305
British Journal of Educational Psychology (2013), 83, 305–328 © 2012 The British Psychological Society www.wileyonlinelibrary.com
Vocational interests of intellectually gifted and highly achieving young adults ¨ 2 , and Gabriel Nagy2 Miriam Vock1 ∗ , Olaf Koller 1 2
University of Potsdam, Germany Leibniz Institute for Science and Mathematics Education, Kiel, Germany Background. Vocational interests play a central role in the vocational decision-making process and are decisive for the later job satisfaction and vocational success. Based on Ackerman’s (1996) notion of trait complexes, specific interest profiles of gifted high-school graduates can be expected. Aims. Vocational interests of gifted and highly achieving adolescents were compared to those of their less intelligent/achieving peers according to Holland’s (1997) RIASEC model. Further, the impact of intelligence and achievement on interests were analysed while statistically controlling for potentially influencing variables. Changes in interests over time were investigated. Sample. N = 4,694 German students (age: M = 19.5, SD = .80; 54.6% females) ¨ participated in the study (TOSCA; Koller, Watermann, Trautwein, & L¨udtke, 2004). Method. Interests were assessed in participants’ final year at school and again 2 years later (N = 2,318). Results. Gifted participants reported stronger investigative and realistic interests, but lower social interests than less intelligent participants. Highly achieving participants reported higher investigative and (in wave 2) higher artistic interests. Considerable gender differences were found: gifted girls had a flat interest profile, while gifted boys had pronounced realistic and investigative and low social interests. Multilevel multiple regression analyses predicting interests by intelligence and school achievement revealed stable interest profiles. Beyond a strong gender effect, intelligence and school achievement each contributed substantially to the prediction of vocational interests. Conclusions. At the time around graduation from high school, gifted young adults show stable interest profiles, which strongly differ between gender and intelligence groups. These differences are relevant for programmes for the gifted and for vocational counselling.
Students who are very intelligent or who excel academically at school seem to have an important prerequisite for a successful vocational development, because their
∗ Correspondence should be addressed to Miriam Vock, Universit¨at Potsdam, Karl-Liebknecht-Str. 24/25, 14476 Potsdam, Germany (e-mail:
[email protected]). DOI:10.1111/j.2044-8279.2011.02063.x
306
Miriam Vock et al.
intellectual abilities allow them to choose from many vocational options – including also very ambitious careers. Intellectual abilities have been found to correlate at least moderately positively with vocational achievements in complex jobs (Gottfredson, 1997; Ones, Viswesvaran, & Dilchert, 2005, Schmidt & Hunter, 2004). Having outstanding abilities, of course, is not a sufficient condition for a positive and successful vocational development. Rather, it seems essential to find a vocation that matches with personal interests, needs, and abilities (Lofquist & Dawis, 1991; Lubinski & Benbow, 2000). Particularly, vocational interests play a central role in the vocational decision-making process and are decisive for the later satisfaction with the chosen job and vocational success (Dawis, 1996; Gottfredson, 1996). Some researchers assumed that intellectually gifted and highly achieving students do not only differ from their peers in their intellectual abilities, but that they also differ in their vocational preferences (e.g., Stapf, 2003; Rice, 1985). Already Lewis Terman’s gifted group, which had been studied extensively during the last century, exhibited specific vocational preferences (Strong, 1945). Yet, more recent empirical findings on the vocational interests of gifted and highly achieving students are rather sparse (Sparfeldt, 2006); and reliable data on vocational interests of these groups at different points in time shortly before and after graduation from high school – when vocational interests are highly relevant for vocational decision-making – are still missing. Vocational interests are still in flux during adolescence, and do not reach stability until early adulthood (Lubinski & Benbow, 2000). Interest profiles can therefore be expected to change during late adolescence, when students have to make first choices with direct relevance for their later vocation (e.g., selection of courses, majors). In the present study, we first describe the vocational interests of different ability and achievement groups shortly before and after graduation from high school; we second analyse the impact of intelligence and achievement on vocational interests in a unrestricted sample while statistically controlling for further possibly influencing variables, and we third analyse the change and stability of young adults’ vocational interests over time. Theoretical background Vocational interests, cognitive abilities, and academic achievement Holland’s (1997) RIASEC model of vocational interests is probably the most prominent model of vocational interests, and there is considerable empirical evidence to support it (e.g., a meta-analysis by Tracey & Rounds, 1993). People are characterized in terms of their interests in six interest dimensions, namely realistic (R), investigative (I), artistic (A), social (S), enterprising (E), and conventional (C), arranged in a hexagonal structure. Interest dimensions are located next to those to which they are most similar and opposite to those to which they are most dissimilar. Holland’s theory assumes that vocational interests develop during childhood and adolescence but can change throughout adult life, depending on environmental influences. Generally, empirical research shows that the stability of vocational interests increases with age, and that vocational interests remain remarkably stable from adolescence on (Swanson, 1999); there are, however, considerable individual differences in this stability (e.g., Lubinski, Benbow, & Ryan, 1995). Gender-specific profiles of vocational interests are a stable finding (Bergmann & Eder, 2005; Holland, 1997; Lippa, 1998). Typically, girls and women are more interested in social and artistic activities, whereas boys and men are more interested in realistic, investigative, and conventional activities.
Vocational interests of the gifted
307
While early measures of vocational interests were designed to be uncorrelated with intelligence (e.g., Strong, 1927), Holland’s (1959) original conception explicitly integrated systematic relationships with intellectual abilities. Hence, Holland did not conceive interests and abilities as completely independent characteristics. Therefore, it seems plausible that gifted and highly achieving students might exhibit different vocational interests than their peers. For a more recent theoretical framework, we can draw on the influential work by Ackerman (1996; Ackerman & Heggestad, 1997) on the interplay of abilities and interests. Ackerman systematically analysed the correlations between specific interests, personality characteristics, and cognitive abilities in order to distil so-called trait complexes across personality, interests, and abilities. Generally, the relationships between cognitive abilities and vocational interests are of moderate strength (|r| = .20–.35). Positive relationships are typically found between investigative interests and verbal, numerical, and spatial thinking; between realistic interests and spatial as well as mathematical abilities; and between artistic interests and verbal abilities (e.g., Ackerman, Kanfer & Goff, 1995; Randahl, 1991). Negative relationships have been found between conventional interests and mathematical abilities; between enterprising interests and cognitive abilities; and between social interests and spatial/mathematical abilities or intelligence (Ackerman & Beier, 2003; Ackerman & Heggestad, 1997). Four ability–interest–personality trait complexes could be identified: social (including social and enterprising interests, extraversion, social potency, and well-being personality traits), clerical/conventional (consisting of perceptual speed abilities; conventional interests; and control, conscientiousness, and traditionalism personality traits), science/math (including visual perception and math reasoning abilities, realistic and investigative interests) and, finally, intellectual/cultural (including crystallized intelligence and ideational fluency, artistic and investigative interests, and as personality traits openness to experience, absorption, and typical intellectual engagement) (Ackerman & Heggestad, 1997). These trait complexes may help to understand the acquisition of knowledge across lifespan. While personality and interests guide the direction of intellectual effort towards certain domains, cognitive abilities determine the complexity of accumulated knowledge. According to the relationship between interests and abilities, Ackerman proposes that ‘abilities and interests develop in tandem, such that ability level determines the probability of success in a particular task domain, and personality/interests determine the motivation for attempting the task’. (p. 243). Hence, the increasing convergence between interests and abilities can be explained by people’s different experiences in different domains. Success in a certain domain will reinforce interest in that domain; failure will often diminish interest. Enhanced interest usually leads to further engagement in a domain. Over the years, this process results in increasingly congruent interest and ability profiles, reflected in prototypical patterns of combinations between cognitive abilities and vocational interests in adulthood. Drawing on these analyses, we can expect that intellectually gifted adolescents and young adults might indeed differ systematically in their vocational interest profiles from their less intelligent peers. As they are by definition very intelligent, it is plausible that during their development they have continuously developed and intensified particularly those vocational interests that are known to be part either of the science/math and/or the intellectual/cultural trait complexes; that is, investigative interests (which are part of both trait complexes), realistic, or artistic interests (cf. Ackerman & Heggestad, 1997). Also, highly achieving students (who are not necessarily also intellectually gifted) can be expected to develop especially those interests described in the science/math and
308
Miriam Vock et al.
