Locus of Control, Self‐Efficacy, and Motivation in Different Schools: Is moderation the key to success?

Self-Concept Research: Driving International Research Agendas Locus of Control, Self-Efficacy and Motivation in Different Schools: Moderation, the Key to Success? Angelika Anderson, John Hattie and Richard J. Hamilton The University of Auckland New Zealand This study used a novel multidimensional locus of control instrument (I-SEE) to investigate the relationship between locus of control, motivation and academic achievement in three schools. The strengths of the I-SEE are that it incorporates the construct of self-efficacy, and that it is embedded in a model of personality and action based on fieldtheoretical conceptions. Further, it includes the role of the environment and personality in determining action. The results support a multidimensional conceptualisation of locus of control and the utility of the I-SEE. There were statistically significant differences between schools for motivation and achievement and also a mediating effect between locus of control and school-type suggesting that interactional models are required in investigations of motivation and achievement. Furthermore, moderate levels of locus of control and self-efficacy appear to be more adaptive than either extremely high or low levels. Locus of control, as originally conceptualised by Julian Rotter (1966) refers specifically to people’s perceptions of control over access to reinforcements. As such it attempted to bridge the gap between operant and cognitive psychology. Control orientations have been among the most studied topics and they have been found to be critical in relation to academic achievement and motivation (Kalechstein & Nowicki, 1997). Over the years the various constructs relating to “control” have undergone development during which new terms and reinterpretations of old notions have been proposed and instruments have been developed. Because of such alternative notions (particularly as they pick up nuances when associated with like-sounding constructs) there has often been a lack of attention to building solid theoretical foundations, causing confusion and making interpretation and integration of findings difficult, particularly in relation to studies investigating the relationship between locus of control and academic achievement. This study presents a highly developed model integrating these notions of control into a sound theoretical framework and a model of action and personality. This model is congruent with and builds upon Rotter’s original conception of the construct ‘locus of control’. The paper also introduces a new locus of control instrument, which is then used to investigate the relationship between locus of control (as defined by the action model), academic achievement and motivation across three different types of schools. The early work with the construct ‘locus of control’ relied on unidimensional notions and instruments, which dichotomised the construct, and divided the world into externals and internals, typically equating internal with good and external with bad. This was particularly striking in studies that investigated cognitive activity (such as memory, attention etc), where externals seemed to be deficient compared to internals (Lefcourt, 1982). Some time ago researchers began to suggest that uni-dimensional conceptualisations of locus of control are inappropriate (Krampen, 1985; Levenson, 1981) as equating internality with health and externality with pathology reflects an overly simplistic view of the construct (Krampen, 1991; Krampen, von Eye, & Brandtstaedter, 1987; Krampen & Wieberg, 1981; Levenson, 1981). Rather both extremes (extreme internality as well as extreme externality) are associated with a loss of reality (Krampen, 1985). Rotter’s work has been significantly extended by Krampen who integrated the construct of ‘locus of control’ into his “Action-theoretical Model of Personality” (AMP). The AMP aims to link personality variables to situational variables systematically (Krampen, 1988). Krampen’s model occupies the middle ground between those action theories on one hand which aim to predict and describe behaviour in specific situations only (with little regard for enduring personality variables, and personality theories), and on the other hand theories that are concerned with the description of relatively stable personality constructs without attempting to link specific experiences to the development of these enduring orientations. The AMP is an action-theoretical model of personality, because it describes personality development as a process of generalisation of situation specific experiences over time and indicates how personality- and person-variables determine action in given situations or domains. The AMP postulates that the prognostic and the descriptive utility of the personality variables co-vary with the extent to which action- or life-situation can be interpreted or understood by individuals. In well-known situation or situations that can be easily classified (strong situations) and for which action-specific cognitions exist in (subjectively) sufficient measure, the influence of the situation is strong and the descriptive and prognostic value of the AMP personality variables is low. In 'weak' (novel, unpredictable) situations for which there are no individual or collective (in the form of social norms) experiences and representations (such as scripts) in sufficient quantity, the descriptive and prognostic value of the AMP personality variables is high. Since the subjective newness / ambiguity of a situation is a continuum, the relative descriptive and predictive value of situation-specific person variables and generalised personality variables are also continua. The relative utility of specific person variables versus generalised personality variables is therefore a question of degree, in accordance with notions of the dynamic interactionism of person - and situation- factors. It is important to remember also that actions and their results as well as consequences, feed back - not only into the situation - but also into the development of personality factors. The AMP attempts to 1 Self-Concept Research: Driving International Research Agendas offer a functional perspective, which connects interactionist ideas with the possibility to make statements about the relative descriptive and predictive value of generalised personality variables and situation-specific person variables. That way the AMP implies ways of measuring for diagnostic and scientific purposes. Depending on the kind and the differentiation of the situation-structure, situation-specific, domain-specific or generalised construct-operationalisations are indicated. situation may depend to some extent on the degree to which it is structured or organised. Generally much more is known about the effect of environments on the development of locus of control. Little is known about other possible moderating effects of environment on the relationship