See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222574471

The fit of Holland’s RIASEC model to US occupations

Article in Journal of Vocational Behavior · May 2007

DOI: 10.1016/j.jvb.2007.04.002

CITATIONS READS

68 7,688

3 authors:

Chi-Ping Deng Patrick Armstrong

National Changhua University of Education Iowa State University
13 PUBLICATIONS 100 CITATIONS 44 PUBLICATIONS 2,476 CITATIONS

SEE PROFILE SEE PROFILE

James Rounds
University of Illinois, Urbana-Champaign
156 PUBLICATIONS 9,353 CITATIONS

SEE PROFILE

All content following this page was uploaded by James Rounds on 08 February 2018.

The user has requested enhancement of the downloaded file.

 Journal of Vocational Behavior 71 (2007) 1–22
www.elsevier.com/locate/jvb

The fit of Holland’s RIASEC model

to US occupations
Chi-Ping Deng a, Patrick Ian Armstrong b, James Rounds a,*

a
Department of Educational Psychology, University of Illinois at Urbana-Champaign,
1310 South 6th Street, Champaign, IL 61820, USA
b
Department of Psychology, Iowa State University, IA, USA
Received 18 December 2006
Available online 7 April 2007

Abstract

Holland’s [Holland, J. L. (1959). A theory of occupational choice. Journal of Counseling Psychology,

6, 35–45; Holland, J. L. (1997). Making vocational choices: A theory of vocational personalities and work
environments (3rd ed.). Odessa, FL: Psychological Assessment Resources, Inc.] RIASEC types were ini-
tially developed using a restricted range of occupational titles. Holland’s type classification system has
been extended to encompass the full range of occupations in the US, using both statistical and expert
rating methods. However, the extent that Holland’s classification model is sufficient to represent the full
range of occupational interests has not been examined. Multidimensional scaling (MDS) was used to
analyze college students’ (266 men, 572 women) interests in occupations representing approximately
85% of the US labor market. A two-dimensional MDS solution of the full set of occupations did not
fit Holland’s model, but limiting the analysis to occupations used in Holland-based measures produced
the expected RIASEC structure. In comparison, a three-dimensional solution included Prediger’s [Pred-
iger, D. J. (1982). Dimensions underlying Holland’s hexagon: Missing link between interests and occu-
pations? Journal of Vocational Behavior, 21, 259–287] dimensions (Things/People and Data/Ideas)
consistent with Holland’s model, but also included prestige and sex-type dimensions that were not
orthogonal to Prediger’s dimensions. These results demonstrate that the RIASEC types are not sufficient
to represent the full range of occupational interests and are confounded with prestige and sex-type.
 2007 Elsevier Inc. All rights reserved.

Keywords: Structural validity of interest measures; Occupational interests; RIASEC model; Occupational
prestige; Sex-type

*
Corresponding author.
E-mail address: [email protected] (J. Rounds).

0001-8791/$ - see front matter  2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jvb.2007.04.002
2 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

1. Introduction

Almost 40 years ago, Holland, Whitney, Cole, and Richards (1969) identified the RIA-
SEC hexagon that has formed the basis of Holland’s (1959, 1997) theory of six vocational
interest types and parallel work environments (Realistic, Investigative, Artistic, Social,
Enterprising, and Conventional; referred to by the acronym RIASEC). The hexagon, or
circumplex (Tracey & Rounds, 1995), a structural model used to represent the inter-rela-
tions among the six types and work environments (see Fig. 1) has been central to Hol-
land’s theory and application in applied setting. Subsequently, numerous structural
studies have supported the typology with measures of the RIASEC personality types
(e.g., Day & Rounds, 1998; Rounds & Tracey, 1993), and with work environments
(Rounds, Smith, Hubert, Lewis, & Rivkin, 1999; Tracey & Rounds, 1992). Despite these
efforts, the fundamental question of the sufficiency of the RIASEC types and work envi-
ronments remains unanswered. To date, there have been no published studies that directly
and systematically evaluate Holland’s model with occupations that represent the full range
of employment opportunities in the US labor market. The purpose of the present study is
to evaluate how well Holland’s RIASEC model generalizes to occupations that represent
majority of the US workforce.
Although there is substantial empirical support for Holland’s (1959, 1997) RIASEC
structure, this evidence is found primarily with measures specifically designed to measure
the six types (see, for example, the meta-analysis by Tracey & Rounds, 1993). At the occu-
pational level, there is surprisingly little research evaluating how well the six types repre-
sent the structure of the US labor market. When Holland (1958) developed the VPI, a
small number of occupational titles were chosen to represent the six types. Subsequently,
different strategies have been used to generate RIASEC codes for occupations in the US,
including incumbent sample scale scores (Harmon, Hansen, Borgen, & Hammer, 1994),
statistical prediction with US Department of Labor databases (Gottfredson & Holland,
1996), and expert ratings (Rounds et al., 1999). However, assuming that the RIASEC

Fig. 1. Holland’s (1959, 1997) interest types with Prediger’s (1982) dimensions.
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 3

structure is the best fitting model for occupations, these strategies do not directly test the
generalizability of Holland’s model to the full range of US occupations.
In the present study, the US Department of Labor’s Occupational Outlook Handbook
(JIST Work Inc., 1997) was used to identify the predominant occupations that represent
the world of work. It is hypothesized that the Holland model may be contingent upon
using sets of occupational titles limited to those occupations clearly identified with a RIA-
SEC type, such as those listed in the Vocational Preference Inventory (VPI, Holland, 1985).
Therefore, using the predominant occupations in the US labor market would either pro-
vide stronger validity evidence for the Holland model, or provide a useful starting point
for the development of a more representative model of occupations.

2. RIASEC model

Occupational titles are central to Holland’s development of the RIASEC typology and
its associated interest measures. In part, the rationale for the development of the Voca-
tional Preference Inventory, which is entirely composed of occupational titles, is the
assumption that preferences for occupations are an expression of personality (Holland,
1958, 1997). Holland (1997) notes that:
Most interest inventories rest heavily on the assumptions that people perceive com-
mon occupations and their associated activities accurately and that these perceptions
remain the same over long period of time. In the same way, a person’s vocational
preferences and choices rest on the same assumptions. If perceptions of occupations
had no validity, interest inventories would have little or no validity, and the average
person would have great difficulty in selecting suitable jobs.’’ (p. 10).
From Holland’s perspective, both occupational perceptions and occupational interest
structures are organized according to the hexagonal model, which facilitate individuals’
attempts to find occupations that match with their occupational interests (Holland,
1976, 1996).
Holland et al. (1969) first proposed that the inter-relations among the six RIASEC
types can be represented by a hexagon structure. The hexagon was identified in an exam-
ination of the correlations among the RIASEC scales of the VPI. As shown in Fig. 1, the
distances between the six types reflect the degree of similarity between the types, with the
most similar types in adjacent positions around the hexagon, and the least similar types at
opposite ends of the hexagon. Early attempts to evaluate the structural hypothesis devel-
oped by Holland et al. (1969), such as Cole and Hanson (1971), often suffered from meth-
odological limitations (Hubert & Arabie, 1987; Rounds, 1995). Over time, however,
comprehensive meta-analyses (Rounds & Tracey, 1996; Tracey & Rounds, 1993), com-
bined with research using more representative samples (Day, Rounds, & Swaney, 1998;
Fouad, Harmon, & Borgen, 1997) and increasingly sophisticated structural methods
(Armstrong, Hubert, & Rounds, 2003) have provided empirical support for Holland’s
structural model. The preponderance of this evidence suggests that the RIASEC hexagon
is valid for large representative samples of the US population, although there continues to
be concern with US minority (Armstrong et al., 2003) and especially, international samples
(Rounds & Tracey, 1996).
A critical limitation of previous structural investigations of Holland’s theory is that
these studies almost invariably use measures explicitly designed to replicate the six
4 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