the intellectual/cultural trait complexes. In several German student samples, Bergmann and Eder (2005) found grades in languages to show positive correlations of a moderate magnitude with artistic interests (r = .17–.29), negligible correlations with investigative interests (r = −.06 to .04), and negative correlations with realistic interests (r = −.09 to −.19). Grades in mathematics correlated positively with investigative interests (r = .12–.23) and negligibly with realistic and artistic interests (r = .00–.05). Intellectual giftedness There is no consensus on the definition of intellectual giftedness in the literature; rather, several theoretical models coexist (Sternberg & Davidson, 2005). In all conceptions, high intelligence is a core criterion, if not the only criterion, of intellectual giftedness (e.g., Robinson, 2005; Rost, 2000). Typically, students scoring at or above the 98th percentile on a general intelligence test – that is, students with an IQ of at least 130 – are defined as ‘intellectually gifted’ (Holling, Preckel, & Vock, 2004). Intellectually gifted students are often also high achievers at school. At the same time, some intellectually gifted students do not fulfil their potential (‘underachievers’), and many less intelligent students nevertheless show outstanding levels of achievement (Preckel, Holling, & Vock, 2006); gifted and high-achieving students are thus overlapping but distinct groups. This distinction is often neglected in research on giftedness, resulting in a confounding of effects. The present study aims to systematically disentangle the vocational interests of the two groups. Vocational interests of gifted and highly achieving students To date, few studies have analysed gifted adolescents’ vocational interests. In an early study, Post-Kammer and Perrone (1983) used Holland’s RIASEC model to analyse the vocations of former participants in a counselling programme for the gifted. They found that most male participants had chosen either investigative or enterprising professions, whereas most female participants had chosen either investigative or social professions. None of the respondents in this gifted sample worked in a realistic job and very few had chosen an artistic or conventional profession. In their seminal Study of Mathematically Precocious Youth (SMPY), Lubinski et al. (1995) and Schmidt, Lubinski, and Benbow (1998) analysed vocational interests according to Holland’s (1997) RIASEC model in a sample of 162 intellectually gifted students attending a special programme for the gifted. The interests of the SMPY subjects were assessed twice during adolescence. At age 13, the gifted students scored high on the scales measuring investigative and artistic vocational interests; the difference between the gifted group and the norm group was about d = .30 (cf. Sparfeldt, 2007). Fifteen years later, the gifted participants’ scores on the realistic, investigative, artistic, and social scales had increased, but their scores on the enterprising and conventional scales had dropped markedly. Their score on the enterprising scale was almost 1 SD below the norm. Thus, profiles of vocational interests were found to change substantially during adolescence. Unfortunately, no data were reported on the gifted respondents’ vocational interest profiles at graduation from high school or in the first years after school, when vocational interests are highly relevant for vocational choices. For gifted and high achieving females, findings have been mixed and it remains to be clarified whether high abilities might moderate typical gender effects. In the SMPY, gifted girls scored higher than gifted boys on artistic and social interests (d = −.89 and d = −1.04, respectively), whereas gifted boys scored higher than gifted girls on
Vocational interests of the gifted
309
realistic interests (d = .60; Schmidt et al., 1998). Mixed findings have been reported for investigative interests: whereas some older studies found no differences between gifted boys and gifted girls (Fox, Pasternak, & Peiser, 1976; Schmidt et al., 1998), another found that although intellectually gifted females reported stronger investigative interests than non-gifted females, the intensity of their investigative interest did not exceed that of non-gifted males (Sparfeldt, 2007). Most studies on gifted populations have used pre-selected samples, such as students attending special schools for the gifted or participating in special programmes (e.g., Lubinski et al., 1995; Platzer, 2002), meaning that the respondents probably also had high levels of achievement motivation. Furthermore, few of these studies have reported findings on less gifted control subjects or referred to adequate norm data, making interpretation of the findings somewhat difficult. These kinds of sample biases restrict the generalizability of results (Sparfeldt, 2006; Vock & Holling, 2007). For example, participants in the SMPY (Lubinski et al., 1995) attended a comprehensive programme for gifted students. Their mere participation in this programme may reflect certain characteristics (e.g., high achievement motivation) and social environments (e.g., supportive families or teachers). At the same time, selection for and participation in such programmes for several years is known to have strong effects on students’ attitudes, self-esteem, and abilities (Kulik, 2004; Vock, Preckel, & Holling, 2007). The vocational interests of gifted students who have attended such programmes might not be representative for the entire gifted population. Furthermore, such designs can lead to socially desirable responding. Respondents know that they are being surveyed as people who were once identified as very bright children. It seems likely that this knowledge influences their answers about interests, and that their ratings might be affected by beliefs about what an intelligent person “should” enjoy and aspire to become. A more recent study by Sparfeldt (2006, 2007) addressed many of the methodological concerns mentioned above. Sparfeldt analysed vocational interest data within a large longitudinal study on giftedness, the Marburg Giftedness Project (Rost, 2000), in which gifted students were selected from a representative sample of third year primary school students by their performance on a test of general intelligence (IQ ≥ 130, N = 107). The gifted sample and a sample of control students were followed up until adulthood. Participants did not attend a specific programme for the gifted (which might have led to the development of different interests). At age 20, the gifted group scored higher on investigative interests (d = .54) and lower on social interests (d = −.38) than the comparison group. In addition, a sample of highly achieving students was compared with a parallel sample of average-achieving students (Sparfeldt, 2006); highly achieving students reported stronger interests in the investigative (d = .44) and artistic (d = .44) domains. These results, therefore, indicate that the interest profiles of gifted students differ in some respects from those of high-achieving students. Most studies on giftedness report – primarily or exclusively – descriptive data on extreme group means. The formation of extreme groups (very intelligent vs. less intelligent) implies considerably smaller samples and, accordingly, a loss of data. Likewise, the possible effects of social factors (e.g., school track, social background) are neglected – sample sizes of gifted students are usually too small to allow for stratification (one frequent exception is a sample split by gender). Many investigations have thus failed to investigate the unique contribution of different factors to gifted students’ interests. We believe that it is necessary to take into account the gifted students’ gender, school type, and socio-economic background in order to obtain a clear picture of their
310
Miriam Vock et al.
vocational interests. Because giftedness is primarily defined by the continuous variable of intelligence, we argue that the classical analytical strategy of comparing group means should be supplemented by analyses of continuous data. Empirical findings indicate that the vocational interests of gifted students may also be moderated by the academic level and profile of the school attended. These differences may be attributable to pre-existing differences in students’ interest profiles that guided their choice of secondary school. Equally, it is possible that the specialized curriculum amplifies existing preferences. Realistic and social interests have been found to discriminate fairly well between the populations of different school types; the effects for investigative and enterprising interests are smaller (Bergmann & Eder, 2005). We assume that the choice of a school with a certain profile is guided by (among other things) the student’s interests, and that attending a school with a certain profile will intensify and consolidate those interests that are congruent with the profile. Drawing on Holland (1997), we expect students in schools specializing in science and technology – relative to students in regular high schools – to have stronger realistic and investigative interests, and students in schools specializing in economics – again relative to students in regular high schools – to have stronger enterprising and conventional interests. Aims This study has three primary objectives. The first objective of the present investigation is to analyse the vocational interests of gifted adolescents and to compare their profiles to those of adolescents with average and low intellectual abilities. We also distinguish intellectually gifted students from high-achieving students – two groups that overlap but that are not identical. Differing results for gifted in comparison to high-achieving students may have implications for (a) theoretical conceptions of giftedness and (b) the design of special programmes for the gifted and for the high achievers. Referring to the analysis of Ackerman and Heggestad (1997) on trait complexes, we hypothesize that intellectually gifted students as well as high achieving students have stronger investigative, realistic, and artistic vocational interests than less able/achieving students. We also look for gender differences within the different ability groups. It remains to be clarified to what extent intellectual giftedness might compensate for the well-known gender effect that women in representative samples generally have lower investigative interests than men. Do gifted young women have comparable investigative interests to gifted young men – or do gifted young women’s interest profiles resemble more those profiles of other women? Second, we analyse the impact of intelligence and achievement on vocational interests while statistically controlling for further, potentially influencing variables. Third, we investigate changes in interests over time. Interests are analysed at two points in time that are crucial for the decision-making process; that is, at the end of school and at the beginning of the vocational or academic post-secondary education. These research questions are approached on the basis of a data design that overcomes many of the problems involved in previous investigations. In addition to the comparison of groups (low, average, and high intelligence; low, average, and high academic achievements) conventionally used in giftedness studies, we adopt a multivariate approach with continuous data. We are thus able to jointly analyse the impact of intelligence and school achievement, controlling for several factors that might influence the development of vocational interests (i.e., gender, SES, and the profile of the school attended).
