Abstract
John Holland’s theory of career orientations advises people to select careers that are congruent with their personalities. Similarly, self-concordance theory, based in self-determination theory, advises people to select personal goals that match their autonomous interests and identifications. We compared the predictive efficacy of the two theories in two studies of undergraduates, using the six career areas of Holland theory (RIASEC: realistic, investigative, artistic, social, enterprising, and conventional) as a common base. Multilevel logit modeling in Study 1 showed that both the Holland score and an aggregate self-concordance score predicted independent variance in the outcome variable, current career choices. These effects were replicated in Study 2. Supplementary analyses showed that the identified motivation subscale was the primary source of these effects. Thus, career counselors may want to consider assessing students’ self-concordance for the six RIASEC domains, in particular their levels of identified motivation for those domains, in addition to assessing their Holland codes.
John Holland’s (1959, 1997) career assessment theory and system “pervades career counseling research and practice” (Nauta, 2010, p. 11). The system assesses peoples’ work interests and aptitudes within six basic areas: realistic, investigative, artistic, social, enterprising, and conventional (RIASEC). People receive feedback from this assessment in the form of a Holland Code, which alerts them to their top three interest domains. Counselors help people to interpret this information and encourage them to consider careers known to match or mirror that particular code. The rationale is that congruence between worker personality type and work environment type affects many important outcomes including job satisfaction, job retention, and job performance (Armstrong, Day, McVay, & Rounds, 2008; Dawis & Lofquist, 1984; Holland, 1997).
Within the framework of self-determination theory (SDT; Deci & Ryan, 1985; Ryan & Deci, 2017), a similar line of research has been conducted by Sheldon and colleagues, examining the match between the personal goals that people consciously endorse and their own deeper or more implicit personality potentials (Koestner, Lekes, Powers, & Chicoine, 2002; Sheldon, 2014; Sheldon & Elliot, 1999). This research shows that congruence between broader personality dispositions and conscious goal selections has important implications for goal attainment and well-being.
Our research brought these two theoretical perspectives together for the first time. Our purpose was to compare the two theories (and their corresponding measures) as to their ability to predict students’ current career preferences. For the sake of theoretical parsimony (Chang, 1998), it is important to compare similar theories to each other when possible, to determine whether they refer to distinct versus redundant psychological processes. We consider each theory in more detail, below.
The Holland Perspective
The Holland system is based upon a typological conception of personality. According to Holland (1959, 1997), over time people develop “habitual or preferred methods of dealing with environmental tasks” (Holland, 1959, p. 35). Each of the six vocational types reflects a distinct constellation of interests, skills, and dispositions, although the six types can coexist and overlap with each other. Research has shown meaningful convergence between the six types and other personality systems including the Big Five traits (Costa, McCrae, & Holland, 1984; Tokar & Swanson, 1995), the Meyers-Briggs types (Armstrong et al., 2008), and the 16 Personality Factors (Cattell, Eber, & Tatsuoka, 1970). For example, meta-analyses show that extraversion is associated with both the social and enterprising types, openness is associated with both the artistic and investigative types, and agreeableness is associated with the social type (Barrick, Mount, & Gupta, 2003; Larson, Rottinghaus, & Borgen, 2002).
Research in the Holland tradition has also categorized a wide variety of different occupations and vocations in terms of the RIASEC system, providing a common framework within which to locate both personalities and occupations and to evaluate the degree of match between them. For example, realistic careers include cook, farmer, fire fighter, and electrician; investigative careers include mathematician, psychiatrist, geneticist, and biomedical engineer; artistic careers include actor, craft artist, graphic designer, and journalist; social careers include nurse, psychologist, dental hygienist, and physical therapist; enterprising careers include urban planner, lawyer, sales manager, and human resources representative; and conventional careers include librarian, statistician, logistics analyst, and accountant.
