Construct validity of success / failure attribution scale among Turkish university students

The aims of this study is to examine the construct validity of the success / failure attribution scale by using exploratory and confirmatory factor analysis, to assess the internal reliability of the scales and to determine the sub-dimensions of the scales by performing structural equation modeling. In the study, success and failure attributions scale are discussed as two separate scales and analyzed separately. Findings from the exploratory factor analysis has revealed three-factor structure for both failure and success attributions. The results of the confirmatory factor analysis (CFA) provided evidence for the three-factor structure of both scales. Finally, It was found that university students tend to attribute their success or failure generally to factors within themselves


INTRODUCTION
Motivation which is one of the factors that determines the continuity and vitality of student participation in the course also provides the continuity of students' attention to the course.The absence of motivation leads to learning difficulties and deficiencies, and an increase in discipline problems in the classroom.For this reason, motivation comes first of the issues to be seriously considered.
As it is in the learning theories, the theories in order to explain what the motive is, are much and they varied.Attribution theory comes first among the theories frequently performed in motivation studies (Pintrich andShunk, 2002, quoted by Schunk, 2011).The term attribution is related to the perception of the causes of an event or result by the individual.Attribution research focuses on the ways used by the individuals while practicing their belief and reaching to causal explanations.Attribution theorists are concerned with the perceived causes of events (Gredler, 1992).Attribution theory explains how the people see the causes of both their own and the other people's behaviors (Weiner, 1985(Weiner, , 2010) ) and investigates how an individual's explanation type of his own success or failure effects motivations (Erden and Akman, 1997).
The most popular attribution theory today is the one developed by Weiner (Açıkgöz, 2012).In Weiner model, it is suggested that the individuals attribute their successes and failures to the causes within or outside themselves and when they encounter similar situations in the future, they think that they will again take the same result (Fidan, 1985).According to Weiner people attribute their success or failure to four causal factors and each reason has two dimensions.

Causal factors:
In order to determine the causes of success and failure, Weiner suggested four typical attributions as skill, effort, difficulty level of the task and chance (Weiner, 2010).For example, if Ayşe gets five in social studies exam, she can attribute this result to her capability ("I'm good in social studies course" ), effort ("I worked very hard for this exam"), the difficulty of the task ("the exam was not too difficult") or chance ("questions were the subjects that I studied").Other attributions include the mood, illness, fatigue, physical appearance, personality and getting help from others (Gredler, 1992;Schunk, 2011).Talent and effort for success and failure are seen as the most dominant reasons.While success is attributed to working hard and / or ability, failure is attributed to the lack of effort / or talent (Gredler, 1992).
Causal dimensions: Weiner (2010) also suggested that there were the results of the causal attributions of the people to reflect their personal history, social norms and causal thinking rein the theory of people's causal attributions their personal history, social norms and causal thinking and investigated causal attributions classification for success and failure in three dimensions studied.These are causal focus that (internal-external) cause focuses inside or outside of the individual, determination (stable -unstable) cause remain the same over time or relatively changes or not, and the controllability (controllable -uncontrollable) that the individual can or cannot control (Açıkgöz, 2012, Gredler, 1992;Perry et al.,1993;Schunk, 2011;Stipek, 1998;Woolfolk, 2001).

Internality -externality dimension:
The first dimension of Weiner's classification is internal-external or control source.It is the feature that internality X externality attributions are related to factors inside or outside of a person (Açıkgöz, 2012).If a student attributes the reason of his failure in the exam to the lack of effort, the reason will be classified internally because efforts will be seen as a reason inside the individual.If a student sees the cause of failure (task) as the difficulty of the questions, the cause will be classified as externally because the reason is outside the individuals.This dimensional can be handled together with the determination dimension.

Stability dimension:
It is the dimension affecting the future goals and hopes of an individual (Grad of 1992) and it is seen to have a close relationship with future expectations (Woolfolk, 2001).It is a feature about whether the attributions in stability dimension changes or not over time.Some attributions may vary over time (Açıkgöz, 2012).For example, if a student is attributed with success or failure against a difficult exam, his future success or failure is expected in more difficult exams (Woolfolk, 2001).Talent does not change relatively or intrinsically meanly, it is stable.Efforts are internal, but variable; if an individual wants, he can study willingly or unwillingly.The difficulty of the task is external, so it does not change relatively because the task's status does not vary from one moment to another; the chance is external and variable, while chance is smiling to a person on one hand, his chance may rotate the next moment (Schunk, 2011).