Vocational interests of the gifted
311
Further, the typical multilevel structure of these kinds of data sets is taken into account (students in different schools) in our analyses. Despite the many methodological advantages of the multivariate approach, we also report findings on group means to allow for the inspection of group-specific interest profiles and to permit comparison with the results of previous studies on gifted students. Another methodological advantage of this study was that the gifted adolescents were selected from a large representative and randomly selected sample of secondary students attending schools in Germany’s highest academic track (Gymnasium). Thus, the findings can be generalized to the population of intellectually gifted adolescents in Gymnasium schools, at least in Germany (i.e., to the vast majority of intellectually gifted students in Germany). We tracked the whole sample, meaning that we could draw on data from all respondents, representing the whole ability range, in analyses with continuous data. Moreover, the participating students did not know whether or not they had been classified as intellectually gifted, which reduced the likelihood of socially desirable responding and resultant biases. Finally, the students’ vocational interests were assessed twice, allowing the stability of profiles to be investigated and different groups to be compared on repeated occasions. Taken together, in this paper we study students’ vocational interests according to the RIASEC model at two points in time during late adolescence (as dependent variables), and intelligence as well as achievement (as independent variables). In a descriptive approach, we first report the RIASEC score means for groups in order to picture the profiles of different ability and achievement groups. These descriptive analyses are then followed by multi-level regression analyses in which we use data from the complete sample for the prediction of interests and can statistically control for potentially influencing variables and model the change of interests over time.
Method The data come from a large-scale longitudinal study entitled “Transformation of the Secondary School System and Academic Careers” (TOSCA) that was initiated in the German state of Baden-Wurttemberg in 2002 (K¨ oller et al., 2004). Specifically, the TOSCA study examines the developmental trajectories of young adults from their final year at high school through the transition to vocational training or university. Data were collected in two waves: wave 1 (spring 2002), when respondents were in their final year at high school, and wave 2 (spring 2004), when they were at university or had started a vocational training. Intelligence was only assessed in wave 1; vocational interests were assessed in both waves. Participants A total of 4,694 students (54.6% female) from 149 schools were tested in wave 1, when their mean age was 19.6 years (SD = .80). Of these, 2,318 students could be reassessed in wave 2, at a mean age of 21.5 years (SD = .86). Most of the students attended a regular Gymnasium school, representing the highest academic track in the German school system (N = 2,836); 1,858 students attended a Gymnasium school specializing in economics, science and technology, nutrition, agriculture, or social education. Fewer than 10% of the students came from immigrant families. For the first, descriptive part of the statistical analyses, we formed groups of students with very high versus average and lower intellectual abilities to permit comparisons
312
Miriam Vock et al.
between gifted, average, and less intelligent students. We classified all students who scored at or above the 98th percentile of the intelligence scale as ‘intellectually gifted’ and students who scored at or below the 2nd percentile as ‘less intelligent’. As a result, 131 students were classified as intellectually gifted and 115 students as less intelligent. In the second wave, we identified 77 gifted respondents and 28 less intelligent respondents. Further, we built a comparison group of students with average abilities by randomly selecting 2% of those students scoring at the 50th percentile. For the first wave 123 students, and for the second wave 61 students were selected for the ‘average’ group. We also defined a group of high achievers and a group of low achievers. High achievers had a grade point average (GPA) in the final school exam (Abitur) at or above the 98th percentile (wave 1: N = 151; wave 2: N = 43); low achievers had a GPA at or below the 2nd percentile (wave 1: N = 112; wave 2: N = 70). Additionally, we formed a group of students with average achievements by randomly selecting 2% of those students with achievements at the 50th percentile (wave 1: N = 131; wave 2: N = 74). In the whole sample, the correlation between GPA and intelligence was r = −.30 (p < .01), which indicates that the overlap between gifted students and high achievers was not very large. Note that subject grades were based on the conventional numeric grading scale employed in German schools, ranging from 1 to 6, with higher values indicating lower achievement. Therefore, negative correlation coefficients between GPA and intelligence represent a positive association between intelligence and achievement. Instruments Intelligence Intelligence (verbal and non-verbal reasoning) was assessed by the highly g-loaded Figure Analogies and Verbal Analogies subscales from the German version of the Cognitive Ability Test 4–13 + R by Heller and Perleth (2000). These scales consist of 25 figural and 20 verbal items in multiple-choice format. Using the ConQuest software (Wu, Adams, & Wilson, 2000), the scores for both subscales were estimated simultaneously using a Rasch model from item response theory. The Rasch scores were subsequently standardized using z scores. The intellectually gifted group scored z ≥ 1.91. The reliability of the composite score was rtt = .91.
Vocational interests The General Structure of Interests Test (German version: Allgemeiner InteressenStruktur-Test; Bergmann & Eder, 1992) consists of six 10-item scales and measures academic and vocational interests according to Holland’s (1997) RIASEC model on 5point scales. Scale scores reflect the mean score of the 10 items in the scale. The psychometric properties of the scales are reasonable (cf. L¨ udtke & Trautwein, 2004). The sample scale reliabilities (internal consistencies) range between ␣ = .85 and .90.
Socio-economic status Participants were asked to give some information on their parents’ education and current job situation. Using this information, the socio-economic status of the students’ families was determined based on the International Socio-Economic Index (ISEI) of occupational
Vocational interests of the gifted
313
status (Ganzeboom & Treiman, 1996). For each student, the highest ISEI (HISEI) score in the family (of either father or mother) was used. Data analysis The vocational interests of intellectually gifted, average, and less intelligent students were compared via analyses of variance. Likewise, the vocational interests of low-, average-, and high-achieving students were compared. In addition to this extremegroup approach, we employed a multivariate approach with continuous data to provide a more conservative analysis of the relationships between intelligence and school achievement, on the one hand, and vocational interests, on the other hand. Beside cognitive variables, several sociological/socio-economic variables were simultaneously used as independent variables to predict vocational interests. The intention was to analyse whether intelligence and school achievement could add to the prediction of vocational interests, beyond what was accounted for by gender and SES. In these analyses, which were run using the software HLM 6.0 (Raudenbush, Bryk, & Congdon, 2004), we used multilevel models to control for the influence of school type, which was coded as a series of dummy variables. As we suspect that vocational interests might differ for highly intelligent and highly achieving persons, in the multivariate analyses we also model a potential interaction between intelligence and achievement by entering an interaction term in the regression equation. The interplay of intelligence and achievement might not be just additive effects. Instead, the effect of achievement might be moderated by intelligence (e.g., high achievers with relatively low intelligence might prefer other vocational activities than high achievers with rather high intelligence). Because many students dropped out of the longitudinal study, there was a substantial amount of missing data in wave 2. We used the NORM computer software (Schafer, 1999) to substitute the missing data by means of multiple imputation (see Schafer & Yucel, 2002; Sinharay, Stern, & Russell, 2001). We opted for multiple imputation (instead of listwise deletion) because it was not reasonable to assume that the individuals who did not respond in wave 2 were a random sample of the target population (i.e., that the missing data were missing completely at random). Moreover, deletion of all cases with missing data in wave 2 would have led to the loss of much valuable information. Multiple imputation accounts for the uncertainty in the missing data by estimating several values for each missing value, and thus does not underestimate the standard errors of the parameter estimates (which would be the case with single imputation methods; Sinharay et al., 2001). There is growing consensus in the literature that multiple imputation is superior to traditional methods, such as listwise or pairwise deletion or mean imputation (L¨ udtke, Robitzsch, Trautwein, & K¨ oller, 2007; Schafer, 1997). Selectivity analyses revealed that intelligence, school achievement, and gender predicted drop-out in wave 2 (Nagy, 2006; Winkelmann, 2004). The variables most strongly related to subject dropout in wave 2 – intelligence, Abitur GPA, and gender – were, therefore, included as auxiliary variables in the imputation model. The use of appropriate auxiliary variables leads to more efficient and less biased parameter estimations (Collins, Schafer, & Kam, 2001). Ten data sets with imputed values were used for the multilevel analyses so that average results could be interpreted. For multivariate analyses in SPSS (MANOVA), separate analyses were performed with all 10 data sets with imputed values. Results of these 10 analyses were highly consistent; therefore, we only report the results for the first imputed data set.
314
Miriam Vock et al.
Results Zero-order correlations of the variables under study (RIASEC scores, school types, GPA, intelligence, GPA × intelligence interaction, gender, and SES of the family) are documented in Table 1. Further, we inspected the pattern of associations between the RIASEC domains and found them to be similar in both waves. The correlations between wave 1 scores and wave 2 scores on the same scales ranged from r = .63 (conventional) to r = .78 (artistic), indicating that the subjects’ interests did not change much over time. In the following, we report group mean scores for the intelligence and achievement groups separately for waves 1 and 2.