RIASEC types. The focus of interest assessment research has been on measures designed
to reflect the characteristics of the six RIASEC types, such as the Strong Interest Inventory
(SII, Harmon et al., 1994), the Unisex edition of the ACT Interest Inventory (American
College Testing, 1995), and the VPI (Holland, 1985). An analysis of one or more of these
measures provides evidence of the extent to which the order predictions of Holland’s the-
ory are matched by Holland type measures. By comparing results between different
groups, important validity evidence is obtained for the equivalence RIASEC measures
with diverse populations (Fouad et al., 1997). This research does not, however, address
the larger question of the extent to which the RIASEC types are sufficient to represent
the full range of individual differences in occupational interests.
When a more extensive number of occupations have been used than are found in the
VPI, a three-dimensional model has emerged. Tracey and Rounds (1996) in developing
the spherical model of occupational interests selected 110 occupations that were not
‘‘pure’’ examples of the RIASEC types and that were at a skilled and nonskilled worker
level. Most important, occupational titles were selected that bridged the six RIASEC types
(c.f., Trapnell, 1989). Einarsdottir and Rounds (2000) used 110 occupational title items
from the Strong Interest Inventory (Harmon et al., 1994). In neither study was there an
attempt to choose a representative sample of US occupations. Nevertheless, expanding
the number of occupations resulted in three-dimensional spatial models of occupational
interests.

3. Dimensional interpretations of RIASEC structure

Several proposals have been advanced to account for the RIASEC interest space. The
most well know is Prediger’s (1982; Prediger and Swaney, 2004) claim that the RIASEC
structure is best represented using two bipolar dimensions (see Fig. 1). These two dimen-
sions reflect preferences for four work tasks: Things, People, Data, and Ideas. Things tasks
involve nonpersonal processes, such as producing and repairing. People tasks involve
interpersonal processes, such as helping and persuading. Data tasks involve impersonal
processes, such as recording and organizing facts. Ideas tasks involve intrapersonal pro-
cesses, such as creating and discovering. Based on an analysis of the 3rd and the 4th edi-
tions of Dictionary of Occupational Titles (DOT), Prediger developed bipolar Things and
People (i.e., T/P) dimension and Data and Ideas (i.e., D/I) dimension, with these two
dimensions being independent. Prediger’s dimensions are essentially predicated on Hol-
land’s circumplex structure: The better that the circumplex model fits RIASEC data,
the better the fit of Prediger’s dimensions (Rounds & Tracey, 1993). In the present study,
we evaluate the fit of Prediger’s dimensions in two- and three-dimensional space.
When researchers began to reinterpret Holland’s (1959, 1997) model in terms of under-
lying dimensions, a logical extension of the model that emerged was the potential for add-
ing a third dimension to the RIASEC structure (Rounds & Day, 1999). Tracey and
Rounds (1996) presented a three-dimensional structure of occupational interests with pres-
tige as the third dimension. They suggested that vocational interests are found to lie on the
surface of a sphere, with the third dimension of prestige orthogonal to two Holland-
circumplex dimensions. Occupational prestige is a broad construct with many highly
related referents such as status, educational level, behavioral control, and responsibility
(Tracey & Rounds, 1996). Typically, as occupations increase in prestige, they require more
education, provide more financial compensation, are more complex, involve exercising
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 5

more authority, and require higher levels of ability (Gottfredson, 1996). In fact, the VPI
(Holland, 1958, 1985) includes a Status scale to measure people’s career concern about
prestige and power.
Tracey and Rounds (1996) identified three methodological artifacts that may contribute
to failures to incorporate a prestige dimension into interest models. First, the prestige
range of occupations used in most studies is limited. Second, studies often fail to sample
prestige variance evenly across the six RIASEC scales, with prestige variance confounded
with RIASEC scales. For example, the R and C types tend to have a larger proportion of
lower prestige occupations. Third, most studies focus on scale scores instead of item
scores, and scale scores confound prestige variance with content variance. Researchers,
therefore, may fail to extract a third dimension of prestige because the measures contain
a limited range of items and are conducting analyses at the scale level instead of the indi-
vidual item level.
In response to Tracey and Rounds (1996), Harmon (1996) suggested that similar meth-
odological concerns may limit the emergence of sex type as an interest dimension, and
hypothesized that sex type may emerge as a dimension if a study included items represent-
ing the full range of variation in sex type. In an item level analysis of SII occupational
titles, Einarsdottir and Rounds (2000) demonstrated that sex-type emerged as an interest
dimension in a three-dimensional structure. They noted that, according to studies of Gott-
fredson’s (1981, 1996) theory, sex-type and prestige are also confounded in research on
occupational aspirations. Male sex-typed occupations vary from high to low prestige while
female sex-typed occupations cluster in the middle to low prestige with few high-prestige
female sex-typed occupations. Einarsdottir and Rounds (2000) concluded that investiga-
tors are unlikely to detect sex-type and prestige as two separate dimensions in the interest
structure. In an attempt to avoid these potential methodological artifacts that impact the
analysis of prestige and sex-type interest dimensions, the present study used occupations
representative of US workforce that include considerable prestige and sex type variation.

4. The present study

In the present study occupational interest structure was investigated using a set of occu-
pational titles identified by the US Department of Labor as representing approximately
85% of the US labor market and workforce. It is proposed that comprehensively surveying
the domain of occupational titles is important for identifying the core dimensions of inter-
est structure. This strategy addresses an important limitation of previous studies, which
have based their findings on RIASEC measures that have a restricted range of interests
represented by the types. We evaluated Holland’s model in a two- and three-dimensional
occupational space. In the two-dimensional test, we compared the results obtained using
the comprehensive set of occupational titles to the results obtained on a benchmark, the
occupational titles from the VPI (Holland, 1985). We expected that VPI occupational titles
would show a better fit than the comprehensive set of occupational titles. To examine the
incremental improvements in fit and the explanatory power of a two- compared to a three-
dimensional models of occupational interests, a dimensional analysis of the occupational
space was conducted We tested the capacity of a two- versus a three-dimensional occupa-
tional structure to capture Prediger’s dimensions of T/P and D/I and the dimensions of
prestige and sex-type. The dimensional analysis was important for accurately locating
6 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

the RIASEC model in three-dimensional space and allowed a test whether prestige and
sex-type dimensions are confounded.

5. Methods

5.1. Participants

Participants were college students enrolled in sections of a career development and

exploration course offered at a large Midwestern university. They completed the survey
as part of a research participation requirement in the course. The surveys were distributed
to the students in class, with a total of 867 surveys returned. Participants (n = 22) who had
more than three missing values were excluded from the study. Participants who did not
report their sex (n = 7) were also excluded because the analysis examined differences
between men and women in interest structure, resulting in a total of 838 study participants,
266 men and 572 women. Participants were asked to indicate their racial or ethnic identity
with 6.6% identifying as Asian American, 6.8% as Hispanic American, 16.6% as African
Americans, 0.1% as Native Americans, and 68.1% as Caucasian (1.8% did not report eth-
nicity). With respect to year in college, 28.9% were freshmen, 30.2% sophomores, 15.4%
juniors, 24.5% seniors, and 1.0% graduate students (1.1% did not report year in school).
Participants reported the following fields of study: 50.6% in Social and Behavioral Science,
15.2% in Business, 10.2% in Science or Engineering, and 21.5% undecided (2.5% did not
report their major field). The average age of participants was 19.78 (SD = 2.04).