between locus of control and action, here specifically academic achievement and related behaviours. Krampen’s “Action-theoretical Partial Model of Personality” (AMP) suggests that both processes are linked (Krampen, 1988). Environments that allow people to develop appropriate situation specific locus of control orientations foster adaptive behaviour in a given situation as well as the development of generalised internality, or locus of control configurations adaptive in a given society. Krampen’s AMP describes how personality and environment both influence action on one hand and personality development on the other hand. Other person variables such as the ability to discriminate and classify environmental cues appropriately are likely to also influence this process. One reason for the strong association between generalised locus of control and achievement could be that the same person variables, which foster adaptive development and appropriate behaviour in specific situations are also predictive of academic achievement. Some of these are essentially cognitive skills, such as the ability to discriminate and classify. According to Krampen (1988; 1991) one important aspect about the environment which may influence the prediction of behaviour is how familiar the environment is to the subject, and how structurable (classifiable) it is. Measures of generalised personality constructs are better predictors of behaviour in novel or unstructured situations, while domain or situation specific measures are better predictors of behaviour in well-known or structured situations. Central variables of the AMP (other than locus of control and competency beliefs) include ‘level of conceptualisation’, which is the subjective knowledge about the dynamics of situations and is related to intelligence and problem solving competence, and value orientations, which are measures of the relative importance or value of outcomes and consequences, similar to Rotter’s ‘need value’. In the AMP model such value orientations not only refer to terminal values, but also emotionally, socially and culturally mediated valuations. According to Krampen all the above variables play a part in determining a person’s behaviour in a given situation. Thus the AMP is seen as a constructivist theory, open to elaborations and changes and the variables, which make up this model should be seen as continua, not as dichotomies. Similar too much of the research into the construct of locus of control, research into antecedents to the development of control orientation has often been atheoretical, lacking the guidance of a model. Krampen’s AMP can be used to address this shortcoming. The model would predict that consistency across time in the same environments or types of situations will foster the development of expectations of control. Consistency across environments and congruence across domains might foster generalised control expectancies congruent with requirements in different situations and domains, leading to a greater likelihood to behave adaptively. At the Mediating Factors Between Locus of Control and Academic Achievement Within the AMP, there are a number of factors that are predicted to mediate the relationship between locus of control and academic achievement. Personal variables including age, sex, reading ability and culture / ethnicity and ability can differentially affect outcomes (Stipek & Weisz, 1981), as do some aspects of the instructional environment (Lefcourt, 1982). In addition, some argue that academic achievement and locus of control are related indirectly, mediated by motivation / motivated behaviour such as task completion, participation and engagement (Finn & Rock, 1997; Stipec & Weisz, 1981). The Environment Previous research has shown that the instructional environment mediates the relationship between locus of control and achievement. Internals generally report greater satisfaction with schooling than externals. Externals prefer high discipline conditions (Parent, Forward, Canter, & Mohling, 1975) and more structured educational settings (Trice, 1980). Skinner (1990) found that pupil perceptions of both teacher involvement and contingency affected measures of engagement in pupils. Boggiano et al (1988) found that in the absence of evaluative pressure there was no difference in the motivation of children with high and low perceptions of control. Krampen (1987) found that different kinds of teacher feedback (social, individual and factual) affected performance outcome and altered locus of control orientation in school children. Individual (noncomparative) feedback was the most beneficial for children both in terms of achievement outcomes and in changes of locus of control orientations in the direction of increased internality. In all, there are relatively few studies investigating the interaction between locus of control and the environment using a multidimensional measure of locus of control, and little of the above research is based on sound theoretical foundations. As outlined earlier, one critical part of the environment in relation to the predictive utility of generalised versus domain specific measures of locus of control is the familiarity or classifiability of a situation. Generalised locus of control is expected to be more predictive of behaviour in novel or unstructured situations (Krampen, 1988, 1991). Domain specific measures are expected to be better predictors of behaviour in known situations where the person has been able to develop domain specific expectations during prior exposure to and experiences in that situation. The extent to which a situation is recognisable, or classifiable as a certain known type of 2 Self-Concept Research: Driving International Research Agendas same time the model suggests that locus of control might be quite malleable (on the specific as well as the general level) and change with changes in life circumstances or through experiences with new situations. There is some evidence for this. For example, prison inmates develop locus of control profiles characterised by relatively high externality levels, reflecting realistic adaptation to their altered life-circumstances where ‘powerful others’ are largely in control (Krampen, 1991). Children who experience different kinds of teacher feedback develop corresponding locus of control profiles again evidencing adaptation to their current experiences (Krampen, 1987). The purpose of the current study is to examine a novel multidimensional instrument measuring generalised locus of control in relation to academic achievement and related behaviours in schools that differed on a number of environmental dimensions that might affect the extent to which generalised locus of control is predictive of behaviour. In accordance with field-theoretical conceptualisations (Lewin, 1952; Rotter, 1954), personality and environment are expected to interact in determining outcomes in terms of academic achievement and related behaviours such as motivated behaviour in the classroom. The