Vocational interests of intelligence and achievement groups Wave 1 First, we look at the preferences of each group separately, and then we compare the groups’ scores on each Holland scale. As shown in Table 2, intellectually gifted adolescents, as a group, favour investigative and enterprising activities over the other activity domains specified in Holland’s RIASEC model, while their less intelligent peers – average as well as low intelligent students – prefer social and enterprising activities. All three intelligence groups, on average, assign their lowest scores to realistic activities. A comparison of the three intelligence groups shows that very intelligent students (relative to less intelligent students) in their final year at school scored significantly higher on realistic and investigative interests and lower on social and enterprising interests. The effect size (based on a comparison between the two extreme groups) for investigative interests is large (d = −1.06), those for realistic and social interests are moderate (d = −.66 and d = .77, respectively), and, finally, the effect for enterprising interests is small (d = .29). Within the female group, mean scores for realistic, enterprising, and conventional interests do not differ as systematically between the three intelligence groups as in the male group. There are, however, trends for investigative and social interests which are comparable to the trends in the male group: more intelligent women report stronger interests in investigative activities and fewer interests in social activities than women with average or less intelligence. Figures 1 and 2 depict the RIASEC profiles (z-standardized scores) of the three intelligence groups by gender. Gender differences in the students’ interests are striking: very intelligent female students scored lower on realistic and investigative interests and higher on artistic and social interests than their male counterparts. Gifted young men, as a group, present a pronounced interest profile with particularly high scores on realistic and investigative interests and low scores on social interests. Gifted young women, as a group, however, show a rather flat interest profile (whereas less intelligent women have a very pronounced profile which is inverted to that of gifted men: low scores on realistic and investigative interests accompanied with high scores on social interests). The biggest differences were thus found between the profiles of very intelligent men and less intelligent women. When these two groups are compared, the gender-specific differences typical of representative samples are striking, with the strongest differences in realistic, investigative, and social interests (each > 1 SD). A 3 × 2 MANOVA with the factors intelligence group and gender revealed significant main effects for intelligence group [F(12, 670) = 4.36, p < .001, 2 = .07] and gender [F(6, 334) = 18.34, p < .001, 2 = .25], but no significant interaction effect between
A
S
E
C
School type economic
−.44∗∗−.30∗∗ .28∗∗ .37∗∗ .02 −.06∗∗ .05∗∗ −.02 .04∗ .07∗∗ .04∗ .04∗∗−.07∗∗−.16∗∗
−.26∗∗ −.06∗∗
.05∗∗
.01 −.01
.02 −.06∗∗ .03∗ .02
.06∗∗ .06∗∗
.00 −.21∗∗−.05∗∗−.01 −.03 −.01 .14∗∗ ∗∗ ∗∗ ∗∗ ∗∗ ∗∗ .24 −.09 −.19 −.04 .03 −.15∗∗ .22 .03
−.11∗∗
1.00
School type science and technology
.06∗∗ .14∗∗−.08∗∗−.12∗∗−.17∗∗
−.09∗∗−.02
−.10∗∗−.10∗∗−.04∗ −.02 .12∗∗ .19∗∗ 1.00 ∗∗ ∗∗ ∗∗ ∗∗ .18 −.13 −.19 −.09∗∗−.02 −.15∗∗ .31
1.00 .61∗∗ 1.00 −.06∗∗−.01 1.00 −.22∗∗−.10∗∗ .41∗∗ 1.00 .01 .01 .25∗∗ .36∗∗ 1.00 ∗∗ ∗∗ .19 −.03 .05∗∗ .58∗∗ 1.00 .27
I
Note. N = 4,487; ∗ p ⬍ .05; ∗∗ p ⬍ .01.
Realistic interests (R) Investigative interests (I) Artistic interests (A) Social interests (S) Economic interests (E) Conventional interests (C) School type: economic School type: science/technology School type: miscellaneous GPA (z-standardized) Intelligence (z-standardized) Interaction z-GPA × z-intelligence Gender Family SES (HISEI)
R
Table 1. Zero-order correlations (wave 1)
.19∗∗ −.10∗∗
.01
.09∗∗ −.14∗∗
1.00
School type: miscellaneous
−.07∗∗ −.14∗∗
.13∗∗
1.00 −.30∗∗
−.22∗∗ .08∗∗
.02
1.00
−.02 −.03
1.00
Interaction Intelligence GPA × GPA (z(zintellistandardized) standardized) gence
1.00 −.04∗
Gender
Vocational interests of the gifted 315
316
Miriam Vock et al.
Table 2. Vocational interests of men and women with low, average, and high intelligence as well as men and women with low, average, and high achievements (wave 1) RIASEC scales (M, SD) Group
Realistic (R) Invest. (I) Artistic (A) Social (S) Enterp. (E) Convent. (C)
Intelligence Low Average High
1.79 (.67) 2.13 (.68) 2.43 (.73)
2.24 (.73) 2.68 (.93) 3.18 (.95) 3.15 (.84) 2.62 (.81) 2.69 (.90) 2.90 (.80) 3.26 (.75) 3.05 (.79) 2.55 (.84) 2.50 (.82) 2.91 (.79)
2.48 (.86) 2.67 (.73) 2.51 (.72)
Male low Male average Male high
2.24 (.87) 2.36 (.70) 2.63 (.68)
2.62 (.82) 2.50 (.92) 2.82 (.91) 3.33 (.96) 2.75 (.78) 2.33 (.79) 2.64 (.87) 3.23 (.78) 3.28 (.65) 2.41 (.80) 2.37 (.79) 2.87 (.78)
2.81 (.84) 2.65 (.57) 2.50 (.71)
Female low Female average Female high
1.61 (.47) 1.97 (.62) 1.96 (.64)
2.09 (.64) 2.75 (.93) 3.32 (.93) 3.07 (.78) 2.52 (.82) 2.94 (.88) 3.09 (.69) 3.28 (.73) 2.53 (.83) 2.85 (.88) 2.81 (.81) 3.00 (.77)
2.35 (.84) 2.68 (.83) 2.52 (.75)
d (intelligence low vs. high) p (intelligence groups) 2 (intelligence groups)
−.66 ⬍.01 .12
Achievement Low Average High
2.18 (.80) 2.08 (.73) 2.07 (.73)
2.48 (.81) 2.59 (.88) 2.84 (.92) 3.07 (.83) 2.65 (.84) 2.62 (.84) 2.81 (.86) 3.13 (.79) 3.19 (.86) 2.78 (.78) 2.95 (.82) 3.32 (.81)
2.48 (.78) 2.53 (.84) 2.53 (.67)
Male low Male average Male high
2.54 (.81) 2.38 (.72) 2.40 (.78)
2.67 (.85) 2.47 (.88) 2.53 (.82) 3.14 (.89) 2.90 (.89) 2.36 (.82) 2.43 (.74) 3.09 (.73) 3.47 (.70) 2.57 (.68) 2.69 (.80) 3.30 (.94)
2.54 (.83) 2.57 (.77) 2.59 (.71)
Female low Female average Female high
1.76 (.55) 1.82 (.65) 1.80 (.57)
2.26 (.72) 2.74 (.86) 3.21 (.90) 2.99 (.74) 2.44 (.75) 2.85 (.79) 3.14 (.82) 3.16 (.84) 2.96 (.92) 2.96 (.83) 3.16(.78) 3.33 (.68)
2.41 (.73) 2.49 (.91) 2.48 (.63)
−.85 ⬍.001 .11
−.07 .82 .00
d (achievement low vs. high) p (achievement groups) 2 (achievement groups)
.14 .42 .01
−1.06 ⬍.01 .16
.15 .38 .01
−.23 .18 .00
.77 ⬍.01 .10
−.13 .44 .00
.29 ⬍.01 .04
−.30 .05 .02
−.04 .13 .01
Note. RIASEC scores represent the mean of the 10 items in a RIASEC scale; scores range from 1 = no interest to 5 = strong interest.
intelligence and gender [F(12, 670) = 1.34, p > .05, 2 = .02]. The univariate analyses of variance for the single scales revealed, however, a significant intelligence by gender interaction effect for investigative and conventional interests [F(2, 339) = 3.38, F (2, 339) = 3.16, respectively, both p < .05 and 2 = .02]. When low- and high-achieving students were compared, the degree of investigative interest was found to differ, with highly achieving students reporting stronger interest in investigative activities (see lower part of Table 2). There were no significant mean differences on the scales measuring realistic, artistic, social, enterprising, or conventional interests. Figures 3 and 4 present the RIASEC profiles (z-standardized scores) of the three achievement groups by gender. The 3 × 2 MANOVA revealed significant main effects (achievement groups: F[12, 732] = 7.94, p < .001, 2 = .12; gender: F[6, 365] = 27.10, p < .001, 2 = .31), but no significant interaction effect (F[12, 732] = .86, p > .05, 2 = .01).