5.2. Occupational preference inventory

We developed an inventory, called the Occupational Preference Inventory (OPI), to

measure occupational interests. The OPI item pool initially contained 277 occupation titles
from the Enhanced Occupational Outlook Handbook (EOOH; JIST Work Inc., 1997).
The EOOH contains 250 major occupations that cover about 85 percent of all jobs in
the US Occasionally, the EOOH has occupations that combine two titles. For example,
accountant and auditor are listed as one occupational title. In such cases, we separated
the occupational title into two titles for the OPI. After 27 titles from the EOOH were sep-
arated into two distinct occupational titles, the 250 EOOH titles yielded 277 occupational
title items. For each occupation on the OPI participants were instructed to ‘‘show how
much you think you would like or dislike doing that kind of work’’ on a 7-point scale, with
the anchors of 1 labeled ‘‘strongly dislike,’’ 4 labeled ‘‘indifferent,’’ and 7 labeled ‘‘strongly
like.’’

5.3. Matching OPI occupations to O*NET occupations

The Occupational Information Network (O*NET) database (O*NET Resource Center,

2003) was used to provide the interest and prestige information for the OPI occupations.
The O*NET was developed by the US Department of Labor as an electronic successor to
the final version of DOT (U.S. Department of Labor, 1991). Four documents were used to
match the OPI occupations to O*NET occupations: job descriptions in EOOH, job titles
and descriptions of O*NET occupations (O*NET Resource Center, 2003), the crosswalk
table between the DOT and the O*NET job titles (O*NET Resource Center, 2003), and
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 7

the crosswalk table between Standard Occupational Classification (SOC) and O*NET job
titles (O*NET Resource Center, 2003). While matching the OPI occupations to O*NET,
nine occupations were excluded because they were either too ambiguous or too general.
For example, the job title painter can be interpreted as either an artist’s occupation or
as a type of construction worker. Among the remaining 268 OPI occupation titles, 244
were exact matches either to O*NET, DOT, or SOC job titles. The remaining 24 titles were
matched by two of the authors based on the similarity of their job descriptions in EOOH
to the job descriptions of O*NET occupations. There was agreement on the matching of
titles for 21 of the 24 occupations, yielding an agreement rate of 87%. Disagreements were
resolved by discussion between raters. The final version of the OPI contained 268 occupa-
tional titles.

5.4. Assignment of Holland codes to occupations

O*NET RIASEC ratings (Rounds et al., 1999) were used to assign each of the OPI
occupation into one of the Holland’s six types. Each occupation in O*NET database
has an occupational interest profile indicating how descriptive and characteristic the occu-
pation is for each RIASEC work environment. Rounds et al. (1999) examined the struc-
tural validity of these ratings, and reported that the O*NET RIASEC ratings were
consistent with the structure of Holland’s theoretical model, supporting the validity of
the ratings. A recent study by Eggerth, Bowles, Tunick, and Andrew (2005) comparing
the RIASEC codes of occupations from the O*NET, the Dictionary of Holland Occupa-
tional Codes (DHOC; Gottfredson & Holland, 1996), and the Strong Interest Inventory
(Harmon et al., 1994) has also demonstrated the validity of first-letter codes of occupa-
tions listed in the O*NET, with 79% agreement with the DHOC, and 71% and 74% agree-
ment with the female and male reference norms (respectively) from the Strong Interest
Inventory. The distribution of the RIASEC types across the 268 OPI occupational titles
(frequency and percent in parentheses) is: R (80, 29.85%), I (41, 15.30%), A (24, 8.96%),
S (38, 14.18%), E (45, 16.79%), and C (40, 14.93%).

5.5. Work-task, prestige, and sex type variables

Four variables were created to represent Prediger’s (1982) work-task dimensions, as

well as measures of prestige, and sex-type. We used Prediger’s (1982) formulas and terms
(i.e., T/P and D/I) to represent the two work-task dimensions based on the RIASEC scale
scores for each occupation. Scores for each occupation on the RIASEC scales were
obtained from the O*NET database. The formula for the T/P score was
2.00 · (R S) + I + C A E; and for the D/I score was 1.73 · (E + C A I). For
example, the T/P score of dancer is 2 · (4 4) + 2.67 + 2.33 6.67 3.67 (i.e., 5.33)
and its D/I score is 5.77.
Three variables were used to construct a measure of prestige for the OPI occupations:
(1) vocational preparation, (2) a composite score of two need-reinforcers, recognition and
social status, and (3) mean annual salary. Vocational preparation required for each occu-
pation came from the O*NET Job Zone ratings (Oswald, Campbell, McCloy, Rivkin, &
Lewis, 1999). A Job Zone is a group of occupations that are similar in how much overall
experience, education, and on-the-job training people need to be able to do the work.
The O*NET Job Zone uses a five-point scale: 1 = little or no preparation, 2 = some
8 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

preparation, 3 = medium preparation, 4 = considerable preparation, and 5 = extensive

preparation. The distribution of the 268 OPI occupational titles across the five Job Zones
(frequency and percent in parentheses) is: little or no preparation (31, 11.57%), some
preparation (47, 17.54%), medium preparation (61, 22.76%), considerable preparation
(90, 33.58%), and extensive preparation (39, 14.55%).
Need-reinforcers, derived from the Theory of Work Adjustment (Dawis & Lofquist,
1984), characterize the nature of the work and the conditions in work environments. Social
Status refers to the extent that workers in the occupations ‘‘would be looked up to by oth-
ers in the company and the community.’’ Recognition refers to the extent that workers in
the occupation ‘‘receive recognition for the work they do.’’ Need-reinforcer occupational
scores were obtained from the O*NET database (McCloy et al., 1999). Because these two
need-reinforcers are highly correlated across O*NET occupations (r = .85, n = 268), a
composite need-reinforcer variable was constructed by calculating the mean of the Social
Status and Recognition scores.
The 2003 mean annual salary (MAS) of each occupation was obtained from the
Occupational Employment Statistics Survey (Bureau of Labor Statistics, 2004b). A
composite variable representing prestige was created from a principle components
analysis. Only the first component had an eigenvalue larger than 1, accounting for
79.05% of the total variance, indicating that the three variables are highly related. The
regression function for the composite prestige variable had similar weights for all three
variables: Prestige = 0.37 * z-Job Zone + 0.37 * z-MAS + 0.38 * z-Reinforcer.
To construct a measure of the sex-typing of occupations, we initially used the approach
of Einarsdottir and Rounds (2000). Occupations were coded according to the percentage
of female employees using the 2003 annual average data of the Current Population Survey
(Bureau of Labor Statistics, 2004a). The information, however, was only available for 213
of the 268 occupations. To develop an alternative method for all of the 268 occupations,
we calculated the difference between female (N = 572) and male (N = 266) mean interest
ratings for each occupation. To evaluate the similarity of these sex-type estimates, we
correlated the percentage of female employees with the interest difference scores for the
213 occupations. For example, the mean interest difference score of Guard is 1.10
(female M = 2.03 and male M = 3.13) and the percentage of female guards is 12% or
the mean interest difference score of Register Nurse is 1.17 (female M = 3.93 and male
M = 2.76) and the percentage of female registered nurses is 92%. The correlation was
.78, indicating that interest-rating difference methodology produced an effective indicator
of the sex-typing of an occupation that is similar to the percent estimates of female
employees (see Rounds, 1995, for application of this method to scale scores).