Controllability dimension:
The third dimension of the attributions is controllability dimension.It is related to being significant or not of a result's cause as controllable (Woolfolk, 1993).Effort is a attribution that can be controlled.The cause of uncontrollable effort depends on the difficulty of the task since the student can not control how difficult an exam is which is made by the teacher.The ability attribution is classified as stable, internal and uncontrollable (Açıkgöz, 2012;Gradler, 1992;Graham, 1997;Peterson and Berger, 1983;Woolfolk, 2001).Even if the attributions are identical internality/externality or stability/instability, they may be different in terms of controllability.For example, while the attributions of "not feeling well" in math class is internal, stable and uncontrollable, the attribution of "not studying sufficiently" that can be thought as a failure cause again is an internal, stable and controllable attribution (Açıkgöz, 2012).Four typical attributions explaining success and failure and their explanations are shown in Table 1.
Attributions are the cognitive and affective consequences shaping the subsequent behavior of individuals and their future potential for success or failure (Weiner, 1985;2010).Besides each of the attributions are the features that cause special effects.Causal focus is associated most with individual's self-confidence.Reasons will either enhance the sense of self-worth or vice versa.If the success is attributed to internal reasons, it leads to boasting, but the failure leads to condemnation, inability situations.If it is attributed to external factors, the success leads to thanks and gratitude emotions; failure leads to the resentment and anger (Fidan, 1996).The feature of controllability constitutes emotions in two ways.First of all, the attributions under the control of the individual lead to feelings of ability or guilt.The second one is reward, punishment and solidarity behaviors create delightfulness (Gredler, 1992).Cognitive and affective responses resulting in success and failure are as follows: 1. Regarding talent (internal and uncontrollable factors) A growing boast in success comes out.Failure may lead a growing sense of lack and self-condemnation.2. Regarding the efforts (internal but controllable factor) It leads to an increasing learning in success and taking credit.If the student increases the efforts, he is thought to be even more successful.In the failure, the student disapproves and blames himself.The student is thought

General condition
The subject was difficult.The exam was difficult.
The difficulty of the task.
The things that I knew were asked.

CHANCE
Source.Stipek, 1983.to get a different success in future.3. Related to the topic difficulty (Uncontrollable external factor) It may reduce boast in success, the student may think he will get the same success in the future.In the failure, a decrease is seen in blaming and disapproving himself.The student thinks that he will get the same success in the future.4. Related to the chance (uncontrollable external factor) It has a role that decreases boasting in success.The student thinks that he will get a different success in the future.Many students try to explain the reasons of their failures.When the successful students are unsuccessful, they make internal and controllable attributions.For example, they express that they misunderstood the instructions, did not studied hard or did not have sufficient knowledge.
The student aware of these attributes the failure to controllable reasons and focuses on new strategies in order to be successful in a short time.This leads the student to being successful, proud and feeling well (Woolfolk, 2001).The attributions of unsuccessful students are generally external and related to the attributions; one of the worst possibilities is being internal, determinant and uncontrollable.For example, if a student thinks that his failure results from disability, he does not make any effort for being successful.It is very hard to motive such type of students (Açıkgöz, 2012).If the students think that they do not make an important thing against the failure, they may develop learned helplessness (Woolfolk, 1993).
Learned helplessness generally emerges when the students who experience lots of failures realize that the result is not affected by their own behaviors and can not succeed when they fail.(Açıkgöz, 2012;Stipek, 1983;Woolfolk, 1993).The reason of the learned helplessness is a student's giving up making an effort and directing to withdraw (Peterson and Berger, 1983).These students do not make an effort to learn and they easily give up when they encounter the difficulties (Stipek, 1983).The important one is to prevent students from experiencing learned helplessness.According to attribution theory, the motivation of the students who attribute the failure to internal and controllable variables are higher and they are more successful in their academic life (Erden and Akman, 1997).Therefore, determining the success or failure attribution of the students is very important to enable them being more successful and to prevent them from experiencing learned helplessness.
The researches on attribution and motivation focused on four attributions stated by Weiner (1985).Apart from these four attributions, it has been detected that motivation and success are affected by social factors such as parents and teachers (McClure et.al., 2011;Wentzel, 1998).Various researchers suggest that social attributions as peer effect, teacher and parents are important in education (Boruchovitch, 2004;Fraser and Killen, 2003;Lebedina-Manzoni, 2004;Lightbody and Siann, 1996;McClure et al, 2011).
Looking at the researches related to the attributions in Turkey, it is observed that these researches focus on Weiner's four attribution dimensions (Baştürk, 2012;Gülveren, 1996;Gürtekin, 1993;Kızgın and Dalgın, 2012;Sipahi, 1995;Yıldız, 1997), and the relationship success and failure attributions with emotional response (Aydin and Berberoğlu, 1990;Baş, 1998).Also there are some researches that investigated success and failure attributions on courses such as English and music, (Besimoğlu et al.,2010(Besimoğlu et al., , Özmenteş, 2012;;).The scale that will help to detect the perceptions about the causes of success / failure especially of the university students and also to identify social attributions is expected to contribute to further researches.The aims of this study were, 1.Failure Attribution Scale by using exploratory and confirmatory factor analysis 2. to assess the internal reliability of the scales 3. to determine the sub-dimensions of the scales by performing structural equation modeling