Vocational interests of the gifted
317
Figure 1. RIASEC profiles of low-, average-, and high-intelligence male students in wave 1 (RIASEC scores were z-standardized for the whole sample).
Figure 2. RIASEC profiles of low-, average-, and high-intelligence female students in wave 1 (RIASEC scores were z-standardized for the whole sample).
Figure 3. RIASEC profiles of low-, average-, and high-achieving male students in wave 1 (RIASEC scores were z-standardized for the whole sample).
Wave 2 The pattern of results from wave 2 was largely consistent with that from wave 1. The mean RIASEC scores in wave 2 are documented in Table 3. The score means in all subsamples and on all six scales tended to be higher than in wave 1. In other words, mean interests in all RIASEC domains increased over the transition from school to vocational or university education.
318
Miriam Vock et al.
Figure 4. RIASEC profiles of low-, average-, and high-achieving female students in wave 1 (RIASEC scores were z-standardized for the whole sample).
Again, the very intelligent group expressed higher interest in the realistic and investigative domains than the less intelligent group, which was more interested in social activities. The effect sizes were stronger than those found in wave 1. For realistic, investigative and social interests, we found effect sizes of about 1 SD between students with less and high intelligence. The mean score differences for the enterprising scale, however, were not significant for wave 2 anymore. For the achievement groups, however, the only striking finding was a moderate effect for investigative interests with the high achievers scoring higher on that scale (d = −.70). The MANOVA for the wave 2 findings revealed – as in wave 1 – significant main effects for intelligence group [F(12, 310) = 2.14, p < .05, 2 = .08] and gender [F(6, 154) = 10.18, p < .001, 2 = .28]. The interaction effect between intelligence and gender failed to reach significance. The MANOVA for the achievement groups also revealed significant main effects (achievement groups: F[12, 354] = 4.00, p < .001, 2 = .12; gender: F[6, 176] = 21.59, p < .01, 2 = .42 ) and – as in wave 1 – no significant interaction effect (F[12, 354] = .65, p < .05, 2 = .05). Analysing continuous data: predicting RIASEC scores by achievement, intelligence, and social variables The findings from group comparisons were supported and extended by a multilevel analysis in which achievement and intelligence were analysed simultaneously as continuous data, and school type, gender, and socio-economic status were controlled. These analyses, which we performed for the whole sample for waves 1 and 2 (with imputed data), are reported in Table 4. The unstandardized b values are given as computed in HLM. As expected, school type and gender were generally the strongest predictors of students’ vocational interests. Female gender predicted stronger interests in artistic and social activities, male gender predicted stronger interests in realistic and investigative activities. Compared with students who attended a regular school, those students at a school specializing in economics reported stronger interests in conventional and enterprising activities. Students in schools specializing in science and technology scored higher on realistic and investigative interests but lower on social and enterprising interests. Students at schools specializing in nutrition, agriculture, or social education reported slightly higher investigative and social interests, but somewhat lower enterprising interests in wave 1. When gender and school type were controlled, intelligence and GPA added to
Vocational interests of the gifted
319
Table 3. Vocational interests of men and women with low, average, and high intelligence as well as men and women with low, average, and high achievements (wave 2; data set includes values from multiple imputation) RIASEC scales (M, SD) Group
Realistic (R) Invest. (I) Artistic (A) Social (S) Enterp. (E) Convent. (C)
Intelligence Low Average High
1.88 (.53) 2.19 (.67) 2.50 (.76)
2.48 (.68) 3.03 (1.03) 3.62 (.73) 3.45 (.67) 2.85 (.66) 3.07 (.91) 3.20 (.73) 3.30 (.78) 3.21 (.76) 2.79 (.85) 2.76 (.78) 3.14 (.75)
2.73 (.79) 2.74 (.67) 2.68 (.64)
Male low Male average Male high
2.45 (.53) 2.64 (.75) 2.76 (.70)
2.67 (.52) 2.55 (.90) 3.11 (.67) 2.53 (.66) 3.46 (.61) 2.64 (.85)
3.10 (.44) 3. 78 (.51) 3.12 (.67) 3.43 (.65) 2.62 (.74) 3.16 (.81)
3.30 (.60) 2.75 (.47) 2.59 (.69)
Female low Female average Female high
1.79 (.48) 1.99 (.53) 2.11 (.68)
2.45 (.71) 3.11 (1.04) 3.71 (.74) 3.40 (.69) 2.73 (.62) 3.31 (.90) 3.24 (.75) 3.25 (.83) 2.83 (.81) 3.04 (.79) 2.98 (.80) 3.12 (.66)
2.63 (.79) 2.73 (.75) 2.83 (.54)
d (intelligence low vs. high) p (intelligence groups)∗ 2 (intelligence groups)∗
−.95 ⬍.01 .10
Achievement Low Average High
2.37 (.81) 2.20 (.78) 2.30 (.67)
2.80 (.88) 2.64 (.77) 2.77 (.71) 2.98 (.83) 3.37 (.74) 3.00 (.75)
3.09 (.88) 3.24 (.77) 3.22 (.80) 3.30 (.73) 3.27 (.72) 3.43 (.67)
2.78 (.85) 2.72 (.69) 2.69 (.62)
Male low Male average Male high
2.75 (.85) 2.54 (.80) 2.65 (.62)
2.93 (.93) 2.22 (.60) 3.03 (.71) 2.39 (.88) 3.63 (.65) 2.75 (.72)
2.55 (.64) 3.16 (.74) 2.86 (.74) 3.28 (.72) 2.97 (.61) 3.47 (.71)
2.58 (.81) 2.79 (.74) 2.64 (.56)
Female low Female average Female high
1.98 (.56) 2.00 (.71) 2.05 (.59)
2.66 (.82) 3.07 (.69) 2.63 (.67) 3.33 (.57) 3.19 (.74) 3.18 (.72)
3.65 (.73) 3.31 (.82) 3.42 (.76) 3.31 (.74) 3.48 (.72) 3.40 (.65)
2.99 (.85) 2.67 (.67) 2.73 (.65)
−.70 ⬍.01 .13
−.22 .49 .01
d (achievement low vs. high) p (achievement groups)∗ 2 (achievement groups)∗
.09 .45 .00
−1.01 ⬍.01 .13
.25 .18 .02
−.47 ⬍.05 .04
1.14 ⬍.01 .16
.44 .14 .02
−.26 .33 .01
.07 .90 .00
.12 .81 .00
Note. RIASEC scores represent the mean of the 10 items in a RIASEC scale; scores range from 1 = no interest to 5 = strong interest. ∗ Results from MANOVA.
the prediction of vocational interests. Students’ intelligence predicted the strength of their realistic, investigative, and social interests; GPA was strongly associated with the strength of investigative interests. As intelligence as well as GPA to a certain amount correlate with the different school types (zero-order correlations ranging between r = −.15 and .14; cf. Table 1), our analysis here might slightly underestimate the impact of intelligence and school achievement on interests. Entering type of school in the model accounts for a (small) part of variance that might also be explained by intelligence or school achievement. In order to find possible moderator effects, we also introduced the interaction between intelligence and GPA as a predictor variable. Although the value reaches statistical significance, this interaction did not, however, add much to the prediction of
−.12∗∗ .43∗∗ .00 .03∗ .10∗∗ −.01 −.56∗∗ −.00
Type of school (reference: regular school) Economics −.08∗ Science and technology .47∗∗ .00 Miscellaneous schoolsa GPA (z-standardized)b .01 Intelligence (z-standardized) .09∗∗ Interaction GPA × intelligence −.00 Gender −.54∗∗ Family SES (HISEI) −.00∗ −.02 .33∗∗ .21∗∗ −.17∗∗ .10∗∗ −.04∗∗ −.50∗∗ −.00
Wave 1
I
−.07 .25∗∗ .13∗∗ −.14∗∗ .09∗∗ −.04∗∗ −.39∗∗ −.00
Wave 2 −.11∗∗ −.17∗∗ .01 −.04∗∗ −.05∗∗ .04∗∗ .45∗∗ .04∗∗
Wave 1
A
−.11∗ −.19∗∗ .00 −.05∗∗ −.04 .04∗ .54∗∗ .00∗∗
Wave 2 −.10∗∗ −.26∗∗ .15∗∗ −.02 −.11∗∗ .04∗∗ .53∗∗ .00∗∗
Wave 1
S
−.09∗ −.25∗∗ .12∗∗ −.02 −.11∗∗ .02 .46∗∗ .00∗∗
Wave 2 .20∗∗ −.21∗∗ −.16∗∗ −.04∗∗ −.04∗∗ .03∗ −.02 .00
Wave 1
E
.15∗∗ −.18∗∗ −.08 .01 −.02 .03 −.01 .00∗
Wave 2
.33∗∗ −.07 −.17∗∗ −.03∗ .02 .02 −.09∗∗ −.00∗∗
Wave 1
C
.24∗∗ −.06 −.11∗ .01 .02 −.01 −.01 −.03∗∗
Wave 2
Note. Nschools = 149; ∗ p ⬍ .05; ∗∗ p ⬍ .01. a Miscellaneous schools: schools specializing in nutrition, agriculture, or social education. b Grade point average: overall grade in the Abitur (German final exam), z-standardized, German school grades range from 1 to 6, with higher numerical values indicating lower achievement.