theoretical expectation is that generalised locus of control is most predictive of behaviour in novel, less structured (weak) situations. One would expect that generalised locus of control should not be predictive of behaviour in the latter years of secondary schools. Students all have had a number of years experience with schools by then, in addition there should be sufficient collective experiences and representations about schools within a given society so that social norms etc. exist. Findings in the research literature have shown that generalised locus of control is predictive of academic achievement and related behaviours. Most of that research has used unidimensional measures of locus of control, neglected the role of environment, and was not based on a sound theoretical model of personality and action. Therefore this study investigated the relationship between generalised locus of control and academic achievement and related behaviour in three different types of school, which differ in the extent to which they offer structured situations. The prediction is that generalised locus of control will be a poor predictor of academic achievement and related behaviour in all schools. There will also be some between school differences in the extent to which generalised locus of control is predictive of achievement and related behaviour, with a stronger association in the less structured school. In addition to structure, competitiveness and cooperativeness of schools were also assessed. Highly competitive schools might affect the degree to which individuals can affect outcomes (success versus failure) directly. For example, in situations where normative evaluations are salient outcomes (success versus failure) depend on the behaviour of others as well as a person’s own behaviour. In such situations some degree of externality might be more adaptive and also the extent to which a persons perception of own competence might be become an important factor. Therefore, in the more competitive schools a high level of competence belief might be associated with higher levels of academic achievement and related outcomes. In addition, one might predict that such schools foster the development of higher levels of externality. Method The participants were year 12 students (age 16 / 17) from 4 classes in each of 3 Auckland secondary schools. In all 215 largely ‘Pakeha” (of European descent) students participated. 121 of them were females and 94 were males. They were selected from schools deemed by expert judges to differ on the dimensions of structure, co-operativeness and competitiveness. Structure was defined as the salience / presence of rules and regulations, competitiveness as concerned with the salience of normative evaluations (such as streaming), and co-operativeness as relating to the emphasis on group-work and positive peer interactions. From a list of all secondary schools in the greater Auckland area, those at the extreme ends of socioeconomic ratings (in NZ these are termed decile rankings, decile 9 – 10 high, and 1 – 2 low) and single sex schools were eliminated to avoid possible confounds and to keep some known mediating variables (i.e. socio-economic status) constant. The list of the 28 remaining schools was given to three expert judges familiar with Auckland secondary schools. These expert judges were asked to identify from the list the five most extreme schools on the following dimensions: highest and lowest on measures of ‘structure’, ‘competitiveness’ and ‘co-operativeness’ (six categories in all). A school could be listed in more than one category, and some schools were never listed. Schools in the resulting listings fell broadly into two groups: those high on measures of structure and competitiveness and low on cooperation and those low on structure and competitiveness, and high on co-operativeness. From these judgments three schools were chosen: one which was judged to be high on structure, competition, and low on cooperation by all three judges (= school 3), one which was judged to be low on structure and competition, and high on cooperation by all three judges (= school 2) and one school, which was considered neutral and not extreme on any of the above dimensions (it was not consistently selected by any of the judges as high or low on any of the above dimensions). The two extreme schools chosen were each listed nine times by the three judges together, while the chosen neutral school (school 1) was only selected three times in total (once as high on structure, and twice as low on structure). Schools 1 and two were large urban schools (roles of 1508 and 2026 students respectively) with a decile ranking of 5. School 2 was smaller (802 students) and had a decile ranking of 7. The Principal of each school nominated four classrooms from their school to participate in this study. The selected samples from the three schools were comparable in terms of ethnic composition. The sample from School 1, had significantly more girls than boys than the samples of the other two schools (65% of the total sample from this school compared to 51% and 54% of the sample from schools 2 and 3 respectively). 3 Self-Concept Research: Driving International Research Agendas Measures There were six measures, which will be described in detail in this section. Locus of Control:English Version of FKK (Fragebogen zu Kompetenz- und Kontrollüberzeugungen), (Krampen, 1991) = I-SEE This instrument was developed by Krampen (1991) in German in 1991, by translating Levenson’s IPC scales (Levenson, 1981) into German and changing all items into less culturally dependent statements. Further, a fourth scale measuring ‘self-concept of own ability’ was added. Thus, the FKK consists of four primary scales each with eight items assessed on a six-point Likert type scale ranging from “not at all true” to “very true”:. FKK – SK: self-concept of own ability (items 4,8,12,16,20,24,28 and 32), FKK – I: internality (1,5, 6, 11, 23, 25, 27 and 30), FKK – P: social externality (items 3, 10, 14, 17, 19, 22, 26, and 29), 4and FKK – C: fatalistic externality (items 2, 7, 9, 13, 15, 18, 21, and 31). The minimum and maximum scores for each scale are eight and 48 respectively, a high score indicating high levels of competency and control beliefs. This instrument was normed on 248 German 14 – 17 year olds as well as 2028 adults. Cronbach alphas for the largest of those studies (n = 2028) were: 0.76 for the competency belief sub-scale, 0.70 for the internality subscale, 0.73 for the powerful others sub-scale, and 0.75 for the chance sub-scale. The corresponding test – retest reliabilities of the sub-scales were established in a different study (n = 127, interval 3 months): 0.75, 0.72, 0.68 and 0.84, respectively. The mean scores for the four primary scales for the adult sample were FKK-SK: 31.9; FKK-I: 32.4; FKK-P: 26.1, FKK-C: 26.8, and for the teenage sample: 29.8, 32.2, 28.3, 26.5, respectively. There are statistically significant intercorrelations in the theoretically expected