The Self-Concordance Perspective
The Holland theory notion that certain occupations, goals, or motives are more congruent or self-concordant for a given person is also mirrored in the contemporary motivation literature. According to Sheldon and colleagues (Sheldon & Elliot, 1999; Sheldon & Kasser, 1998; Sheldon, 2009), when people select idiographic goals for themselves, they do not always do a good job. This is because self-stated goals are in part an expression of explicit conscious beliefs and processes (Emmons, 1989). These operate in what is now termed “System 2” (Kahneman, 2011; Sheldon, 2014), the domain of self-concepts and self-presentations. As such, peoples’ conscious goals may not accurately represent deeper inclinations or preferences within what is now termed “System 1,” the domain of implicit preferences and automatic behavioral orientations. And indeed, mismatches between these two domains have been shown to predict lower well-being and performance in nonconcordant individuals (Brunstein, Schultheiss, & Grassman, 1998; Sheldon, Prentice, Halusic, & Schuler, 2015). In Socrates’s words, it is important to try to “know thyself,” in part, so one can make more self-concordant and satisfying goal choices in life.
Recent research indicates that merely providing participants with the opportunity to rate their potential motivations for “candidate goals,” prior to selecting a smaller subset of those goals for actual pursuit, can promote more adaptive goal selection. Specifically, Sheldon, Prentice, and Osin (2019) conducted experiments showing that participants randomly assigned to prerate their possible motivations for candidate goals subsequently selected more growth, intimacy, and community-oriented goals for pursuit. These are examples of “intrinsic” goals which are known to be more generally conducive to health and well-being (Kasser, 2002). Participants who did not get a chance to consider their motivations in advance were more likely to select money, appearance, or status goals, “extrinsic” goals known to be generally less conducive to health and well-being. Sheldon et al. (2018) argued that prompting people to consider their motivations for potential choices can provide them with concrete information about which choices would be most enjoyable and meaningful to pursue, aiding in the choice process.
Linking the Two Theories
In sum, the RIASEC counseling methodology (based in Holland theory) and the goal self-concordance approach (based in SDT) converge in the assumption that it is beneficial for people to pursue goals or career initiatives that are congruent with their personality orientations. In the RIASEC tradition, those who pursue career types that are congruent with their Holland codes tend to perform better and derive greater satisfaction (Nye, Su, Rounds, & Drasgow, 2012). Similarly, in the SDT tradition, those who pursue personal goals congruent with their interests and identifications perform and have greater well-being (Sheldon, 2014; Sheldon & Elliot, 1998).
Surprisingly, the Holland theory and SDT approaches have never been examined together by researchers (a search in November 2018 combining the two terms yielded 0 hits). We believe such a comparison is overdue, given the converging goals and assumptions of the theories and their long-standing research traditions. Scientific advancement can often be forwarded when two theories that have been developed in isolation, but which address similar issues, are brought together within a single study or set of studies (Hogg, Terry, & White, 1995; Marsh et al., 2019; Nigg, Allegrante, & Ory, 2002). This enables researchers to evaluate both theories in a new way, by evaluating their predictive and incremental validity with respect to each other (Chang, 1998; Vargas-Salfate, Paez, Liu, Pratto, & de Zúñiga, 2018). Direct comparisons of competing theories can help to prune redundant theories or concepts and can also help to identify “jingle jangle” fallacies wherein the same concepts and processes are given different names by different research groups (Marsh, Craven, Hinkley, & Debus, 2003).
As our dependent measure for this comparison, we focused on predicting uncommitted students’ current selections from among a range of career areas. This was in keeping with the Sheldon et al. (2018) self-concordance research, which focused on predicting participants’ current goal choices from among a range of candidate goals. Notably, relatively little research in the Holland tradition has focused on momentary career choice as a dependent measure, despite the fact that counseling practice often focuses on helping clients in making this choice (but see Betz, Borgan, & Harmon, 2006; Rottinghaus, Betz, & Borgen, 2003; Tracey & Hopkins, 2001). That is, research studies in this area have not focused on predicting undecided participants’ current career preferences at a particular moment in time. Instead, studies have typically focused on person–career congruence, a fit index which is typically used as a predictor of favorable job outcomes such as greater career self-efficacy, career goal stability and persistence, and job satisfaction and performance (Betz, Harmon, & Borgen, 1996; Nauta, 2010; Nye et al., 2012). In these types of studies, the selection has already been made, that is, the person is already working within a particular job or context. Thus, the outcomes of choice are what is being studied rather than choice itself.