5.6. Analysis

To evaluate sex-differences in interest structure, separate correlation matrices were

calculated for women and men. Because there was missing item data, pairwise deletion
was used in calculating the correlations. These correlation matrixes were then scaled using
the three-way (individual differences) MDS. The fit index of Variance Accounted For
(VAF) and the salience weights were examined to determine if there are sex differences
in interest structure.
We used several techniques to determine if Holland’s model was best represented in
three dimensions. First, MDS yields a spatial representation that was examined to
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 9

evaluate Holland’s hypothesis that the types are ordered in a circular fashion of R-I-A-S-
E-C. Secondly, a correspondence index (CI; Hubert & Arabie, 1987) was calculated to
examine the violations and agreements of Holland’s order hypothesis (i.e., the relations
between the adjacent types are stronger than the relations between the alternate types,
and the relations of the alternate types are stronger than the relations between the opposite
types). The CI is calculated from the number of agreements minuses the number of viola-
tions, and then it is divided by the total number of agreements and violations (or
predictions).
Although there is a probabilistic interpretation of CI, the p-value only tests the random-
labeling hypothesis, i.e., the random selection of order conjectures from the given data.
Therefore, this procedure does not provide direct information whether the spatial arrange-
ment of Holland’s model holds (Rounds, 1995). To avoid confusion, the p-values of CI
were not reported. Instead, the CI values of this study are compared to a US benchmark,
developed on 73 US RIASEC correlation matrices (Rounds & Tracey, 1996). The mean CI
value of the benchmark is .70 with a lower bound 99% confidence interval of .66. As a
third test of the Holland structure, we used the technique of Circular Unidimensional Scal-
ing (CUS, Hubert, Arabie, & Meulman, 1997) to evaluate the circular ordering of the
RIASEC types, following established procedures for using the CUS technique to evaluate
Holland’s model (see Armstrong et al., 2003). The CUS technique fits the observed dis-
tances between types to a circumplex model, and uses a VAF statistic to evaluate model
fit. To address potential concerns with the effects of local optima, 500 random starting per-
mutations were used for each analysis. Finally, we examined the fit of the Prediger’s T/P
and D/I dimensions. If the occupations show a reasonable fit to Holland’s model, these
dimensions should be well represented in the MDS solution. We used the R2 measure of
VAF from the property vector fitting results to evaluate how well Prediger’s dimensions
were represented.

6. Results

The evaluation of the extent to which Holland’s model is sufficient to account for occu-
pational structure begins with fitting of Holland’s model in a two-dimensional structure,
using both the 268 OPI occupations and the subset of 49 VPI occupations. Next, we eval-
uated the fit of Holland’s model in a three-dimensional structure using the same data.
Evaluations of Holland model fit are based on a direct test of Holland’s RIASEC order
hypotheses, and the extent to which either two- or three-dimensional solutions account
for variables representing Prediger’s (1982) T/P and D/I dimensions, and measures of
Prestige and Sex-Type. Holland’s model was also evaluated by comparing the results
obtained with the full set of OPI occupations and the subset of VPI occupations. Before
using a combined sample of men and women in the MDS analyses, we first conducted an
analysis of sex differences, testing whether the occupational interest structures for men and
women were similar.

6.1. Sex differences in occupational interest structure

As a first step in the data analysis, the potential for gender differences in structure were
examined. We expected to find similar occupational interest structures for women and men
since it is rare to find sex differences in vocational interest structures (e.g., Anderson,
10 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

Tracey, & Rounds, 1997). Correlations among the 268 OPI items were calculated sepa-
rately for male and female students, and were then submitted to a three-way (individual
differences) multidimensional scaling (MDS) analysis. Two dimensions were extracted.
Table 1 presents the stress, proportion of variance accounted for (VAF), and dimensional
salience weights from the two-dimensional solution by sex. The stress, VAF, and the sal-
ience weights are very similar for the male and female samples, suggesting that a common
occupational interest structure is present for both groups. The combined (male and
female) two-dimensional MDS configuration is shown in Fig. 2. To represent the locations
of the six RIASEC types, centroid coordinates were calculated by averaging the coordi-
nates of the occupations with same letter RIASEC code.

6.2. Holland’s model in two-dimensions

The MDS configuration presented in Fig. 2 does not support Holland’s model, with the
A-type occupations clustering between the S and the E occupations rather than between I
and the S occupations. The order of the mean points is RISAEC with a reversal of A and
S. The CI is .47, considerably below the lower bound of 99% confidence interval of .66.
Furthermore, as illustrated in Fig. 3, the Circular Unidimensional Scaling (CUS) analysis
of the 268 occupation solution (with the S and A types reversed) provides stronger evi-
dence that the order hypothesis of the RIASEC model is not supported. Overall, these
results indicate that Holland’s circular order model is not well represented in two dimen-
sions when examining the structure of occupational interest items that represent the full
range of occupations in the US labor market.

Table 1
Three-way MDS fit indices and salience weights for two- and three-dimensional solutions by occupational sets
Solution Stress VAF Dimensional weight
Dim 1 Dim 2 Dim 3
268 OPI occupations
Two-dimensional
Women 0.27 0.62 0.68 0.65 —
Men 0.29 0.58 0.66 0.66 —
Combined 0.28 0.60 — — —
Three-dimensional
Women 0.20 0.72 0.59 0.57 0.51
Men 0.21 0.68 0.56 0.56 0.55
Combined 0.20 0.70 — — —
49 VPI occupations
Two-dimensional
Women 0.21 0.77 0.70 0.68 —
Men 0.20 0.79 0.71 0.66 —
Combined 0.20 0.78 — — —
Three-dimensional
Women 0.14 0.84 0.56 0.54 0.60
Men 0.15 0.83 0.63 0.58 0.47
Combined 0.15 0.84 — — —
Note. VAF, Variance Accounted For; Dim, Dimension; OPI, Occupational Preference Inventory; VPI,
Vocational Preference Inventory.
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 11

Fig. 2. Two-dimensional MDS configuration of the 268 OPI occupations plotted by first-letter RIASEC code.
Centroid coordinates for each RIASEC type are circled.

Rounds (1995) found strong support for Holland’s model by using scales from mea-
sures designed to assess Holland’s RIASEC model. To further examine whether the pres-
ent data can provide similar evidence, we analyzed a subset of OPI occupations that
matched occupations in the VPI, a measure used by Holland et al. (1969) to develop
the hexagonal model. The VPI contains 84 occupations titles, 14 occupations for each
RIASEC type. Two graduate students who are familiar with Holland coding of occupa-
tions independently matched the 84 occupations to O*NET occupations. They agreed
on 70 of the 84 occupations, including both OPI-VPI occupational matches and non-
matches, for an overall agreement rate of 83%. The first author resolved the 14 disagree-
ments between raters. A total of 49 OPI occupations were matched to VPI occupations,
and were assigned RIASEC codes from the VPI, resulting in 10 R, 11 I, 8 A, 4 S, 7 E,
and 9 C occupations.
Three-way MDS of the 49 VPI occupations again indicated that men and women share
a similar structure (see Table 1). Fig. 4 illustrates the two-dimensional MDS solution for
these occupations. These results provide support for Holland’s model, with the occupa-
tions of the same type clustering together and distributed in a RIASEC-consistent order-
ing. The mean RIASEC item coordinates also support Holland’s arrangement of the six
types. The CI was calculated on the distances among the mean RIASEC item coordinates.
The CI value for this set of occupations was .67, within the benchmark CI 99% confidence
interval. Also, as illustrated in Fig. 3, the CUS analysis of the 49 VPI occupations sup-
ported the RIASEC order hypothesis. In short, these results show that Holland’s model
is well represented in two dimensions when the analysis is restricted to the 49 VPI
12 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

Fig. 3. Circular Unidimensional Scaling results for RIASEC centroids obtained from the two-dimensional MDS
solution for 268 OPI and 49 VPI occupations and the Holland Plane of the rotated three-dimensional MDS
solution for 268 OPI occupations.

occupations, but when the full range of occupations in the US labor market is evaluated,
the Holland model is not supported.