directions between the primary scales support the formation of the secondary scales (SKI and PC). Combining two primary sub-scales derived these secondary scales. The SK scale (competency belief) together with the internality scale (I) gave a measure of self-efficacy (SKI). A combination of the two-externality scales (C and P) gave a measure of overall externality (PC). This instrument was translated into English by the backtranslation method. All of the items, which were clearly straight translations of Levenson’s original items, reverted to the original items. Therefore in the end 12 of these 32 items are Levenson’s original items (items 2, 3, 6 (=5), 7 (=6), 9 (=7), 14 (=11), 19 (=15), 23 (=18), 25 (=19), 27 (=21), 29 (=22), 30 (=23). See Table 1 for a list of all items by scale. For an extensive discussion of the ISEE and the AMP, particularly in relation to developmental issues, see Greve, Anderson and Krampen (2001). measures were used: a self-report measure, a teacher rating, and a quantitative measure of task-completion. The self-report scale was the 10 - item ‘Involvement’ scale of the Classroom Environment Scales (Tricket & Moos, 1974), which measures the “extent to which students are attentive and interested in class activities, participate in discussions, and do additional work on their own” (see above) was adapted, by changing statements into “I” statements, to obtain a self-report measure of ‘engagement’, or motivated classroom behaviour. Teachers were asked to rate the level of participation of each student on a scale from 1 to 7, where 1 = this student always participates actively, with relevant contributions, and 7 = this student never participates at all and seems disinterested. A measure of task completion was obtained by working out the proportion of Year 12 English assignments the student had completed at the time of testing, as a proportion of numbers of assignments he / she should have completed at that time (expressed in percent). Measure of Academic Achievement School Certificate English results from the previous year were obtained from the school records. In New Zealand School Certificate is a national, norm-referenced examination-based measure of achievement at end of year 11 in student’s schooling. Additional information sought included socio-demographic variables (age, gender, ethnicity and other motivational variables (need-value English; a rating of how important it is to do well in English, both for the individual and his / her parents). Data Analysis For each scale and sub-scale descriptive data were calculated (means and standard deviations, coefficient alphas for each scale and sub-scale). Structural equation models were used (AMOS, (Arbuckle, 1995). The RMSEA index of goodness of fit was used in this study, as it is not influenced by the size of the sample, has a theoretically known distribution, and is regarded as the most effective measure to assess the quality of fit of the model to the data. Browne and Cudeck (1993) have commented that a value of the RMSEA of 0.05 or less is indicative of a close fit of the model, and that a value for the RMSEA of 0.08 or less indicates a reasonable error of approximation (Browne & Cudeck, 1992). Means and standard deviations of measures of locus of control, measures of motivated behaviour, and previous School Certificate result were calculated. These were examined for statistical significance by carrying out MANOVA for class within school, sex by class within school, sex by school, school and sex. MANOVA was used to allow for the intercorrelations between dependent variables, and thus increasing the power of the study. Where there were statistically significant MANOVA results, then univariate ANOVAs were used to ascertain Measures of Motivation the scales contributing most of the overall differences. A K-Means cluster analysis was carried out on the primary locus of control scale scores to identify locus of In order to validly and reliably measure students’ motivated behaviour in the classroom three different control groupings in the sample in order to analyse it for 4 Self-Concept Research: Driving International Research Agendas person – environment interactions. K-means cluster analysis was used in this instance because the relatively large number of cases to be grouped (N = 215, it is the recommended method when there are in excess of 200 cases (Coakes & Steed, 1999)). In addition this method allowed us to specify a given number of clusters, based on theoretical expectations. Discrepancy analyses were carried out in order to determine if there were any statistically significant differences in the distribution of locus of control clustermembers between classrooms, schools and by classroom climate cluster. To investigate school-effects an ANOVA was carried out for between-group differences in previous school Certificate results for School, I-SEE cluster and School by I-SEE cluster Lambda = 0.606, F (27,523) = 3.62, df = p< 0.001; School: Wilks Lambda = 0.844, F (6,358) = 5.28 p<0.001]. Univariate Analysis of variance showed that classes within schools differed on levels of CES E (self-reported engagement) [F (9, 181) = 2.92 , p< 0.005], levels of Participation [F (9, 181) = 5.36, p<0.001], and task completion [F (9, 181) = 3.11 p<0.005]. There was also a significant main effect between schools for all measures of motivation: CES E (self-reported engagement) [F (2, 181) = 3.36, p< 0.05], task completion [F (2, 181) = 13.93 p< 0.001], and participation [F (2, 181) = 3.14 p< 0.05]. There were statistically significant differences between groups in previous School Certificate results between schools [F (2, 166) = 6.56 p< 0.005] between classes within schools [F (9, 166) = 10.97 p<0.001], for sex [F (1, 166) = 33.28 p<0.001] and for sex within classes [F (9, 166) = 3.81 p<0.001]. Table 4 shows that overall girls have higher scores in their School Certificate English Results results than boys do, and that schools 1 and 3 have higher scores in their School Certificate English results than Reliability and Validity of Measurement Instruments school 2 does. In order to investigate further the Table 2 provides the means and standard deviations of relationship between these measures of achievement and the major scales obtained by the sample as a whole and the locus of control the sample was grouped by I-SEE estimates of reliability (coefficient alphas). All scales and configuration. sub-scales are sufficiently reliable to be included in further Person – Environment Interactions analyses. Figure 1 represents the path diagrams for the I-SEE According to the field-theoretical roots of the construct locus of control scales used in this study (as estimated in AMOS), giving the standardised factor loadings for all of locus of control and the AMP in which the I-SEE items within scales and for all sub-scales. For the I-SEE integrated both, development and action are the result of scales all correlations between sub-scales are as expected, person – environment interactions. In