Our guiding study hypothesis was that the Holland and SDT measures would each account for independent variance in current career selections. Such a hypothesis is not necessarily obvious based on an examination of the two theories and their associated measures. This is because the two theories focus on somewhat similar motivational processes, revolving around the concept of intrinsic motivation. Our guiding hypothesis is based on the observation that nevertheless, there remain some potentially important differences between the two theories. To illustrate, we consider the primary measurement strategies of each theory.
Many different measures of the RIASEC model have been proposed and used over the decades. Shared in common by all Holland measures is that they present participants with a list of specific activities, asking whether they would like to do them versus would dislike or be indifferent to doing them. Examples of “realistic” activities include “repair cars,” “build things,” and “work with tools.” Examples of “investigative” activities include “work on a scientific project” and “develop a research study.” Examples of “artistic” activities include “write novels or plays” and “play a musical instrument.” Examples of “social” activities include “work as a volunteer” and “take a human relations course.” Examples of “enterprising” activities include “promote a product” and “start my own business.” Examples of “conventional” activities include “work in an office” and “organize meetings.” As can be seen, these activities are quite concrete and the items reference only “liking.” Aggregate count scores are computed for each domain by summing the number of “like” ratings for work areas of a similar type.
Notably, from the perspective of SDT, the Holland measure is a measure of intrinsic motivation. According to SDT, intrinsically motivated behaviors are enacted because one likes doing them, enjoys doing them, and is challenged by doing them (Deci & Ryan, 1985; Ryan & Deci, 2017); they are inherently rewarding to perform. However, SDT has also moved considerably beyond the concept of simple intrinsic motivation, to distinguish multiple additional types of motivation, each of which can be located at a different position on a relative autonomy continuum (Ryan & Connell, 1989; Ryan & Deci, 2017; Sheldon, 2014; Sheldon, Osin, Gordeeva, Suchkov, & Sychev, 2017). These motivations range from very controlled or nonautonomous motivation at one end to fully autonomous motivation at the other end, with gradations in between. In order, the six motivations are amotivation (“I don’t know why I keep doing it”), external motivation (“I do it for the rewards gained or punishments avoided”), introjected motivation (“I do it to avoid guilt or shame”), self-esteem motivation (“I do it to bolster my self-esteem”), identified motivation (“I do it because it is important and meaningful, even when it is not enjoyable”), and intrinsic motivation (“I do it because I enjoy and choose it”). Self-concordance is operationally defined as the extent to which goals are pursued for autonomous reasons rather than controlled reasons, as computed by a formula described below. Higher self-concordance scores indicate that the set of goals as a whole feels more internalized into the self, suggesting that those goals likely better represent the person’s deeper or more implicit personality preferences and potentials.
In sum, the SDT methodology provides six predictor variables, including intrinsic motivation, which can be aligned on a relative autonomy continuum. The Holland methodology provides a single predictor variable, closely akin to intrinsic motivation. Given their overlap, it seemed possible that Holland measure effects might be accounted for intrinsic motivation measure effects. Our guiding hypothesis made the assumption that this would not be the case because the Holland measure provides a more concrete and detailed assessment of participants’ domain affinities, compared to the abstract domain focus of the intrinsic motivation measure. For example, the realistic domain measure used in Study 1 assesses liking of eight highly specific activities including test the quality of parts before shipment, lay brick or tile, work on an offshore oil-drilling rig, assemble electronic parts, operate a grinding machine in a factory, fix a broken faucet, assemble products in a factory, and install flooring in houses. This level of detail should provide a robust measure which supplies unique predictive variance.