Wave 2
Wave 1
R
Table 4. Prediction of vocational interests (RIASEC) by cognitive and social/socio-economic variables (regression weights b from a two-level model, modelled in HLM)
320 Miriam Vock et al.
Vocational interests of the gifted
321
vocational interests. Hence, the two variables have only a small joint effect on interests over and above the combination of the separate effects. Finally, the SES of students’ families had virtually no influence on vocational interests when school type, gender, intelligence, and GPA were controlled. We found only a small positive effect on artistic interests in wave 1 and a small negative effect on conventional interests in wave 2. (Note that, due to the different metrics of the measures, very small SES coefficients can also yield significant results here.)
Stability and change of vocational interests Changes in the students’ vocational interests between waves 1 and 2 were modelled in further HLM regression models for each RIASEC interest score (Table 5). Here, we evaluated the effects of the predictor variables on change in the RIASEC scores by entering the scale score in wave 1 as a predictor of the respective scale score in wave 2. The resulting dependent variables are the wave 2 RIASEC scores, controlled for the effects of the wave 1 RIASEC scores. The wave 1 RIASEC scores had the largest effects on wave 2 RIASEC scores (b = .59–.73), indicating a high stability of the vocational interests. All other coefficients in these change models were much smaller than in the preceding models, but GPA, intelligence, and gender still had some predictive power beyond the wave 1 interest score for some wave 2 interest scores. Highly achieving students’ interests in investigative and artistic activities increased, whereas their interests in realistic and enterprising activities decreased. The interests of very intelligent students in realistic and investigative activities increased; their interest in social activities decreased further. Gender-specific profiles were consolidated and intensified in wave 2: men’s interests in realistic and investigative activities became stronger, as did women’s interests in artistic and social activities. Graduates from schools specializing in science and technology became more interested in realistic activities and less interested in social activities 2 years after leaving school. Table 5. Prediction of vocational interests (RIASEC) by cognitive and social/socio-economic variables in wave 2 and intensity of interest in wave 1 (regression weights b From a two-level model, modelled in HLM) R
I
A
S
Type of school (reference: regular school) Economics Science and technology Miscellaneous schoolsa
−.06 .10∗∗ .00
−.06 .05 .01
−.03 −.07 −.01
−.03 −.09∗ .03
GPA (z-standardized)b Intelligence (z-standardized ) Interaction GPA × intelligence Gender Family SES (HISEI) Intensity of interest at wave 1
.03∗ .04∗ −.01 −.18∗∗ .00 .71∗∗
−.04∗∗ .03∗∗ −.02∗∗ −.11∗∗ .00 .61∗∗
−.02∗ −.01 .01 .21∗∗ .00 .73∗∗
.00 −.04∗ .00 .13∗∗ .00∗ .63∗∗
E .02 −.04 .03 .03∗∗ .00 .01 .00 .00 .67∗∗
C .05 −.02 −.01 .02 .01 −.02 .04 .00 .59∗∗
Note. Nschools = 149; ∗ p ⬍ .05; ∗∗ p ⬍ .01. a Miscellaneous schools: schools specializing in nutrition, agriculture, or social education. b Grade point average: overall grade in the Abitur (German final exam), z-standardized, German school grades range from 1 to 6, with higher numerical values indicating lower achievement.
322
Miriam Vock et al.
Discussion and conclusions In this study, we describe and compare the vocational interests of young men and women with different levels of intelligence and academic achievements. Further, we analyse the impact of intelligence and achievement on specific interests in a sample of young adults with a large range of abilities. Vocational interests play a central role during different phases of vocational development and decision-making. For example, they guide the direction of intellectual effort towards certain domains and hence powerfully influence the acquisition of knowledge (Ackerman, 1996), and they determine whether a person will evaluate a chosen job or work environment as satisfying and rewarding (Dawis, 1996). Ackerman and Heggestad (1997) describe four broad trait complexes, which comprise different abilities, personality traits, and interest domains. Their PPIK theory (intelligence-as-Process, Personality, Interests, intelligence-as-Knowledge) suggests that gifted as well as high achieving adolescents might differ substantially in their vocational interests from their less able peers. Furthermore, based on this theory, intelligence as well as academic achievement should be good predictors of vocational interests. Previous studies with gifted students have not yielded entirely consistent results, however. Whereas findings leave no doubt that intellectually gifted students, that is, very intelligent students, are more strongly interested in investigative activities than their peers, findings on realistic and artistic interests as well as on the interaction between gender and giftedness have been mixed (Lubinski et al., 1995; Sparfeldt, 2007). In the present study, by comparing group means, we find that gifted students in their final year at school reported stronger investigative and realistic interests and weaker social and enterprising interests than did less intelligent students. Two years later, the same pattern of differences between the groups can be found. However, the differences for realistic, investigative, and social interests had grown stronger. We find smaller differences for the three achievement groups, only investigative interests differ substantially (with high achievers reporting stronger investigative interests than low achievers). In wave 2, both achievement groups also differ in their artistic interests, which is in line with Sparfeldt’s (2006) findings. Hence, the vocational interests profile of the gifted group differs markedly from that of the average and less intelligent groups and also slightly from that of the highly achieving group. These results demonstrate that a separate analysis of the interests of intellectual gifted, on the one hand, and the interests of highly achieving students, on the other hand, is reasonable as those groups exhibit different profiles. A limitation of this part of the analysis is that we operationalize achievement by teacher-assigned grades. Grades do not only reflect the students’ academic achievements but are affected by various other factors including type of school, teacher–student relationship, and a teacher’s personal tendency for leniency or stringency. Our results therefore refer to those high achievers who get outstanding grades from their teachers. A replication of our findings with standardized achievement tests as the basis for group formation would reveal whether the results also hold for those students whose achievements might be underestimated by their grades. The RIASEC profile of gifted young men differs from that of gifted young women. While gifted men score high on realistic and investigative interests and low on social interests, the gifted women’s profiles are rather flat, meaning that they – on average – do not particularly favour certain vocational activities over others. On the other hand, less intelligent women, as a group, show a pronounced profile, which is inverted to that
Vocational interests of the gifted
323
of gifted men, with low scores on realistic and investigative interests and high scores on social interests. An explanation based on socialization and parental influence for these gender differences among the gifted has been offered by Kerr and Cohn (2001), who had observed in longitudinal studies that gifted boys receive more pressure from their parents to follow linear career paths and to decide early, while gifted girls are actively encouraged by their parents to keep their options open and to decide later. Following this interpretation, we suppose that many gifted girls might profit from counselling and pull-out programmes which focus on helping female students to sharpen their interest profile and to find out which activities they really like and wish to pursue in the future – irrespective of society’s female role expectations. Also, gifted boys might often profit from counselling in order to find out about their real interests, which might differ from the male stereotypes. Another interpretation of the flat profile of gifted females might be, of course, that the individual women in fact have pronounced profiles but that the group is very heterogeneous, resulting in a group profile with low mean scores on each individual interest scale. Further in-depth analyses of the profiles of gifted girls will help to clarify the issue but are beyond the scope of this paper. To simultaneously analyse the impact of cognitive and social factors on vocational interests, we also used a multivariate approach to analyse continuous data, with intelligence and achievement (as well as the interaction between both) being used as continuous predictor variables. Because the whole, representative student sample was followed longitudinally in the present study (and not just the gifted subsample and a comparison subsample, as is usually the case in giftedness studies), we were able to perform more conservative analyses with continuous data, and did not have to restrict the analyses to small, somewhat arbitrary extreme groups. As expected, the multilevel regression analyses show a strong impact of gender on vocational interests. But beyond this gender effect, intelligence, school achievement, and type of school each contribute substantially to the prediction of vocational interests. To study at a school with a certain profile yields the expected effects: graduates of schools specializing in science and technology have higher realistic and investigative interests and lower artistic, social, and enterprising interests than graduates of regular schools. Graduates of schools specializing in economics have stronger enterprising and conventional interests and lower realistic, artistic, and social interests than graduates of regular schools. However, socio-economic background is largely irrelevant for the prediction of interests when school type, gender, GPA, and intelligence are statistically controlled. Also, the interaction between intelligence and achievement only adds very little to the prediction of vocational interests. Taken together, our analyses suggest that intellectually gifted young adults reveal stronger investigative and enterprising interests than their less intelligent peers (who score higher on social and enterprising interests). The biggest differences between the intelligence groups can be found between their realistic and investigative activities (stronger interest in the gifted group), and in their interest in social activities (stronger interest in the group of less intelligent participants). In wave 1, at least, intelligence is also negatively associated with enterprising and artistic activities. The latter finding seems to contradict predictions made on the basis of Ackerman’s intellectual/cultural trait complex; in their analyses, Ackerman and Heggestad (1997) found artistic activities to be typically positively associated with intelligence. Conventional interests, however, seem to be rather independent of intellectual ability in our data. This finding is coherent with Ackerman’s idea of the clerical/conventional complex. When high and low achievers are
324
Miriam Vock et al.