order to investigate with strong positive correlations between the I-SEE I, and the utility of the I-SEE to investigate such questions the the I-SEE SK scales, supporting the formation of the following set of analyses were carried out. The purpose of secondary scale of self-efficacy, and strong positive these analyses was to identify interaction effects of locus correlation between the I-SEE C and the I-SEE P scales, of control and school environments in relation to supporting the formation of the secondary scale of motivated behaviour and development. The sample was grouped by I-SEE configuration. To combined externality. The correlations between the I-SEE accomplish this a K-means cluster analysis was carried out I, and SK on one hand, and the I-SEE P and C scales on for the primary scales of the I-SEE scales. A small group the other, was negative, as expected. The chi-square is of subjects (n=3) with extreme high externality / low 968.82, df= 458, and the RMSEA is .072, which is a internality configuration (in relation to the norm = here the reasonable fit. mean-scores of the total sample overall) was not included in this analysis. A 4-cluster solution provided good fit to Description of Sample in Terms of Locus of Control the data and can be described in relation to the meanA MANOVA showed that there were no statistically scores of the total sample in a 2 x 2 grid: significant differences between groups in the current High Internality sample in levels of locus of control across for schools, High Externality Cluster 2 classes and sex in the I-SEE primary scales (I-SEE SKI (High I/E) and I-SEE PC), or in the I-SEE secondary scales (I-SEE Low Externality Cluster 3 SKI and I-SEE PC). There were also no statistically (High I / Low E) significant interaction effects. Low Internality Cluster 1 (Low I / High E) Cluster 4 (Low I/E) An ANOVA showed that the between-cluster differences on I-SEE scores are statistically significant For I-SEE I [F (2, 206) = 7.017, p<0.005], I-SEE C [F (2, 206) = 114.549, p<0.001], I-SEE P [F (2, 206) = 109.313, p<0.001], and I-SEE SK [F (2, 206) = 67.602, p<0.001]. Table 5 shows the means and standard deviations of I-SEE scores for the final I-SEE clusters. A discrepancy analysis was carried out by crosstabulation of school / class and I-SEE Cluster. There are statistically significant discrepancies between observed Measures of Motivated Behaviour Table 3 shows that schools 1 and 2 have higher levels of CES E than school 3 does. Schools 1 and 2 also have higher levels of task-completion than school 3. School 1 has higher levels of participation than schools 2 and 3. There were statistically significant differences between groups in levels of motivation between classes within schools and between schools. [Class x School: Wilks 5 Self-Concept Research: Driving International Research Agendas by environment or personality. The self-report measure of CES engagement was most dependent on I-SEE cluster membership, there were no significant between school differences on this measure. The teacher rating (participation) was largely a function of school (no significant between I-SEE cluster differences). Both personality and school influenced task completion. There might be school-wide policies on homework completion. At the same time task-completion depends on behaviours and environmental factors outside of school. Participation also showed a weak interaction effect. This suggests that findings of investigations about the influence of personality and environment on motivation might differ depending on how motivation is assessed. These results illustrate some of the complexities in investigating motivation and academic achievement. Within this study a measure of academic achievement was obtained by School Certificate results in English. The cumulative effect of experiences in one particular school in interaction with a generalised measure of personality (here locus of control) might be reflected in such performance measures, which are the result of behaviour and action across a larger time frame and in a number of contexts. Therefore an ANOVA was carried out to test for statistically significant between-group effects and possible interaction effects for School Certificate results (by I-See cluster and school). There are statistically significant main effects for school [F (2,166) = 3.83, p<0.05] and I-SEE cluster [F (2,166) = 3.268 p< 0.05] but no interaction effects. Table 6 shows the means and standard deviations for School Certificate results English for the I-SEE clusters. The ‘average’ cluster has the highest levels of School C Certificate results, closely followed by the internal cluster. Consistent with theoretical expectations the external cluster has lower scores of school Certificate results. This result also supports the argument that the most adaptive locus of control configuration is one of realistic internality and externality rather than extremes on either measure. Thus while both, the environment of instruction (school type) and personality (generalised locus of control) impact on pupil’s motivation, no interaction effects were observed. The environment has a much more powerful effect than this generalised personality measure, and this might be one reason why no interaction effect was observed. and expected frequencies in the distribution of students by locus of control profiles among schools and classrooms (Chi-square = 33.92, df = 22, which is significant at p < 0.05.). Two classes in school 2 had significantly more external students than would be expected (critical ratios = 6.16 and 6.50 for the external cluster in the two classes respectively - any z-score > 3.84, is significant at p< 0.01) . Motivated Behaviour by I-SEE Cluster by School There were statistically significant differences in measures of motivation by school and I-SEE Cluster membership [School: Wilks Lambda = 0.799, F (6,376) = 7.243 p<0.001; I-SEE cluster: Wilks Lambda = 0.916, F (6,376) = 2.793, p<0.05; School by I-SEE cluster: Wilks Lambda = 0.900, F (12,376) = 1.686, p<0.10]. Univariate Analysis of variance showed that schools differed significantly on measures of Participation [F (2, 199) = 4.307, p<0.05] and task-completion [F (2, 199) = 20.588, p<0.001]. I-SEE clusters differed significantly on measures of task completion [F (2, 199) = 3.481, p<0.05], and CES engagement [F (2, 199) = 3.489, p<0.05]. There is a weaker interaction effect for task completion [F (4, 199) = 2.328, p= 0.058] and participation [F (4, 199) = 2.194, p= 0.071]. Figures 2 and 3 show the means for measures of motivation (task-completion and levels of participation) by school and I-SEE cluster membership. For Taskcompletion the schools vary in the extent to which there are between I-SEE cluster differences, with School 1 (=neutral) having consistently high levels of taskcompletion for all cluster-groups, School 2 (= low structure) showing some degree of between I-SEE cluster variability, and School 3 (= high structure) showing large between I-SEE