We tested our guiding hypothesis by comparing the Holland measure against an aggregate self-concordance measure, the quantity employed by Sheldon (2014) and colleagues in most past research. We also tested the hypothesis in a second way by examining the six specific subtypes of SDT motivation separately rather than in aggregate. In both types of test, we expected the Holland measure to explain unique variance in the current career choice outcome, for the reasons explained above. We also expected that SDT’s two autonomous motivations, that is, intrinsic motivation and identified motivation, would have the most robust relations with the outcome. This was because the choice measure, described below, was designed to measure what the participant wants to do (autonomous motivation), not what they feel they have to do (controlled motivation).
Study 1
Method
Participants and Procedure
Participants were 246 students from undergraduate psychology courses at the University of Missouri who participated in exchange for extra credit or research credit. Ages ranged from 18 to 25 (M = 19.56, SD = 1.36), and 81% of participants were female. Eighty-one percent of the participants were Caucasian, 6% were African American, 7% were Asian American, 3% were Hispanic/Latino, and 3% were “Other.” Participants completed the survey online. Materials for the study were presented as a part of a larger questionnaire.
Participants completed a Holland Occupational Themes assessment, presented first to help them become familiar with the six career domains. Then, they completed 41 filler items, to help remove their former responses from memory. Participants were then presented with a shorthand description of each of the six occupational domains, prior to completing a self-concordance assessment specific to that domain. The domain modules were presented in the following order: social, enterprising, conventional, artistic, realistic, and investigative. Finally, participants were asked to choose three of six domains that they currently found most appealing.
Measures
Holland occupational orientations
To measure the six RIASEC orientations, we used the research instrument developed by Liao, Armstrong, and Rounds (2008). This instrument was explicitly designed to for usage in the public domain, as an alternative to commercial instruments which can limit the range of questions investigated in empirical research (Goldberg, 1999), and it has attracted considerable research use (Bai & Liao, 2018; Fouad, Singh, Cappaert, Chang, & Wan, 2016). The scale consists of 48 items that describe different work tasks, with items rated on a scale from 1 (greatly dislike) to 5 (greatly enjoy). A separate work interest score is calculated for each domain. Some sample items are as follows: “Give career guidance to people” (social), “Do research on plants or animals” (investigative), and “Assemble electronic parts” (realistic). αs for the six domains ranged from .78 to .91.
Self-concordance
To measure self-concordance, we adapted the 24-item “comprehensive relative autonomy index” (C-RAI) developed by Sheldon, Osin, Gordeeva, Suchkov, and Sychev (2017). The C-RAI was derived from a thorough content analysis of the complex RAI literature, an analysis which identified all prototypes of shared meaning across the different existing measures of the different motivational subscales. These prototypes were then turned into items, meaning that the C-RAI combines all relevant domain content within a single measure containing six subscales. In Study 1, we used a shortened 12-item measure to reduce item burden, since the measure need to be administered 6 times, within each of the six RIASEC domains. The measure was shortened by combining two related concepts within slightly longer items (e.g., instead of the 2 items of “because it is pleasurable” and “because it is enjoyable,” the single item “because it is pleasurable and enjoyable” was constructed). This procedure was practical because the C-RAI consists of very short statements which can be combined into longer items when desired (Sheldon et al., 2017).
As an introduction to the task, participants read: “Career Choices! Now we will ask you to rate your possible motivations for pursuing each of 6 career areas. Before you rate each area, we will describe it and give examples.” Next they read: “AREA 1: Helping. Helpers enjoy doing things for others. Example careers are teachers, coaches, advisers or counselors.” Then participants read: “People do things for many different reasons. Below are 12 statements about why you might like helping people. Please rate the accuracy of each statement. Many of them may apply.” After making the 12 ratings, they were then presented with Area 2, then Area 3, and so on. The labels and descriptions provided for the six RIASEC career areas were taken from the Guide to the Holland Code provided by the University of Missouri to students undergoing counseling at the career center. For example, the RIASEC Social domain is therein described as “Helpers: people who like to work with people to enlighten, inform, help, train, or cure them or are skilled with words.”