6.3. Comparisons between two- and three-dimensional models

To compare the fits of a two-dimensional and a three-dimensional model, a three-way

MDS was performed for both the 268 OPI occupations and the 49 VPI occupations. Both
three-way MDS solutions showed that men and women have similar three-dimensional
structures (see Table 1). Therefore, the three-way MDS solution from the combined sam-
ple was used. For the 268 OPI occupations, the VAF increases 10% from a two-dimen-
sional solution (VAF = .60) to a three-dimensional solution (VAF = .70). For the 49
VPI occupations, the addition of a third dimension increased VAF by 6% (from .78 to
.84). Property vector fitting (Kruskal & Wish, 1978) was used to examine whether a
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 13

Fig. 4. Two-dimensional MDS configuration of the 49 VPI occupations plotted by first-letter RIASEC code.
Centroid coordinates for each RIASEC type are circled.

three-dimensional solution represented Holland’s interest structure more effectively than a

two-dimensional solution. Measures of T/P, D/I, and Prestige were regressed on the two-
dimensional and then the three-dimensional stimulus coordinates. Characteristics that
achieved R2 (VAF) values greater than .50 were included in figures.
Table 2 shows the MDS and property vector fitting results for the 268 OPI occupations
and 49 VPI occupations. For the OPI occupations, T/P can be located as a dimension in
the two-dimensional configuration with a VAF of .56. D/I and Sex Type, however, are not
as well represented as T/P, and Prestige does not emerge as a dimension. However, in the
three-dimensional configuration, all three properties are well represented. Specifically, the

Table 2
Variance Accounted For (VAF) from property vector fitting analysis of four interest variables
Solution Interest variable
Things/People Data/Ideas Prestige Sex type
268 OPI occupations
Two-dimensional 0.56 0.38 0.08 0.42
Three-dimensional 0.68 0.61 0.51 0.54
49 VPI occupations
Two-dimensional 0.86 0.69 0.13 0.68
Three-dimensional 0.92 0.75 0.12 0.85
Note. OPI, Occupational Preference Inventory; VPI, Vocational Preference Inventory.
14 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

addition of a third dimension increased the explained variances (R2) of D/I from .38 to .61,
Prestige from .08 to .51, and Sex Type from .42 to .54. For the VPI occupations, in com-
parison, T/P, D/I, and Sex Type are well represented in a two-dimensional MDS config-
uration, and Prestige did not emerge in either the two-dimensional or the three-
dimensional configuration.
The property vector fitting results indicated that a three-dimensional model represented
the interest structure of US occupations better than a two-dimensional model. Prestige
cannot be located in the VPI occupations in a two-dimensional or a three-dimensional
model. To examine whether the lack of Prestige dimension is due to restriction of range,
we calculated the mean and standard deviation of Prestige for both the 268 occupations
and the 49 VPI occupations. The prestige range for the 268 occupations is from 2.01
to 3.50 with a M of .00 and a SD = 1.00, while the prestige levels of the 49 VPI occupa-
tions range from 1.29 to 2.48 (M = .33 and SD = .77). These results suggest that restric-
tion of range may be the reason why the prestige dimension could not be found in the VPI
occupational configuration.

6.4. Holland’s model in three-dimensions

The regression results have shown that the four interest dimensions fit well in the three-
dimensional OPI solution (see Table 2). The T/P and D/I vectors form a two-dimensional
plane that represents Holland’s model, embedded in the larger interest space of the three-
dimensions extracted in the MDS analyses of OPI occupations. Conceptually, this repre-
sents a subspace, or Holland plane, in the larger interest space of occupations in the US
labor market.
By creating a vector that is orthogonal to the Holland plane, a new three-dimensional
space was formed by rotating the original MDS coordinates. Table 3 shows the regression
results of T/P, D/I, Sex-type, and Prestige on this new three-dimensional space. As illus-
trated in Fig. 5, in the rotated three-dimensional solution, dimensions 1 and 2 are now
strongly aligned with measures representing Prediger’s T/P and D/I dimensions. In this
solution, as illustrated in Figs. 5 and 6, prestige is now aligned somewhat with both dimen-
sions 2 and 3, and sextype is aligned with dimension 1. These results clearly demonstrate
that prestige and sextype are embedded within Holland’s model when examining occupa-
tional interests. Prestige is orthogonal to both Sex Type and T/P, and has a negative asso-
ciation with D/I. Sex Type is also essentially orthogonal to D/I and is almost parallel with
T/P, but in opposite direction. The CI calculated on the distances between the RIASEC

Table 3
Interest variable directional cosines for three-dimensional MDS solution
Interest variable VAF Dimension weight
Dim 1 Dim 2 Dim 3
Things/People 0.68 0.82 0.00 0.00
Data/Ideas 0.61 0.16 0.78 0.00
Prestige 0.51 0.03 0.43 0.62
Sex-type 0.54 0.72 0.02 0.17
Note. VAF, Variance Accounted For; Dim, Dimension.
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 15

Fig. 5. Property vector fitting results for the Holland plane (dimensions one and two) of the rotated three-dimensional
MDS solution, with occupations plotted by first-letter RIASEC code. D/I, Data/Ideas; T/P, Things/People.

Fig. 6. Property vector fitting results for dimensions one and three of the rotated three-dimensional MDS
solution, with occupations plotted by first letter RIASEC code. T/P, Things/People.
16 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

mean coordinates on the Holland Plane is .72 which is higher than the US benchmark of .70
(Rounds & Tracey, 1996). When the distances between the six RIASEC types on this Hol-
land Plane are evaluated using CUS (see Fig. 3), the ordering is consistent with Holland’s
theory. Overall, these results strongly support the use of a three-dimensional model to rep-
resent occupational interest structure that includes Holland-based T/P and D/I dimensions.
To further evaluate the embedding of prestige and sex-type in Holland’s RIASEC
types, we computed the mean prestige and sex-type level of the RIASEC occupations
(see Table 4). The results of the two ANOVAs indicated there were significant differences
among the RIASEC types for both Prestige, F (5, 262) = 40.45, p < .001, g2 = .44, and Sex
Type, F (5, 262) = 26.68, p < .001, g2 = .34. Consistent with Gottfredson’s (1978, 1980)
findings, the R and C occupations have the lowest prestige level, and I occupations have
the highest prestige level. For sex-type, the R occupations have the lowest score (in the
direction of male type), and the S occupations have the highest score (in the direction
of female type). These results demonstrate that men prefer the R occupations more than
women, and women prefer the S occupations more than men.