cluster variability for task completion (range 64.62 – 85.83%). In School 2 the external I-SEE cluster has the highest levels of task-completion, and in School 3 it is the balanced I-SEE cluster, which has the highest levels of Task completion. For participation it is again School 3, which has the greatest between I-SEE cluster variability. School 1 and 2 have similar levels of variability between I-SEE clusters, but in School 1 it is the balanced group, which has the lowest level of participation, while in School 2 it is the internal cluster, which has the lowest levels of Participation. In School 3 the average cluster has the highest, and the external cluster the lowest levels of participation. These results showed that both the environment (schools) and personality (generalised locus of control) impact on pupil’s motivation. Generally the environment has a more powerful effect than the generalised personality measure. Only a weak interaction effect was observed between I-SEE cluster membership and school. The highly structured school appears to accommodate less well for diversity in terms of locus of control orientation, while the school judged to be neutral on this dimension shows the least extreme between I-SEE differences in one of the measures of motivation employed (Task completion), as well as fostering higher levels of motivation in general. It is also apparent that the different measures of motivated behaviour differ in the extent to which they are influenced Conclusions The AMP is based on the assumption that actiontheoretical personality variables described here develop as a result of learning processes in specific person-situation interactions. Personality variables develop by generalisation. They represent ‘summary terms’ for consistently applied or observable behaviours. The model can help explain why at times (or only in certain environments) generalised locus of control is predictive of outcomes and behaviour and why we can see differences in locus of control configurations within similar populations in different contexts. The purpose of this study was to evaluate the power of this action model of control which states that action is the 6 Self-Concept Research: Driving International Research Agendas result of a dynamic interaction between a person and his / her meaningful environment. A further aim was to test a new multi-dimensional locus of control instrument in relation to academic achievement and motivation. The ISEE measures generalised locus of control, and selfconcept of own ability (SK). The Internality scale combined with the SK scale constitutes a measure of selfefficacy beliefs. These constructs are integrated into a theoretical model of personality and action (the AMP). The context for testing the new I-SEE locus of control instrument was an investigation of the relationships between motivated behaviour, academic achievement and locus of control for a group of 16 – 17 year old students in three very different types of schools. It was felt that this context would be a good test of the model and the instrument because both locus of control and the AMP include both personality as well as environmental factors as determinants of development and action. This study confirms previous findings in the literature that there is a relationship between locus of control and academic achievement. This multi-dimensional measure suggests, however that high externality might have a detrimental effect on academic achievement, rather than high internality having a beneficial effect. In addition Krampen’s (1985) expectation that non-extreme control orientations might be associated with more realistic expectations and hence more adaptive outcomes is supported by the findings reported here. The results reported here also confirm that locus of control is a multidimensional construct – people can be high on both internality as well as externality. The cluster analysis illustrates this as clusters included a high internality/externality cluster, and a low internality / externality cluster. This shows that low externality, for example, is not automatically associated with high internality and self-efficacy, as uni-dimensional conceptualisations would suggest. In addition, although internality (I) and self-concept of own ability (SK) covary, the three clusters are distinguished more by variations in levels of SK and externality (P and C) than levels of I. The range of scores on the I- scale between the three clusters is only 33.37 – 36.98, whereas the ranges of the other measures are much wider (SK: 27.31 – 38.47, C: 29.98 – 17.02, and P: 29.34 – 15.91). This gives strength to the argument that the SK scale is a useful addition in this instrument. In New Zealand, as well as elsewhere boys tend to achieve significantly less well than girls do, particularly in English (Borg, Falzon, & Sammut, 1995; Fergusson & Horwood, 1997). This is confirmed by the data reported here, where boys have significantly lower achievement levels in English than do girls, in terms of School Certificate English. There are no statistically significant sex differences in this data set in motivation or locus of control scores. There are also no interaction effects between sex and either school, or class-within-school in relation to any of these measures. It is not levels of control that account for these differences. These findings are congruent with findings reported in the literature. Kalechstein and Nowicki (1997) in a meta-analytic review concluded that on the whole there are no sex differences in locus of control, nor are there sex differences in the relationship between locus of control and achievement at secondary school age. The schools were selected to maximise the likelihood of sampling diverse environments. The three schools differed, according to expert judges, on the dimensions of structure, competitiveness and co-operation. School 3 was judged to be the most structured, competitive and least cooperative, and School 2 the least structured, competitive and most co-operative. School 1 was judged to be somewhere in between schools 2 and 3 on all dimensions. Evidence for school-wide variables that operate are the between-school differences in School Certificate English results. School 1 has significantly higher School C results than School 2. School 1 also has the highest levels of motivated behaviour of all schools lending support to the contention that a sufficiently structured (rather than extreme) environment is most likely to foster adaptive behaviour and development. Such a small sample, however, mitigates against any strong conclusions. Even though pupils at School 3 are comparable in terms of ‘ability’ (previous School Certificate English) to pupils in School 1, pupils in school 3 exhibit lower levels of motivation than pupils in school 1. It is the non-extreme school judged to have medium levels of structure, competitiveness and co-operation that was associated with higher levels of motivation and achievement than the other