The C-RAI consists of six subscales, three on the autonomous and three on the controlled side of the relative autonomy continuum. As listed and illustrated above, these included intrinsic reasons, identified reasons, and self-esteem reasons, the three autonomous reasons. They also included introjected reasons, external reasons, and amotivated reasons, the three controlled reasons. A 1 (not at all for this reason) to 5 (very much for this reason) scale was employed. Six aggregate self-concordance scores were computed (Ryan & Connell, 1989; Sheldon & Elliot, 2000; Sheldon, Ryan, Rawsthorne, & Ilardi, 1997), one for each RIASEC domain, by first subtracting the total motivation score from each subscale score to control for quantity (independent of quality) of motivation and then by adding the identified, intrinsic, and self-esteem reason ratings and subtracting the introjected, amotivated, and extrinsic ratings in each case (see Sheldon et al., 2017, for new psychometric support for this typical scoring procedure within SDT). α for the aggregated self-concordance measures ranged from .63 to .84. As noted earlier, we also intended to use the six subscales separately, in order to evaluate which ones contribute the most unique variance in the prediction of current career choices.
Current career choice outcome measure
As a third task, participants read: “If you were going to decide right now which of these areas were MOST appealing to you, which ones would you pick? Please click and drag the 3 choices that look attractive, for whatever reason, over to the box on the right.” The six domains were listed on the left, and participants clicked and dragged three of them into a box on the right. Thus, for each participant, each of the six RIASEC domains received a score of either 0 (not chosen) or 1 (chosen). This was the primary dependent measure for our analyses.
We asked participants to pick three of six domains for three reasons. First, this mirrors the procedure of Sheldon et al. (2018), which asked participants to select three of six candidate goals. Second, we wanted to create an engaging task in which participants were not being asked to make a single selection just now; rather, they were being asked to narrow the possibilities down, from six to three. Third, we wished to maximize the variance in the outcome variable, by giving each domain a 50% chance of being selected.
Results
Table 1 provides descriptive statistics and correlations for the six Holland and the six self-concordance scores. As can be seen, the average correlation between corresponding Holland and aggregate self-concordance scores was r = .14, with four of the six reaching significance. The relatively low average correlation indicates that the two assessment approaches are assessing distinct constructs, to a considerable extent.
Means, SDs, and Correlations for Predictors, Study 1.
Note. For correlations ≥.14, p < .05. For correlations ≥.16, p < .01. HOL = Holland scale ratings; SC = self-concordance assessment of each domain of Holland careers.
Again, our primary study hypothesis was that the Holland and SDT measures would account for independent variance in career selections. Because each participant had supplied data from each of the six RIASEC domains, we focused on the career level of analysis, not the person level of analysis. There were thus 1,476 cases for the main analysis (246 Participants × 6 Domains for Each Participant). Because of the repeated measures design, dependence between scores needed to be addressed. We used a multilevel modeling procedure with maximum likelihood estimation. Participants’ IDs (N = 246) and career domains (N = 6) were Level 2 variables, and the Holland scores, self-concordance scores, and career choice scores were Level 1 variables (N = 1,476). Specifically, we used the lme4 package in R to analyze the data (Bates, Mächler, Bolker, & Walker, 2015; R Development Core Team, 2008). Due to the dichotomous dependent variables (selected or not), we tested logistic models of odds ratios, meaning that the coefficients are unstandardized. However, all variables were standardized before being included in the models.
