. E., & Rice, K. G. (1995). Attachment to parents in late American Psychologist, 34, 834-841.
adolescent college students: Current status, applications, and Stern, D. N. (1985). The interpersonal world of the infant. New
future considerations. The Counseling Psychologist, 23, 433- York: Basic Books.
456. Stevens, G., & Hoisington, E. (1987). Occupational prestige and
Kohut, H., & Wolf, E. S. (1978). The disorders of the self and their the 1980 U.S. labor force. Social Science Research, 16, 74-105.
treatment: An outline. International Journal of Psycho-Analysis, Terry-Short, L. A., Owens, R. G., Slade, P. D., & Dewey, M. E.
59, 413-425. (1995). Positive and negative perfectionism. Personality and
Lewinsohn, P. M., & Rosenbaum, M. (1987). Recall of parental Individual Differences, 18, 663-668.
behavior by acute depressives, remitted depressives, and nonde- Weissman, M. M., Sholomskas, D., Pottenger, M., Prusoff, B. A., &
pressives. Journal of Personality and Social Psychology, 52, Locke, B. Z. (1977). Assessing depressive symptoms in five
611-619. psychiatric populations: A validation study. American Journal of
Miller, S. B. (1996). Shame in context. Hillsdale, NJ: Analytic Epidemiology, 106, 203-214.
Press.
Minarik, M. L., & Ahrens, A. H. (1996). Relations of eating
behavior and symptoms of depression and anxiety to the Received January 19, 1999
dimensions of perfectionism among undergraduate women. Revision received August 2, 1999
Cognitive Therapy and Research, 20, 155-169. Accepted August 18, 1999 •