schools. These findings suggest that school type may have a direct influence on classroom behaviour. The theoretical expectation was that generalised locus of control should have a greater impact on classroom behaviour in relatively unstructured (novel – unpredictable) schools (Krampen, 1988, 1991). In such schools externals would be expected to be disadvantaged and evidence less adaptive behaviour. Alternatively, in schools where the high ‘structure’ is associated with a high degree of perceived external control (strong extrinsic incentives to perform, perceived as controlling), internals might be disadvantaged. The ideal environment should be one that is sufficiently structured to be predictable, but not so structured as to limit action-alternatives unduly (Krampen, 1991). In this study, a weak interaction effect between school and locus of control in determining motivated behaviour in the classroom does not support the expectation that in a highly structured school locus of control should be less predictive of behaviour and that such a setting, should be preferred by externals. The school judged to be highly structured was associated with the largest between-group differences in motivated behaviour by locus of control cluster. In addition for participation (the most immediate measure here of classroom behaviour) externals had the lowest levels at school 3, (the school rated as highly structured) compared to the other two loci of control clusters at this school, as well as in comparison to externals at the other two schools. That is at school 3 members of the external locus of control cluster were likely to participate less than members of the other locus of control clusters as well as other externals in other schools. There was no difference between schools 1 and 2 in terms of the between-group differences by locus of control in their levels of participation. 7 Self-Concept Research: Driving International Research Agendas There is support for the negative effect of external control on motivation in locus of control internals. Internals in School 3 had the lowest levels of task completion, both compared to members of the other locus of control clusters in school 3 as well as compared to other internals in the other two schools. One reason why School 3 was judged as highly structured could be because of tight and controlling homework policies. The kind of factors considered to represent ‘structure’ might not be factors that make action – outcome relationships more predictable for students. Interestingly, for task completion it is School 1 which is associated with both, the highest levels of task completion overall as well as practically no between group differences by locus of control cluster in measures of task completion. This strengthens the argument that some degree of structure is beneficial. School 1 might have elements of structure that make outcomes predictable, or is structured in such a way as to be predictable enough without limiting action alternatives unduly. In conclusion, the significant school-level differences in environment, and the significant school-level effects in terms of locus of control, motivated behaviour and achievement mean that the effect of interventions, or other classroom-level research across several schools need to be interpreted with caution. School-level effects could interfere with the efficacy of interventions, or compromise change at the classroom-level. The results reported here also suggest that environmental interventions can override personality attributes that might disadvantage some students. On the whole the environmental variables had a more powerful effect on academic achievement and related behaviour than personality (here locus of control) did. Furthermore, some school-environments eliminate between group differences on those measures by personality type. These findings suggest that it is better to direct energy and effort at environmental interventions to create school environments that foster achievement and motivation for all and minimise between group differences in achievement and related behaviours than attempt to change student’s control orientations. Future research should look to identify the ideal school environment to foster motivation and achievement as well as the development of adaptive personality orientations. The AMP could be useful in guiding such research. Dr. Richard J. Hamilton is co-director of the Research Institute for Interventions in Teaching and Learning (RCITL). His research interests include issues around motivation in classrooms and teacher education. Contact Details Dr .Angelika Anderson RCITL, School of education The University of Auckland Private Bag 92019 Phone: + (09) 373 7599 x3043 Fax: + (09) 367 7191 Email: an.anderson@auckland.ac.nz References Arbuckle, J. L. (1995). Amos for Windows. Analysis of moment structures. Version 3.5. Chicago: SmallWaters Corp. Boggiano, A. K., Main, D. S., & Katz, P. A. (1988). Children's preference for challenge: the role of perceived competence and control. Journal of Personality and Social Psychology, 54(1), 134 - 141. Borg, M. G., Falzon, J. M., & Sammut, A. (1995). Age and sex differences in performance in an 11-plus selective examination. Educational Psychology, 15(4), 433-443. Fergusson, D. M., & Horwood, L. J. (1997). Gender differences in educational achievement in a New Zealand birth cohort. New Zealand Journal of Educational Studies, 32(1), 83-96. Finn, J. D., & Rock, D. A. (1997). Academic success among students at risk for school failure. Journal of Applied Psychology, 82(2), 221 - 234. Kalechstein, A. D., & Nowicki, S., Jr. (1997). A metaanalytic examination of the relationship between control expectancies and academic achievement: an 11year follow-up to Findley and Cooper. Genetic, Social, and General Psychology Monographs, 123(1), 27 - 56. Krampen, G. (1985). Zur Bedeutung von Kontrollueberzeugungen in der klinischen Psychologie. Zeitschrift fuer Klinische Psychologie, 14(2). Krampen, G. (1987). Differential effects of teacher comments. Journal of Educational Psychology, 79(2), 137 - 146. Krampen, G. (1988). Toward an action-theoretical model of personality. European Journal of Personality, 2, 39 55. Krampen, G. (1991). Fragebogen zu Kompetenz- und Kontrollueberzeugungen (FKK), Handanweisung (1 ed.). Goettingen: Hogrefe Verlag fuer Psychologie. Krampen, G., von Eye, A., & Brandtstaedter, J. (1987). Konfiguration generalisierter Kontrollueberzeugungen. Zeitschrift fuer Differentielle und Diagnostische Psychologie, 8(2), 111 - 119. Krampen, G., & Wieberg, H.J.W., (1981). Three aspects of locus of control in German, American, and Japanese university students. The Journal of Social Psychology, 113, 133 - 134. About the Authors Dr. Angelika Anderson recently graduated with her PhD, which involved the exploration of a dynamic and interactive model of personality and action, supervised by Professor John Hattie and Dr. Richard Hamilton. She also has interests in issues around inclusive provision for special education needs, with a focus on catering for a more diverse student population in mainstream schools. Professor John Hattie is the head of the School of Education at the University of Auckland. His research interests include issues of assessment and evaluation, and meta-analytical examinations of key issues around teaching effectiveness. 