We wished to test random intercepts models in which Level 2 differences are merely controlled, so that the analysis can focus only on Level 1. For example, if participant X’s Holland social score is 65 and participant Y’s Holland social score is 45, this between-subject difference would be removed, and if the average Social domain score across participants is higher than the average realistic domain score (as was found in our data), this between-domain difference would be removed. This would render all 1,476 cases equivalent at Level 1. Before testing random intercept models, we first tested a maximal model in order to determine whether slopes vary across people (i.e., does the effect of the Holland Social domain score upon the selection of the Social career option varies across participants?). In the maximal model, both intercepts and slopes between each nested variable are allowed to vary. In our case, we allowed for unique slopes and unique intercepts for each participant and for each domain. Additionally, this model allowed for all the intercepts and slopes to be potentially correlated (Matuschek, Kliegl, Vasishth, Baayen, & Bates, 2017). As expected, however, this maximal model did not provide a good fit for the data because between-person variations in slopes were very small and within-person slopes were large. Thus, we proceeded to the simplified random intercepts models. Additionally, we compared fit statistics for the maximal model (AIC = 676.45 and BIC = 740.99) and Model 1 (AIC = 662.89 and BIC = 684.41), and they suggested that Model 1 had better fit for our data.
In Model 1, we simply entered the Holland measure and the aggregate self-concordance measure together, controlling for participant age, ethnicity, and gender at Level 2. The results suggested that both self-concordance and Holland score were significant positive predictors of participant career domain choice (see Model 1 in Table 2). This means that with every increase of one point on the self-concordance scale for a domain, the chances for choosing a career in a domain increased by 65%. Similarly, with every point increase on the Holland Scale within a domain, the chances for choosing the domain increased by 80%. The control variables of participant sex, age, and ethnicity had no significant effects.
Results of Multilevel Logistic Regression Analyses for Study 1.
Note. All coefficients are standardized. Bolded coefficients are significant.
In Model 2, we examined the six self-concordance motivation subscales separately rather than combining them into a single RAI (see Model 2 in Table 2). Consistent with expectations, only two subscales were significant, namely, the two autonomous motivations of intrinsic motivation (“because it is interesting and enjoyable”) and identified motivation (“because it is important and meaningful”). Again, demographic variables had no effects.
In Model 3, we inserted the Holland measure into Model 2, expecting it to predict significant variance in choice. This was indeed the case (see Table 2); in this third model, intrinsic motivation and identified motivation continued to predict significant variance.
Study 2
The first goal of Study 2 was to replicate the Study 1 findings. Would both the Holland and the self-concordance scores predict concurrent career selections? Another goal was to generalize the patterns to a somewhat different sample. Rather than focusing on generic undergraduates, Study 2 focused on a sample of self-selected students who were actively engaged in noncredit career exploration processes at the career center, during the period of the study. Thus, Study 2 could illustrate the generalizability of the Study 1 effects to highly motivated career explorers and show that the processes have direct relevance within career counseling settings.
Method
Participants and Procedure
Participants were members of career exploration classes taught at the University of Missouri career center, who were administered the Holland measure (described below) within an early class session. The self-concordance and career selection measures were also administered, via an online survey invitation e-mailed to all class participants. Ninety-two students (72% Caucasian, 12% African American, 5% Asian American, 3% Hispanic/Latino, and 8% “Other”) completed the survey, 30 men and 62 women; they ranged from 18 to 36 in age, with a median age of 18.
Measures
Holland occupational themes
To measure the six RIASEC domains, we used the Work Interest Scales of the Focus 2, which is administered to all career exploration classes at the University of Missouri (MU) career center. This is a well-validated RIASEC measure provided by the Career Dimensions Corporation, which has been in business since 1967; the Focus 2 is the current update. As with the measure used in Study 1, the focus provides participants with concrete work examples and asks them to check the ones they like. We received scale-level scores, not raw data; thus, we were unable to calculate α coefficients. However, the company reports good factor structures and high reliabilities (exceeding .85) for the Focus 2 subscales (Career Dimensions, Inc., 2010), findings which were confirmed by Tirpak and Schlosser (2013). A 2018 Technical Report from Career Dimensions, Inc., indicated that the Focus 2 is used by more than 1,000 colleges, universities, and service agencies.
Self-concordance
In order to measure domain self-concordance in Study 2, we used the full 24 items C-RAI (Sheldon et al., 2017), with the same instructions and response scale/anchors as before. Cronbach’s α reliability coefficients ranged from .68 to .81. As before, six aggregate self-concordance domain scores were computed by summing the autonomous motivation items and subtracting the controlled motivation items, one for each RIASEC domain. However, we again planned to examine the six motivational subscales separately, as in Study 1.