METHODOLOGY
The study group The study group of the research includes first and last grade students primary school department of Dokuz Eylül University Buca Faculty of Education during spring term of 2010 to 2011 academic year.The subjects with extreme values who did not fill one of the scales or filled it missing or incorrectly were excluded from the scope of research.559 students participated in the scale development part of the research for success attribution scale, 369 students participated for failure attribution.In the second aspect of the study, 369 students participated in both success and failure attribution scale.67.2% (n = 250) of the students who participated in the study were female and 32.8% (n = 241) of them were male.The distribution of the students according to the departments and classes are included in Table 2.

Success / failure attribution questionnaire
Scale was developed for the first time by Açıkgöz and Kasap in order to detect the success and failure causes of primary school and high school students elementary and high school students and used for Kasap (1996) and Sucuoğlu (2003)'s researches.The scale was developed and harmonized to university students.The scale is 5-point Likert-type and consists of two scales as Success Attributions and Failure Attributions.

Procedures
In order to obtain evidences about factor structure validity of success/failure attribution scale a three-stage routes were followed.The total test scores were calculated for the data obtained from primarily success attributions scale with 559 people then failure attributions scale with 369 people and whether there are test scores with too low scores and having negative correlations (r < .20 ) or not was investigated and no negative item with negative or low test score was found in both success and failure attributions scale.In the second stage, in order to identify how many dimensional structures the failure attribution scale with 33-item success and 33item failure attributions has, exploratory factor analysis was conducted.After receiving the results of factor analysis, reliability analysis was made separately for each dimension structure determined separately for each dimension by considering the factor structures determined (Şencan, 2005).Therefore, in order to determine the internal consistency in the scope of reliability studies, Cronbach's alpha coefficients were regarded.In the third stage, it was subjected to (confirmatory) factor analysis depending on the results of exploratory factor analysis and to prove the hypothesis that the relevant variables will take place mainly on the factors identified.Thus, the factorial validity of the scale has been tested with two factor analysis applications.Cronbach's Alpha coefficient and factor analysis studies of success/failure attributions scale was performed with statistical package for the social sciences (SPSS) software package.Confirmatory factor analysis was performed by using the software of linear structural relationships (LISREL) 8:53.