compared, we find statistically significant differences for investigative interests (wave 1: d = −.85; wave 2: d = −.70) and artistic interests (wave 2: d = −.47). We believe that our findings differ from findings in previous studies with gifted students for two main reasons: first, we aimed at distinguishing intellectually gifted students from high achievers at school by assessing giftedness with intelligence scales and school achievement with GPA. Many former studies have not had the necessary data to make such a distinction, meaning that effects were probably confounded. Second, there is no standard assessment tool for diagnosing intellectual giftedness. Instead, the intelligence scales commonly used are based on different conceptions of intelligence and differ in terms of both the intelligence construct and the test material administered. An intelligence scale measuring mainly fluid intelligence will produce different findings from scales that focus on, for example, verbal abilities, mathematical reasoning, or processing speed, as different intelligence factors are part of different trait complexes (cf. Ackerman & Beier, 2003). In this study, we used a combined score of indicators for verbal and nonverbal (figural) reasoning, thereby covering at least the verbal and the figural domain, as well as a more fluid component (figural reasoning) and a more crystallized component (verbal analogies) of intelligence. Because we study the same sample at two points in time crucial for vocational decisions, we are able to analyse changes in vocational interests in both the intelligence and achievement groups. We expected vocational interests to still be in flux during adolescence, and previous studies had found that the occupational interests of gifted students had changed substantially between adolescence and adulthood (Schmidt et al., 1998). Our assessment took place shortly before graduation from German high school (in grade 13), when students were around 19.5 years old, and 2 years later, when they were around 21.5 years old, at which time they had usually opted for vocational or academic post-secondary training. The correlations between the RIASEC scores in waves 1 and 2 ranged from r = .63 to r = .78, in line with previous findings on the stability of vocational interests (e.g., Johansson & Campbell, 1971). Contrary to findings from the SMPY (Schmidt et al., 1998), we do not observe a decrease of interest in any of the RIASEC scales. Instead, we found a slight mean increase in the interest reported for every interest scale and for every subsample. We interpret these findings such that young adults’ experiences at university or in vocational training may enhance their interest in different occupational activities. One reason for these diverging findings may be the time of the assessment: the SMPY participants were tested at age 28, when the first excitement of choosing and entering a profession may already have faded. Their interest profiles may also be better defined by that time, resulting in a decrease of mean scores. The HLM change models, in which the intensity of one interest domain at wave 1 is included as a predictor for the same interest at wave 2, indicate considerable stability of vocational interests. Still, GPA, intelligence, and gender are significant predictors in these models, indicating that change in occupational interest is partly attributable to these variables. For example, intelligence predicts higher interests in realistic and investigative activities and lower interests in social activities, even when the intensity of that interest in wave 1 is controlled. These results might be interpreted as further evidence for the increasing convergence of interests and abilities, resulting from accumulated experiences with one’s own competences, over lifelong development (cf. Ackerman, 1996). The analyses show that, despite considerable stability in the young adults’ vocational interests, the factors cognitive ability, achievement, and gender still extend their influence during this developmental phase of vocational orientation.
Vocational interests of the gifted
325
The advantages of longitudinal designs, which allow the development of a target group to be traced over time, are inevitably accompanied by typical drawbacks – in particular, dropout of participants over time. Particularly high drop-out could be expected at our second point of measurement – after graduation from school, when many students leave their homes and move away for post-secondary training. We addressed this problem by multiple imputation of missing values, based on sample information about those participants who did not respond in wave 2. It is widely accepted that this approach is preferable to the common alternatives (e.g., listwise deletion). Nevertheless, it is clear that our results from wave 1 (which were obtained from a representative sample) can be generalized to the population with a somewhat higher reliability than the results from wave 2. So far, we could incorporate only data from two waves in the analysis. For a detailed analysis of personal growth trajectories, however, a minimum of three waves would be required in order to rule out typical threats to the internal validity of repeated measures designs (e.g., regression to the mean). Future analyses of the subjects’ interests will allow for a more valid interpretation of the students’ development. Respondents were at the very beginning of their occupational careers when they answered the questionnaires. Future waves of the study will reveal how the gifted group and the highly achieving group make their way in the world of work. Analyses will then show whether the different vocational interest profiles predict different vocational developments and vocational success.
References Ackerman, P. L. (1996). A theory of adult intellectual development: Process, personality, interests, and knowledge. Intelligence, 22, 227–257. doi:10.1016/S0160-2896(96)90016-1 Ackerman, P. L., & Beier, M. E. (2003). Intelligence, personality, and interests in the career choice process. Journal of Career Assessment, 11(2), 205–218. doi:10.1177/1069072703011002006 Ackerman, P. L., & Heggestad, E. D. (1997). Intelligence, personality and interests: Evidence for overlapping traits. Psychological Bulletin, 121, 219–245. doi:10.1037/0033-2909.121.2.219 Ackerman, P. L., Kanfer, R., & Goff, M. (1995). Cognitive and noncognitive determinants and consequences of complex skill acquisition. Journal of Experimental Psychology: Applied, 1, 270–304. doi:10.1037/1076-898X.1.4.270 Bergmann, C., & Eder, F. (1992). Allgemeiner Interessen-Struktur-Test (AIST) [General Structure of Interests Tests]. Weinheim, Germany: Beltz. Bergmann, C., & Eder, F. (2005). Allgemeiner Interessen-Struktur-Test (AIST-R) [General Structure of Interests Tests Revised]. G¨ ottingen, Germany: Beltz Test. Collins, L. M., Schafer, J. L., & Kam, C.-M. (2001). A comparison of inclusive and restrictive strategies in modern missing data procedures. Psychological Methods, 6, 330–351. doi:10.1037/1082989X.6.4.330 Dawis, R. V. (1996). The theory of work adjustment and person-environment-correspondence counseling. In D. Brown, L. Brooks, & Associates (Eds.), Career choice and development (3rd ed., pp. 75–120). San Francisco: Jossey-Bass. Fox, L. H., Pasternak, S. R., & Peiser, N. L. (1976). Career-related interests of adolescent boys and girls. In D. P. Keating (Ed.), Intellectual talent: Research and development. Baltimore: The Johns Hopkins University Press. Ganzeboom, H. B. G., & Treiman, D. J. (1996). Internationally comparable measures of occupational status for the 1988 International Standard Classification of Occupations. Social Science Research, 25(3), 201–239. doi:10.1006/ssre.1996.0010
326
Miriam Vock et al.