Career choice
During the online assessment, participants completed the same current career preference measure used in Study 1, in which they dragged three choices of six into a selection box.
Results
Table 3 presents descriptive statistics and correlations for the variables. The average correlation between corresponding Holland and self-concordance scores was r = .16, with three of the six reaching significance. The relatively low average correlation (up from .14 in Study 1) again indicates that the Holland and self-concordance assessment approaches are assessing distinct constructs, to a considerable extent.
Means, SDs, and Correlations for Predictors, Study 2.
Note. For correlations ≥.21, p < .05. For correlations ≥.25, p < .01. HOL = Holland scale ratings; SC = self-concordance assessment of each domain of Holland careers.
As in Study 1, participants’ self-concordance and Holland scores were nested within participants and also within domains (given that each participant rated self-concordance and Holland measures within each of six domains). Therefore, we again used a maximum likelihood estimation multilevel model to test our hypotheses, with participants’ IDs and domains as Level 2 variables and Holland and self-concordance scores as Level 1 variables. As before, a maximal model allowing varying slopes across participants did not fit the data and thus were able to proceed to the preferred random intercept models. Again, such models simply allow researchers to treat each within-subject data point as a separate case, expanding the effective sample size from N to 6 × N.
Predicting Career Choice
Replicating the cross-sectional results of Model 1 in Study 1, the logistic multilevel model analysis showed that aggregate self-concordance and the relevant Holland score were both significant positive predictors of participant career domain choice (see Model 1 of Table 4). These results indicate that with every increase of one point on self-concordance scale, the chances for choosing a career in a domain increase by 67%, and with every increase on Holland scale of one point, the odds of career selection increased by 46%.
Results of Multilevel Logistic Regression Analyses for Study 2.
Note. All coefficients are standardized. Bolded coefficients are significant.
Next, we tested the six motivation subscales separately (see Model 2 of Table 4). Replicating the results of Study 1, intrinsic motivation and identified motivation, the two clearly autonomous motivations, were significant predictors of career choice. In addition, introjected motivation, a controlled type of motivation, emerged as a significant negative predictor of career choice. That is, in Study 2, this guilt-based motivation was negatively predictive of career preferences.
Then, we tested Model 3, created by adding the Holland measure to Model 2 (see Table 4). Intrinsic, identified, and introjected motivation continued to predict career choice even after adding the Holland score to the model (the introjection effect was negatively signed). Additionally, as expected, Holland score was a significant predictor in the model.
General Discussion
This study was the first to combine Holland vocational theory and SDT (via the concept of self-concordance) in a study of career preferences. The study was relevant because self-concordance research has recently begun to supply participants with self-concordance information before they make goal choices rather than assessing self-concordance only after goal choices have been made (as has been typical; Sheldon, Prentice, & Osin, 2019). This is similar to the Holland counseling method, which provides students with information to use before making career choices. We hypothesized that the self-concordance measures might account for additional variance in career-related outcomes because these measures draw from the full relative autonomy continuum of SDT (Ryan & Connell, 1989; Sheldon et al., 2017) rather than simply asking students whether they like exemplars of a career domain or not. However, we also hypothesized that the Holland measures would supply independent predictive variance from intrinsic motivation because the Holland measures assess liking of a domain in a more detailed and concrete way, providing participants with many specific exemplars of each domain.
To summarize the results, cross-sectional Study 1 surveyed psychology students. The study assessed participants’ Holland scores for each of the six RIASEC domains and also assessed self-concordance scores regarding “why you might pursue” each of the RIASEC career domains. Scores derived from the two assessment systems predicted independent variance in participants’ concurrent career selections. Specifically, the aggregate self-concordance score and the Holland score were both significant in our analyses. Supplementary analyses showed that the two autonomous motivations, namely, intrinsic and identified motivation, accounted for the aggregate self-concordance effect; however, the Holland measure continued to supply independent predictive variance beyond these two SDT measures. The Study 1 findings both newly validate the Holland method and also suggest that the self-concordance methods may supplement or in some ways go beyond the Holland method.