7. Discussion

In the present study, the structure of interests was investigated using a set of occupa-
tions representing approximately 85% of the US labor market. To date, almost all studies
of vocational interest structure have focused on existing interest inventories. These mea-
sures often reflect the characteristics of the six Holland RIASEC types, because that is
what they are designed to measure, and perhaps it is not surprising that the structural
hypothesis of Holland’s model is supported by analyses of these inventories. This RIA-
SEC-based research does not, however, address the larger question of identifying the
key dimensions of individual differences in occupational interests. By using a more repre-
sentative set of occupations, the underlying dimensions of interest structure were examined
without preordaining the emergence of a Holland model.

Table 4
Means and standard deviations of prestige and sex-type variables by RIASEC type
Type N M SD Min Max
Prestige
Realistic 80 0.61 0.75 2.01 2.01
Investigative 41 1.07 0.76 0.06 3.50
Artistic 24 0.65 0.40 0.20 1.54
Social 38 0.12 0.88 1.84 1.76
Enterprising 45 0.48 0.83 1.87 2.48
Conventional 40 0.69 0.72 1.74 1.29
Sex-type
Realistic 80 0.71 0.45 1.59 0.52
Investigative 41 0.38 0.54 1.14 1.08
Artistic 24 0.09 0.64 0.54 1.77
Social 38 0.39 0.58 1.14 1.31
Enterprising 45 0.16 0.62 1.57 1.87
Conventional 40 0.20 0.39 1.17 0.80
Note. Min, minimum value; Max, maximum value.
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 17

The results for Holland’s model in the current analyses were somewhat mixed: The
RIASEC model was not identified in a two-dimensional occupational space, as expected,
but was identified in a more complex three-dimensional space. This finding has important
implications for the interpretation of Holland’s model and RIASEC-based measures.
These results also provide a starting point, similar to lexical hypothesis research in the per-
sonality area (Saucier & Goldberg, 1996), for the continued development of a model of
interest structure that reflects the full range of employment opportunities in the US.
Does the present three-dimensional occupational map adequately define the domain of
vocational interests? In part, the answer varies depending on the kinds of items researchers
use to assess interests (see Rounds, 1995, for a review of the vocational interest domain).
Kuder (1970) and Strong (1943) had different ideas on the kinds of items to include in a
vocational interest measures. Kuder (1970) argued for the use of activity items only.
Strong (1943) used a broad set of items including occupations, activities, leisure activities,
school subjects, preferences, types of people, and person characteristics. Holland and Ray-
mond (1986) began with occupations in the VPI but later with the Self-Directed Search, he
expanded his item domain to include activities and self estimates of skills (competencies).
Because the present study tested the generality of Holland’s theory, occupational titles
were used. Using the emergence of the Big 5 (Digman, 1989), as an exemplar, a structure
and taxonomy of vocational interests will involve a variety of methods to assess interests
and multiple studies that converge on a structural model.

7.1. Prestige as an interest dimension

This study supports the contention by Rounds and Day (1999) that a prestige dimen-
sion will emerge when items vary in their levels of prestige. Representing Holland’s RIA-
SEC model in two dimensions does not appear to capture vocational interest variability
caused by prestige differences across occupations. Besides the work-task oriented dimen-
sions, prestige is a variable contributing to occupational interests. Holland (1997) has sta-
ted that earlier versions of his interest theory focused more on the kind of work a person
performs than on level of work (i.e., prestige, income, level of talented required, etc.), but
he has also noted that level of work is a factor in vocational choice (Holland, 1959). Gott-
fredson and Holland (1996) proposed complexity as a third dimension, which they
defined as the cognitive demands placed on workers, as a distinguishing feature of occu-
pations. However, Holland’s hypotheses about the level of vocational choices and
achievement, and level of cognitive demand placed on workers have received little
research attention.
Studies using occupations (Tracey & Rounds, 1996) or activities (Tracey, 1997) indi-
cated that prestige can be interpreted as a third dimension that is independent, or orthog-
onal, to the two dimensions underlying Holland’s types. The present study, however,
demonstrated that prestige is not independent of Holland’s types. The key differences
between the results of Tracey and Rounds (1996) and the present study involve the num-
ber and type of occupations studied and how the spherical model was created. In the pres-
ent study, the I and A occupations had the highest prestige levels, and the C and R
occupations had the lowest prestige level. This replicates Gottfredson’s (1978) findings
that there is an association between Holland’s occupational classification and prestige.
As noted by Gottfredson, Holland, and Gottfredson (1975) these prestige differences in
types of work may limit career options for individuals who aspire towards either
18 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

high- or low-level occupations. For example, in career counseling there would be few
occupations to explore if an individual has strong C or R interests and also a strong pref-
erence for high prestige. Similarly, there may be limited occupational options for clients
who have strong A or I interests combined with a preference for working in a less
demanding, lower prestige occupation. The non-orthogonality of RIASEC types with
prestige could also confound the assessment of interests with prestige preferences. For
example, when completing an interest measure, an individual with prestigious aspirations
may tend to prefer I and A occupations, and reject C and R occupations, irrespective of
the work activities involved in the occupations.
The present study shows how work values and interests can be integrated with interests
as an organizational framework. Prestige was assessed using need-reinforcers (social status
and recognition) from the theory of work adjustment (Dawis, 2005). These need-reinforc-
ers from the environmental side are considered occupational reinforcers and from the per-
son side of the equation –work values. The present study found that these occupational
reinforcers from the O*NET can be incorporated within an interest space. Similarly, Arm-
strong, Smith, Donnay, and Rounds (2004) developed the Strong Ring, a model that has
interests as an organizational framework to integrate work values and skills. These studies
and others (e.g., Ackerman & Heggestad, 1997; Mount, Barrick, Scullen, & Rounds, 2005)
demonstrate that interest structures form the base for understanding how personality
traits, values, and abilities are interrelated.

7.2. Sex-type as an interest dimension

The present study demonstrates that sex-type is also a dimension of occupational

interest structure. Sex-type was found to be highly correlated with T/P, and orthogonal
to the D/I and Prestige dimensions. Men scored higher than women on the Things end
of T/P dimension (i.e., R occupations) and women higher than men on the People end
(i.e., S occupations). These results are similar to previous studies that have examined
sex-differences in vocational interests. For example, Tracey and Rounds (1992) analyzed
SDS norm data that Holland collected from college and high school students in 1985
and 1987. In their database, more males had R and I as their first letter code than
females and more females had S as their first letter code than males. Lubinski (2000)
commenting on Lippa’s (1998) research on gender and interest structure reported that
the People (Social) and Things (Realistic) dimension is the ‘‘largest of all sex differences
on major psychological dimensions’’ (p. 421). Helwig (2002) conducted a longitudinal
study on children’s occupational aspiration. His study showed that girls reported signif-
icantly higher levels of working with people than boys through Grade 2 to Grade 12,
and boys reported significantly higher in working with things than girls from Grade
2 to Grade 10.
Gottfredson (1996) proposed developing cognitive maps of occupations by plotting
dimensions based on sex-type and prestige level, and then embedding interests in the
resulting occupation space. Her proposal represents an initial attempt to integrate
interests, prestige, and sex-type. In comparison, The present results show that sex-type
and prestige are confounded with Holland’s type definitions. The current findings sup-
port the continued development of a more integrative approach to representing
interests as part of a cognitive map that includes prestige level and sex type as
dimensions.
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 19