8 Self-Concept Research: Driving International Research Agendas Lefcourt, H. M. (1982). Locus of Control (2nd ed.). Hillsdale: Lawrence Erlbaum Associates. Levenson, H. (1981). Differentiating among internality, powerful others, and chance. In H. M. Lefcourt (Ed.), Research with the Locus of Control Construct (Vol. 1, pp. 15 - 63). New York: Academic Press. Lewin, K. (Ed.). (1952). Field Theory in Social Science. London: Tavistock Publications TTD. Parent, J., Forward, J., Canter, R., & Mohling, J. (1975). Interactive effects of teaching strategy and personal Locus of Control on student performance and satisfaction. Journal of Educational Psychology, 67(6), 764 - 769. Rotter, J. B. (1954). Social Learning and Clinical Psychology. New York: Prentice-Hall, INC. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied, 80(1), whole number 609. Skinner, E. A., Wellborn, J. G., & Connell, J. P. (1990). What it takes to do well in school and whether I've got it: a process model of perceived control and children's engagement and achievement in school. Journal of Educational Psychology, 82(1), 22 - 32. Stipek, D. J., & Weisz, J. R. (1981). Perceived personal control and academic achievement. Review of Educational Research, 51(1), 101 - 137. Trice, A. D. (1980). Students' locus of control and ratings of a structured educational environment. Psychological Reports, 46, 782. 9 Self-Concept Research: Driving International Research Agendas Table 1 I-SEE Items by Scale Item Scale I-SEE I 1 Whether or not other people respect my wishes is mostly up to me. 5 Whether or not I have an accident depends entirely on my behaviour. 6 When I make plans, I am almost certain to make them work. 11 I can do a lot to protect myself from disease. 23 I can pretty much determine what will happen in my life. 25 I am usually able to protect my personal interests. 27 When I get what I want, it’s usually because I worked hard for it. 30 My life is determined by my own actions. I-SEE SK 4 Sometimes I feel like I have no ideas and don’t want to do anything. 8 I don’t like ambiguous situations, because I don’t know how to behave or what to do. 12 I often don’t know what to do to make my wishes come true. 16 I know many ways of protecting myself from diseases. 20 In unclear or dangerous situations I always know what to do. 24 Sometimes I just don’t know at all what to do in a given situation. 28 I can usually think of many alternative ways of dealing with even difficult situations. 32 I can usually think of many ways of solving problems. I-SEE P 3 I feel like what happens to me in my life is mostly determined by powerful people. 10 Other people often prevent my plans from becoming reality. 14 My life is chiefly controlled by powerful others. 17 I have very little chance of protecting my personal interests when they conflict with those of other people. 19 Getting what I want requires pleasing those people above me. 22 My wellbeing depends to a great extent on the behaviour of other people. 26 Whether or not I have an accident depends to a large extent on the behaviour of others. 29 In order to have my plans work I make sure that they fit in with the desires of people who have power over me. I-SEE C 2 To a great extent my life is controlled by accidental happenings. 7 Often there is no chance of protecting my personal interests from bad luck happenings. 9 When I get what I want it’s usually because I’m lucky. 13 Much of what happens to me in my life is a matter of coincidence. 15 Whether or not I have an accident is mostly a matter of luck 18 It’s not wise for me to plan too far ahead because many things turn out to be a matter of good or bad fortune. 21 It is sheer coincidence when somebody else ever considers my wishes 31 Whether I fall ill or not is a matter of fate. 10 Mean 4.04 3.58 4.40 4.65 3.29 4.55 4.79 5.00 3.40 3.58 3.73 4.63 3.92 3.63 4.38 4.49 2.63 2.96 2.43 2.93 sd 1.39 1.62 1.21 1.24 1.64 1.07 1.20 1.09 1.69 1.41 1.49 1.25 1.24 1.31 1.09 1.11 1.44 1.41 1.43 1.33 3.31 2.95 3.12 1.48 1.28 1.41 3.35 1.36 2.73 3.28 2.75 2.85 2.90 3.11 1.29 1.46 1.45 1.34 1.43 1.52 2.76 3.47 1.32 1.61 Self-Concept Research: Driving International Research Agendas Table 2 Description of Scales Scale No of Items means sd α I-SEE Internality 8 34.30 5.52 0.62 I-SEE Self Concept 8 31.56 6.32 0.73 I-SEE Powerful Other 8 23.50 6.66 0.74 I-SEE Chance 8 23.73 6.48 0.70 I-SEE PC 16 47.2 11.52 I-SEE ISC (selfefficacy) 16 65.9 10.11 Table 3 Means and Standard Deviations of Measures of Motivation School CES Completion means sd means sd Participation* Means sd 1 16.68 2.64 96.72 8.86 2.99 1.57 2 16.57 2.26 92.25 19.64 3.48 1.43 3 15.65 2.48 78.93 23.57 3.51 1.91 All schools 16.29 2.50 89.00 20.07 3.33 1.68 Note that Participation is reverse scored: a low number indicates a high level of participation. Table 4 Means and Standard Deviations for Previous School Certificate Results School Girls Boys All Means SD Means SD Means SD 1 63.61 17.80 55.29 17.34 60.37 17.93 2 58.43 15.18 45.93 12.11 52.29 15.02 3 62.44 15.89 55.00 12.08 59.15 14.70 All schools 61.66 16.35 51.81 14.28 57.32 16.19 11 Self-Concept Research: Driving International Research Agendas Table 5 Means and Standard Deviations of I-SEE Scores for Final I-SEE Clusters Cluster 2 (n=59) external 3 (n=107) balanced 4 (n=43) internal Overall (n=209) I-SEE I 33.37 3.91 34.22 5.44 36.98 5.03 34.55 5.11 I-SEE SK 27.31 4.31 22.93 4.70 38.47 4.76 31.80 6.14 I-SEE C 29.98 4.17 22.60 4.48 17.02 4.35 23.54 6.32 I-SEE P 29.34 4.74 22.93 4.70 15.91 3.82 23.30 6.49 Table 6 Means and Standard Deviations of School Certificate Results by I-SEE Cluster I-SEE Cluster School Certificate Means SD 2, external 51.60 15.70 3, average 60.52 15.61 4, internal 57.91 16.21 Total 57.41 16.13 12 Self-Concept Research: Driving International Research Agendas Figure 1: Path Diagram for I-SEE Scales FKK4 FKK30 0.37 0.53 FKK27 0.36 0.74 FKK25 0.58 0, 0, 0.43 FKK23 FKK8 0.45 FKK12 0.4 I-SEE SK I-SEE I 0.4 0.47 0.55 FKK20 FKK11 0.56 FKK6 FKK16 0.56 FKK24 0.13 0.65 0.38 FKK5 0.69 FKK32 FKK1 -0.33 -0.42 -0.44 -0.14 FKK31 FKK28 FKK3 0.31 0.58 FKK21 0.54 0.53 FKK10 FKK18 0.6 0.67 FKK14 FKK15 0.46 0.58 FKK13 FKK9 FKK7 0, 0, FKK17 I-SEE P I-SEE C 0.63 FKK19 0.48 0.46 0.47 0.36 FKK22 0.76 0.35 0.38 0.49 FKK26 FKK29 FKK2 13 Self-Concept Research: Driving International Research Agendas Figure 2: Levels of Task-Completion for I-SEE Clusters by School 100 95 % Task Completion 90 85 Balanced 80 External 75 70 65 Internal 60 1 2 3 School 14 Self-Concept Research: Driving International Research Agendas Figure 3: Levels of Participation for I-SEE Clusters by School 4.5 External Rating of Participation 4 Internal 3.5 3 Balanced 2.5 2 1 2 3 School 15