Gottfredson, L. S. (1996). Gottfredson’s theory of circumscription and compromise. In D. Brown, & L. Brooks (Eds.), Career choice and development (pp. 179–232). San Francisco, CA: JosseyBass. Gottfredson, L. S. (1997). Why g matters: The complexity of everyday life. Intelligence, 24(1), 79–132. doi:10.1016/S0160-2896(97)90014-3 Heller, K. A., & Perleth, C. (2000). Kognitiver F¨ ahigkeitstest f¨ ur 4. bis 12. Klassen. Revision. KFT 4-12 + R [Cognitive Ability Test 4–12. Revision]. G¨ ottingen, Germany: Beltz Holland, J. L. (1959). A theory of vocational choice. Journal of Counseling Psychology, 6(1), 35–45. doi:10.1037/h0040767 Holland, J. L. (1997). Making vocational choices. A theory of vocational personalities and work environment (3rd Ed.). Englewood Cliffs, NJ: Prentice Hall. Holling, H., Preckel, F., & Vock, M. (2004). Intelligenzdiagnostik [Assessment of intelligence]. G¨ ottingen, Germany: Hogrefe. Johansson, C. B., & Campbell, D. P. (1971). Stability of the SVIB for men. Journal of Applied Psychology, 55, 34–36. Kerr, B. A. & Cohn, S. J. (2001). Smart boys: Talent, manhood, and the search for meaning. Scottsdale, AZ: Great Potential Press. K¨ oller, O., Watermann, R., Trautwein, U., & L¨ udtke, O. (2004). Wege zur Hochschulreife in Baden-W¨ urttemberg. TOSCA – Eine Untersuchung an allgemein bildenden und beruflichen Gymnasien [Ways to qualification for university entrance. TOSCA – A study at traditional and vocational Gymnasium schools]. Opladen, Germany: Leske + Budrich. Kulik, J. A. (2004). Meta-analytic studies of acceleration. In N. Colangelo, S. G. Assouline, & M. U. M. Gross (Eds.), A nation deceived: How schools hold back America’s brightest students (pp. 13–22). The Templeton National Report on Acceleration. Iowa City: University of Iowa. 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. doi:10.1037/0022-3514.74.4.996 Lofquist, L. H., & Dawis, R. V. (1991). Essentials of person–environment correspondence counseling. Minneapolis: University Press. Lubinski, D. & Benbow, C. P. (2000). States of excellence. American Psychologist, 55, 137– 150. doi:10.1037/0003-066X.55.1 Lubinski, D., Benbow, C. P., & Ryan, J. (1995). Stability of vocational interests among the intellectually gifted from adolescence to adulthood: A 15-year longitudinal study. Journal of Applied Psychology, 80(1), 196–200. doi:10.1037/0021-9010.80.1.196 L¨ udtke, O., Robitzsch, A., Trautwein, U., & K¨ oller, O. (2007). Umgang mit fehlenden Werten in der psychologischen Forschung: Probleme und L¨ osungen [Handling of missing data in psychological research. Problems and solutions]. Psychologische Rundschau, 58(2), 103– 117. L¨ udtke, O., & Trautwein, U. (2004). Die gymnasiale Oberstufe und psychische Ressourcen: Gewissenhaftigkeit, intellektuelle Offenheit und die Entwicklung von Berufsinteressen [Upper secondary education at Gymnasium schools and psychological resources: Conscientiousness, intellectual openness, and the development of vocational interests]. In O. K¨ oller, R. Watermann, U. Trautwein, & O. L¨ udtke (Eds.), Wege zur Hochschulreife in Baden-W¨ urttemberg. TOSCA – Eine Untersuchung an allgemein bildenden und beruflichen Gymnasien (pp. 367–401). Opladen, Germany: Leske + Budrich. Nagy, G. (2006). Berufliche Interessen, kognitive und fachgebundene Kompetenzen: Ihre Bedeutung f¨ ur die Studienfachwahl und die Bew¨ ahrung im Studium [Vocational interests, cognitive and scholastic abilities: Their role in choice of major and success at university]. Doctoral thesis, Berlin, Germany: Free University. Retrieved from http://www.diss.fu-berlin. de/2007/109/ Ones, D.S., Viswesvaran, C., & Dilchert, S. (2005). Cognitive ability in selection decisions. In O. Wilhelm & R. Engle (Eds.), Handbook of understanding and measuring Intelligence. London: Sage.
Vocational interests of the gifted
327
Platzer, S. (2002). Erfolg ist nicht alles. Zum Studierverhalten von Absolventen eines Sonderf¨ orderzweiges f¨ ur Hochbegabte [Success is not everything. On the study behavior of graduates of a special curriculum for the gifted] [Dissertation]. Catholic University of Nijmegen, The Netherlands. Post-Kammer, P., & Perrone, P. A. (1983). Career perceptions of talented individuals: A follow-up study. Vocational Guidance Quarterly, 31, 203–211. Preckel, F., Holling, H., & Vock, M. (2006). Academic underachievement: Relationship with cognitive motivation, achievement motivation, and conscientiousness. Psychology in the Schools, 43(3), 401–411. Randahl, G. (1991). A typological analysis of the relations between measured vocational interests and abilities. Journal of Vocational Behavior, 38, 333–350. doi:10.1016/00018791(91)90034-J Raudenbush, S., Bryk, A., Congdon, R. (2004). Hierarchical linear and nonlinear modelling (HLM 6.0) [Computer software]. Lincolnwood, IL: Scientific Software International. Rice, J. P. (1985). Gifted children: developing total talent (2nd ed.). Springfield: Charles C. Thomas. Robinson, N. M. (2005). In defense of a psychometric approach to the definition of academic giftedness. A conservative view from a die-hard liberal. In R. J. Sternberg & J. E. Davidson (Eds.), Conceptions of giftedness (pp. 280–294). Cambridge: Cambridge University Press. doi:10.1017/CBO9780511610455.016 Rost, D. H. (2000). Hochbegabte und hochleistende Jugendliche [Gifted and highly achieving adolescents]. M¨ unster, Germany: Waxmann. Schafer, J. L. (1997). Analysis of incomplete multivariate data. New York: Wiley. doi:10.1201/9781439821862 Schafer, J. L. (1999). NORM: multiple imputation of incomplete multivariate data under a normal model, version 2.03. Software for Windows 95/98/NT. Retrieved from www.stat.psu.edu/∼jls/misoftwa.html Schafer, J. L., & Yucel, R. M. (2002). Computational strategies for multivariate linear mixed-effects models with missing values. Journal of Computational and Graphical Statistics, 11(2), 437– 457. doi:10.1198/106186002760180608 Schmidt, F. L., & Hunter, J. (2004). General mental ability in the world of work: Occupational attainment and job performance. Journal of Personality and Social Psychology, 86(1), 162– 173. doi:10.1037/0022-3514.86.1.162 Schmidt, D. B., Lubinski, D., & Benbow, C. P. (1998). Validity of assessing educational-vocational preference dimensions among intellectually talented 13-year-olds. Journal of Counseling Psychology, 45(4), 436–453. doi:10.1037/0022-0167.45.4.436 Sinharay, S., Stern, H. S., & Russell, D. (2001). The use of multiple imputation for the analysis of missing data. Psychological Methods, 6(4), 317–329. doi:10.1037/1082-989X.6.4.317 Sparfeldt, J. R. (2006). Berufsinteressen hochbegabter Jugendlicher [Vocational interests of gifted adolescents]. M¨ unster, Germany: Waxmann. Sparfeldt, J. R. (2007). Vocational interests of gifted adolescents. Personality and Individual Differences, 42(6), 1011–1021. doi:10.1016/j.paid.2006.09.010 onlichkeit, entwicklung, f¨ orderung [Gifted children: Stapf, A. (2003). Hochbegabte kinder: pers¨ personality, development, promotion]. M¨ unchen: Beck. Sternberg, R. J., & Davidson, J. E. (2005). Conceptions of giftedness. Cambridge: Cambridge University Press. doi:10.1017/CBO9780511610455 Strong, E. K. (1927). Vocational Interest Test. Educational Record, 8, 107–121. Strong, E. K. (1945). Vocational interests of men and women. Stanford: Stanford University Press. Swanson, J. L. (1999). Stability and change in vocational interests. In M. L. Savickas, & A. R. Spokane (Eds.), Vocational interests: Meaning, measurement, and counseling use (pp. 135–158). Palo Alto, CA: Davies-Black Publishing.
328
Miriam Vock et al.
Tracey, T. J., & Rounds, S. B. (1993). Evaluating Holland’s and Gati’s vocational-interest models: A structural meta-analysis. Psychological Bulletin, 113, 229–246. doi:10.1037/00332909.113.2.229 Vock, M., & Holling, H. (2007). Begabung und Berufserfolg [Giftedness and vocational achievements]. In K. A. Heller, & A. Ziegler (Eds.), Begabtsein in Deutschland [Being gifted in Germany] (pp. 234–263). M¨ unster, Germany: LIT. Vock, M., Preckel, F., & Holling, H. (2007). F¨ orderung Hochbegabter in der Schule. Evaluationsbefunde und Wirksamkeit von Maßnahmen [Fostering the gifted at school. Evidence from evaluation studies and effectiveness of interventions]. G¨ ottingen, Germany: Hogrefe. Winkelmann, H. (2004). Pr¨ adiktoren differentiellen Teilnahmeverhaltens in L¨ angsschnittstudien: Eine Untersuchung der Panelmortalit¨ at in der TOSCA-Studie [Predictors of differential participation behavior in longitudinal studies: An investigation of panel mortality in the TOSCA study] (Unpublished diploma thesis). Free University, Berlin, Germany. Wu, M. L., Adams, R. J., & Wilson, M. R. (2000). ConQuest [Computer software]. Melbourne: Australian Council for Educational Research (ACER). Received 12 January 2011; revised version received 31 October 2011