7.3. Interpreting Holland’s model

In the area of interest research, attempts have been made to draw parallels between the
six Holland RIASEC types and the Big Five model of personality (see, for example, Lar-
son, Rottinghaus, & Borgen, 2002; Mount et al., 2005). The emergence of the Big Five
model of personality from lexical hypothesis research provides a comprehensive taxonomy
for describing individual differences in personality (Goldberg, 1993). Although some
researchers have argued that additional factors are necessary to capture the full range
of individual differences (e.g., Block, 1995; Paunonen & Jackson, 2000), the debate over
personality structure is now framed by the lexical hypothesis-based goal of representing
the full range of personality adjectives. Attempts to draw parallels between the Big Five
and the RIASEC types, as the initial development of the types, as outlined in Holland,
1959, did not involve a systematic, empirically-based evaluation of occupational interest
structure that would be comparable to the lexical hypothesis. The present results, obtained
using a representative set of occupational titles, clearly suggest that the six Holland types
are not sufficient to represent the full range of individual differences in occupational inter-
ests. Therefore, calling the Holland types the ‘Big Six’ may be somewhat premature, and
results obtained supporting the Holland’s model in previous research may be an artifact of
using the limited range of interest items in RIASEC-based measures.
In his critique of the lexical hypothesis and the five factor model of personality, Block (1995)
identified two issues that are relevant to the study of occupational interest structure. First,
Block claims that it is possible for researchers to bias the results of a lexical hypothesis-based
study by imposing an a-priori structure. For example, with personality adjectives this may hap-
pen as researchers group words into categories to reduce the number of redundant synonyms.
Rounds and Zevon (1983) have raised similar concerns with interest measures. The external
criteria of using occupations that represent the most frequently employment opportunities
in the US labor market was used in the present study to select occupations. This was done
to evaluate the impact of item selection bias on the obtained results, and the selection of occu-
pations does appear to impact the results of structural analyses. Holland’s RIASEC model can
be identified in a two-dimensional structure, but only if the analysis is limited to occupations
from Holland’s VPI. The RIASEC model is not identified when occupations are used that are
representative of the full range of employment opportunities for the US workforce. The poten-
tial effects on the career counseling process of using assessment measures with a limited range
of interest items is an important issue to explore in future research.
Another concern raised by Block (1995) is that research conducted with college students
may bias results towards a simplified picture of personality structure due to developmental
issues. Similarly, when occupational interests are examined, the emerging structure may
reflect in some ways the limited work experiences of college students. This may limit the
implications of the current findings to applications in college settings. Replication of the
current study with different populations is required to address this concern. However,
because Holland’s model was originally developed with high school and college students,
it seems unlikely that the limitations found here will be less of an issue when working with
other groups. It is also worth noting that college students remain an important target audi-
ence for interest assessments, and more generally for career counseling interventions, mak-
ing the structure of interests for this population and the validity of interest measures used
with college students an important research question irrespective of the extent to which
obtained results generalize to other populations.
20 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

8. Summary and conclusions

The finding that Holland’s (1959, 1997) types are not sufficient to represent the
full range of occupations in the US has important implications for the use of RIA-
SEC-based assessment measures. For individuals whose interests fall within the occu-
pational space represented by the types, current measures may be sufficient.
However, with individuals considering career choices that fall outside the RIASEC
types, current interest measures may present an overly restricted and distorted range
of career choices. We also found that prestige and sex type were embedded in Hol-
land’s categorization of interest types, demonstrating the importance of level of work
and the influences of socialization in occupational interest. It is important to incor-
porate prestige and sex type with work-task interests in practice. Finally, the present
study demonstrated that the prestige dimension is attenuated if occupations with
limited prestige levels are used for scale construction. In future investigations of
occupational interest structure researchers should use a set of occupational titles that
are more representative of the full range of employment opportunities considered by
individuals.

References

Ackerman, P. L., & Heggestad, E. D. (1997). Intelligence, personality, and interests: Evidence for overlapping
traits. Psychological Bulletin, 121, 219–245.
American College Testing, Inc. (1995). Technical manual: Revised Unisex Edition of the ACT Interest Inventory
(UNIACT). Iowa City, IA: Author.
Anderson, M. Z., Tracey, T. J., & Rounds, J. (1997). Examining the invariance of Holland’s vocational interest
model across gender. Journal of Vocational Behavior, 50, 349–364.
Armstrong, P. I., Hubert, L., & Rounds, J. (2003). Circular unidimensional scaling: A new look at group
differences in interest structure. Journal of Counseling Psychology, 51, 297–308.
Armstrong, P. I., Smith, T. J., Donnay, D. A. C., & Rounds, J. (2004). The Strong ring: A basic interest model of
occupational structure. Journal of Counseling Psychology, 51, 299–313.
Block, J. (1995). A contrarian view of the five-factor approach to personality description. Psychological Bulletin,
117, 187–215.
Bureau of Labor Statistics (2004a). Household data-Annual averages. Current Population Survey. Available from
http://www.bls.gov/cps/cpsaat11.pdf.
Bureau of Labor Statistics (2004b). National employment and wage data. The Occupational Employment
Statistics Survey. Available from http://www.bls.gov/oes/oes_dl.htm#2003.
Cole, N. S., & Hanson, G. R. (1971). An analysis of the structure of vocational interests. Journal of Counseling
Psychology, 18, 478–486.
Dawis, R. V. (2005). The Minnesota theory of work adjustment. In S. D. Brown & R. W. Lent (Eds.), Career
development and counseling: Putting theory and research to work. Hoboken, New Jersey: Wiley.
Dawis, R. V., & Lofquist, L. H. (1984). A psychosocial theory of work adjustment. Minneapolis, MN: University
of Minnesota Press.
Day, S. X., & Rounds, J. (1998). Universality of vocational interest structure among racial and ethnic minorities.
American Psychologist, 53, 728–736.
Day, S. X., Rounds, J., & Swaney, K. (1998). The structure of vocational interests for diverse racial-ethnic
groups. Psychological Science, 9, 40–44.
Digman, J. M. (1989). Five robust trait dimensions: Development, stability, and utility. Journal of Personality, 57,
195–214.
Eggerth, D. E., Bowles, S. M., Tunick, R. H., & Andrew, M. E. (2005). Convergent validity of O*NET Holland
code classifications. Journal of Career Assessment, 13, 150–168.
Einarsdottir, S., & Rounds, J. (2000). Application of three dimensions of vocational Interests to the Strong
Interest Inventory. Journal of Vocational Behavior, 56, 363–379.
 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22 21