Study 2 employed students enrolled in a noncredit career exploration class. Although the study was limited by its small sample size, it was also valuable because Study 2 participants were actively engaged in a noncredit career exploration class. Study 2 replicated all of the results of Study 1. That is, both the Holland score and the aggregate self-concordance score predicted career selections. Also, the supplemental analyses again showed that intrinsic and identified motivations were positive predictors of career choice, along with the Holland measure.
Notably, in Study 2, introjected motivation was also a negative predictor of current career choice. Introjected motivation is located on the controlled side of SDT’s motivation continuum and is associated with guilt and negative affect. Thus, it is perhaps unsurprising that introjection negatively predicted career choice, especially since participants were asked to pick the choices “that look most appealing and attractive to you right now.” However, since there was no significant negative effect of introjection in the larger sample of Study 1, it is premature to draw conclusions about introjection.
We believe it is noteworthy that the identified motivation measure predicted additional variance beyond the Holland measure in both studies herein. Identified motivation involves the perception that the activity is pursued because it is meaningful and valuable. The identified motivation concept was incorporated into SDT as a second type of autonomous motivation, in order to address cases in which an activity still feels autonomously chosen, even when it may not feel pleasurable and enjoyable (Ryan & Deci, 2017). Identified motivation means doing X because it is important, even if it is not fun. Thus, career counselors looking to supplement the Holland measures may want to consider employing SDT’s identified motivation measure as well, as given in Sheldon et al. (2017). In this way, they will be able to take into account the second important facet of autonomous motivation and give voice to career domains that a counselee might be drawn to pursue, not because they will be “fun,” but because they represent meaningful and important values.
It is also noteworthy that the intrinsic motivation subscale significantly predicted career choice, in both studies. As discussed earlier, intrinsic motivation and the Holland measure are similar, in that both address what participants like to do. We hypothesized and found that the Holland measure would supply unique predictive variance in the analyses because the Holland measure assesses domain liking in a specific way, by providing participants with concrete behaviors to consider. However, the finding that SDT’s intrinsic motivation subscale remained significant in the analyses suggests that it too supplies unique predictive variance, potentially supplementing the Holland measure. This may be due to the fact that the intrinsic motivation measure uses other terms besides liking to assess intrinsic motivation, such as “interesting,” “pleasure,” “fun,” and “enjoy.” Because the intrinsic motivation more broadly assesses the range of the intrinsic motivation concept, whereas the Holland measure more broadly assesses the range of domain-relevant behaviors, both may provide important information for career counselees.
These studies have a number of limitations including the fact that they were conducted only on college students; they were conducted at a single university in the U.S. Midwestern region; and the “career choices” outcome variable had no concrete weight for students, perhaps appearing to them as just another career preference measure. Another major limitation of these research is the sample size of Study 2. As we wanted to recruit students who were signed up for an actual career exploration course, we were limited to the number of students enrolled in this course which resulted in a smaller number of participants than desired. However, by using MLM statistical techniques, we were able to gain better power for our statistical analyses partially mitigating this limitation. Future research could attempt to extend the findings to high school or adult populations, within other geographic or cultural regions, using outcome measures that have more impact (e.g., students’ actual choice of majors or careers). Future research could also include other outcome measures such as peoples’ performance levels in or enjoyment of the majors or careers they choose. If they have chosen “wrongly,” might exposure to self-concordance information help them to modify or correct their career paths?
In conclusion, we hope that readers agree that the self-concordance assessment methodology provides a promising alternative or complement to conventional career assessment techniques. Applying this method allows for the application of SDT (Ryan & Deci, 2017), a theory which provides many conceptual tools for analyzing the “quality” of peoples’ motivations. We hope other researchers will further develop this promising linkage.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project has been partially funded by the Russian Academic Excellence project “5–100.”