Fouad, N. A., Harmon, L. W., & Borgen, F. H. (1997). Structure of interests in employed male and female members
of U.S. racial-ethnic minority and nonminority groups. Journal of Counseling Psychology, 44, 339–345.
Goldberg, L. R. (1993). The structure of phenotypic personality traits. American Psychologist, 48, 26–34.
Gottfredson, G. D. (1996). Prestige in vocational interests. Journal of Vocational Behavior, 48, 68–72.
Gottfredson, G. D., & Holland, J. L. (1996). Dictionary of Holland occupational codes (3rd ed.). Odessa, FL:
Psychological Assessment Resources, Inc.
Gottfredson, G. D., Holland, J. L., & Gottfredson, L. S. (1975). The relation of vocational aspirations and
assessments to employment reality. Journal of Vocational Behavior, 7, 135–148.
Gottfredson, L. S. (1978). An analytical description of employment according to race, sex, prestige, and Holland
type of work. Journal of Vocational Behavior, 13, 210–221.
Gottfredson, L. S. (1980). Construct validity of Holland’s occupational typology in terms of prestige, census,
Department of Labor, and other classification systems. Journal of Applied Psychology, 65, 697–714.
Gottfredson, L. S. (1981). Circumscription and compromise: A developmental theory of occupational
aspirations. Journal of Counseling Psychology, 28, 545–579.
Gottfredson, L. S. (1996). Gottfredson’s theory of circumscription and compromise. In D. Brown & L. Brooks
(Eds.), Career choices and development (3rd ed., pp. 179–232). San Francisco, CA: Jossey-Bass.
Harmon, L. W. (1996). Lost in space: A response to The spherical representation of vocational interests. Journal
of Vocational Behavior, 48, 53–58.
Harmon, L. W., Hansen, J. C., Borgen, F. H., & Hammer, A. L. (1994). Strong Interest Inventory applications and
technical guide. Stanford, CA: Stanford University Press.
Helwig, A. A. (2002). Sex and developmental differences by complexity of functions of occupational aspirations of
school children across ten years. Psychological Reports, 90, 597–605.
Holland, J. L. (1958). A personality inventory employing occupational titles. Journal of Applied Psychology, 42,
336–342.
Holland, J. L. (1959). A theory of occupational choice. Journal of Counseling Psychology, 6, 35–45.
Holland, J. L. (1976). Vocational preferences. In M. D. Dunnette (Ed.), Handbook of industrial and organizational
psychology. Chicago: Rand McNalley.
Holland, J. L. (1985). Vocational Preference Inventory manual (1985 ed.). Odessa, FL: Psychological Assessment
Resources, Inc.
Holland, J. L. (1996). Exploring careers with a typology: What we have learned and some new directions.
American Psychologist, 51, 397–406.
Holland, J. L. (1997). Making vocational choices: A theory of vocational personalities and work environments (3rd
ed.). Odessa, FL: Psychological Assessment Resources, Inc.
Holland, J. L., & Raymond, J. R. (1986). In W. B. Walsh & S. H. Osipow (Eds.), Advances in vocational
psychology. The assessment of interest (Vol. 1, pp. 55–82). Hillsdale, NJ: Lawrence Erlbaum Associates.
Holland, J. L., Whitney, D. R., Cole, N. S., & Richards, Jr., J. M. (1969). An empirical occupational classification
derived from a theory of personality and intended for practice and research (ACT Research Report No. 29).
Iowa City, IA: American College Testing.
Hubert, L., & Arabie, P. (1987). Evaluating order hypotheses within proximity matrices. Psychological Bulletin,
102, 172–178.
Hubert, L., Arabie, P., & Meulman, J. (1997). Linear and circular unidimensional scaling for symmetric
proximity matrices. British Journal of Mathematical and Statistical Psychology, 50, 253–284.
JIST Work, Inc. (1997). The enhanced occupational outlook handbook. Indianapolis, IN: JIST Works, Inc.
Kruskal, J. B., & Wish, M. (1978). Multidimensional scaling. Newbury Park, CA: Sage.
Kuder, F. (1970). Some principles of interest measurement. Educational and Psychological Measurement, 30,
205–226.
Larson, L. M., Rottinghaus, P. J., & Borgen, F. H. (2002). Meta-analyses of big six interests and big five
personality factors. Journal of Vocational Behavior, 61, 217–239.
Lippa, R. (1998). Gender-related individual differences and the structure of vocational interests: The importance
of the People-Things dimension. Journal of Personality and Social Psychology, 74, 996–1009.
Lubinski, D. (2000). Scientific and social significance of assessing individual differences: Sinking shafts at a few
critical points. Annual Review of Psychology, 51, 405–444.
McCloy, R., Waug, G., Medsker, G., Wall, J., Rivkin, & Lewis, P. (1999). Determining the Occupational
Reinforcer Patterns for O*NET Occupational Units. O*NET Consortium.
Mount, M. K., Barrick, M. R., Scullen, S. M., & Rounds, J. (2005). Higher order dimensions of the big five
personality traits and the big six vocational interest types. Personnel Psychology, 58, 447–478.
22 C.-P. Deng et al. / Journal of Vocational Behavior 71 (2007) 1–22

O*NET Resource Center (2003). The O*NET analyst database. O*NET Consortium. Available from http://
www.onetcenter.org/database.html#archive.
Oswald, F., Campbell, J., McCloy, R., Rivkin, D., & Lewis, P. (1999). Stratifying Occupational Units by Specific
Vocational Preparation. O*NET Consortium.
Paunonen, S. V., & Jackson, D. N. (2000). What is beyond the big five? Plenty! Journal of Personality, 68,
821–835.
Prediger, D. J. (1982). Dimensions underlying Holland’s hexagon: Missing link between interests and
occupations? Journal of Vocational Behavior, 21, 259–287.
Prediger, D. J., & Swaney, K. B. (2004). Work task dimensions underlying the World of Work: Research results
for diverse occupational databases. Journal of Career Assessment, 12, 440–459.
Rounds, J. (1995). Vocational interests: Evaluating structural hypotheses. In D. Lubinski & R. V. Dawis (Eds.),
Assessing individual differences in human behavior: New concepts, methods, and findings (pp. 177–232). Palo
Alto, CA: Davies-Black Publishing.
Rounds, J., & Day, S. X. (1999). Describing, evaluating, and creating vocational interest structures. In M. L.
Savickas & A. R. Spokane (Eds.), Vocational interests: Meaning, measurement, and counseling use
(pp. 103–133). Pato Alto, CA: Davies-Black.
Rounds, J., Smith, T., Hubert, L., Lewis, P., & Rivkin, D. (1999). Development of occupational interest profiles for
O*NET. O*NET Consortium.
Rounds, J., & Tracey, T. J. (1993). Prediger’s dimensional representation of Holland’s RIASEC circumplex.
Journal of Applied Psychology, 78, 875–890.
Rounds, J., & Tracey, T. J. (1996). Cross-cultural structure equivalence of RIASEC models and measures.
Journal of Counseling Psychology, 43, 310–329.
Rounds, J. B., & Zevon, M. A. (1983). Multidimensional scaling research in vocational psychology. Applied
Psychological Measurement, 7, 491–510.
Saucier, G., & Goldberg, L. R. (1996). The language of personality: Lexical perspectives on the five-factor model.
In J. S. Wiggins (Ed.), The five factor model of personality: Theoretical perspectives (pp. 21–50). New York:
Guilford.
Strong, E. K. (1943). Vocational interests of men and women. Stanford, CA: Stanford University Press.
Tracey, T. J., & Rounds, J. (1992). Evaluating the RIASEC circumplex using high-point codes. Journal of
Vocational Behavior, 41, 295–311.
Tracey, T. J., & Rounds, J. (1993). Evaluating Holland’s and Gati’s vocational-interest models: A structural
meta-analysis. Psychological Bulletin, 113, 229–246.
Tracey, T. J. G. (1997). The structure of interests and self-efficacy expectations: An expanded examination of the
spherical model of interests. Journal of Counseling Psychology, 44, 32–43.
Tracey, T. J. G., & Rounds, J. (1995). The arbitrary nature of Holland’s RIASEC types: A concentric-circles
structure. Journal of Counseling Psychology, 42, 431–439.
Tracey, T. J. G., & Rounds, J. (1996). The spherical representation of vocational interests. Journal of Vocational
Behavior, 48, 3–41.
Trapnell, P. D. (1989). Structural validity in the measurement of Holland’s vocational typology: A measure of
Holland’s types scaled to an explicit circumplex model. Unpublished master’s thesis, University of British
Columbia.
U.S. Department of Labor. (1991). Dictionary of occupational titles (Rev. 4th ed.). Washington DC: U.S.
Government Printing Office.

View publication stats