Exploratory factor analysis
The suitability of the data for factor analysis suitability was tested with Kaiser-Meyer-Olkin (KMO) coefficient and Barlett sphericity test and that KMO is higher than .60 and Barlett test is significant show the suitability of data for factor analysis (Büyüköztürk, 2005;Şencan, 2005).Accordingly, in the analysis on data, for factor analysis of 33-item success attributions scale, the Keiser-Meyer-Olkin's value was found to be .956and for the failure attributions, it was found to be .933.Similarly, Bartlett test result was found for success attributions as [x 2 = 12885.930,df = 528, p <, 000], and for failure attributions as [x 2 = 6558.247,df = 528, p <, 000] respectively.With these results, the data was assumed to come from multivariate normal distribution.This result shows that it is suitable for factor analysis.
In order to determine the structure validity of the scale, factor analysis was conducted and to determine how many important factors or structures the scale measures, the factors with 1.00 or above 1.00 eigenvalues were taken.In conclusion of rotation process made with principal component factor analysis and Varimax technique, the items with the eigenvalue that is bigger than 1.00 and factor load which is greater than .30were analyzed.In conclusion of factor analysis, five components whose eigenvalue is more than 1.00 were determined for 33item achievement scale.The contribution of these components to the total variance is 64.44%.When these five components identified were evaluated in the light of thier contribution to the total variance by examining both the total variance table, and slope -sediment graph (scree plot), it was seen that three components have important contribution to variance and after fourth component, they are both small and about the same.In the frame of all these information, it was decided to apply the analysis for three factors.When the line graph drawn based on the eigenvalues (scree plot) is examined, it is seen that the factor with high-acceleration and rapid declines show the number of important factors, and the breaking point occurred after the first factor.(As shown in Figure 1 and 2) The same process was followed for 33-item success attributions scale it was detected that there were six components with eigenvalues above 1.00 and the contribution of these components to the total variance that six components and their contribution to the total variance of these components were found to be 61.05%.By investigating the six components identified, the total variance table described and slope-deposit graph (scree plot) it was decided to repeat the analysis for three factors.These decisions made for both success attributions scale and failure attributions scale are seen as significant in terms of being compatible with the theoretical structure identified during the development process of the tool.
In conclusion of the analysis repeated for both the scales, it was seen that the contribution of the factors to the total variance was 34.49% for the first factor, 14.07 % for the second factor and 8.73% for the third factor and the total contribution of these three factors identified to the variance was 57, 29%.In failure attribution scale, the contribution of the factors to the total variance was detected as 20.18% for the first factor, 19.59% for the second factor, and 10.30% for the third factor, and the total contribution was 50.07%.
In the exploratory factor analysis performed in order to reveal the factor pattern of success attribution scale, the acceptance level for factor loading value was determined to be .32.In the success attributions scale, in the analysis performed for three factors, when the items were evaluated in terms of meeting the acceptance level of cyclicality and factor loading value, three items (24, 27 and 33) were seen as cyclical.By excluding these items from the analysis, factor analysis was repeated.In failure attribution scale, seven items were found to be cyclical (19,14,22,10,4,29,9).The factor pattern obtained in conclusion of being excluded of this item, factor loading value of the item and common factor variances are shown in Table 3.
In conclusion of the analysis, it was seen that the items described theoretically gathered under their own factors.When Table 2 is analyzed, the success attribution scale consists of three-factor structure.When examined subscales level, the factor attribution values for the subscale of "attribution to the teacher and the course" are between .470 and .849, the factor attribution values for the subscale of "attribution to the family" attribution are .696and .829,and the factor attribution values for the subscale of "attribution to himself" range from .545 to .713.
The common factor variance takes a value between 0 and 1.For an item common factor variance approaching to 1 shows that the contribution of indication to variance is high, approaching to 0 shows its being low and common factor variance indicates the ratio described by the factors (Çokluk et al., 2010).Based on this information, when factor attribution values are examined in terms of size, it is possible to describe attribution values from "good" to "excellent" except for two items (three and six).The attribution values of these two substances can be considered as "inadequate".
Following the items repeated for three factors and excluded from the analysis, the contribution of factors to total variance was found to be 1) 35.076% for the first factor, 2) 15.009% for the second factor, 3) 8.705% for the third factor.The total contribution of these three factors to the variance is 58.790%.
When Table 4 is examined, it is seen that failure attribution scale consists of three-factor structure.These subscales are 1) Attribution to the family, 2) Attribution to teachers and the course, and 3) Attribution to himself.
When it is examined on the level of sub-scales for the " attribution to the family " that is the first sub-scale, factor attribution value is between .518 and .861,for the subscale of "attribution to teachers and courses" the factor attribution value is between .388 and .797and for the sub-scale of "attribution to himself" is between .54 and .80.In line of information given before, when the failure attributions scale factor attribution values are examined in terms of size, it is possible to describe the item attributions as from "good" to "excellent" except for six substances (33 , 28 , 11, 16 , 2, and 3) are considered to be "inadequate".
In conclusion of the analysis repeated for both the scales, it was seen that the contribution of the factors to the total variance was 34.49% for the first factor, 14.07% for the second factor and 8.73% for the third factor and the total contribution of these three factors identified to the variance was 57, 29%.In failure attribution scale, the contribution of the factors to the total variance was detected as 20.18% for the first factor, 19.59% for the second factor, and 10.30% for the third factor, and the total contribution was 50.07%.
Following the items repeated for three factors and excluded from the analysis of the substances, it was seen that the contribution of the factors to the total variance was 1) 21.278% for the first factor, 2) 20.843 % for the second factor and 3) 11.009% for the third factor.
The total contribution of these three factors identified to the variance was 63.130%.In a multi-factorial design, the variance described is considered to be sufficient between 40% and 60%.(Çokluk et al., 2010;Tavşancıl, 2006).It is seen that the total contribution of described factors to the total variance is sufficient for both success attribution and failure attribution.

Confirmatory factor analysis (CFA)
In conclusion of exploratory factor analysis, three-factor structure was obtained for both success and failure attributions and in order to test the structural validity of these three factors, CFA was applied by using Lisrel 8.53 program.Confirmatory factor analysis is based on statistics of compliance.Compliance statistics is intended to elicit to what extent the predicted model overlaps with the observation data (Şencan, 2005).Compliance statistics are interpreted by using some kinds of limit values related to whether the model is regarded as acceptable or not.So the fix indexes produced in conclusion of the analysis are required to be above or below certain values.One of the most commonly used compliance chisquare (χ²).It is calculated by dividing χ² by degree of freedom.χ² value and being two or more lower of this value shows being good of the model, being five or lowershows being acceptable competence of the model (Şimşek, 2007) However, for some researchers, this ratio is in excellent competence with ≤ 2, and in bigger samples, it is in excellent competence with ≤ 3 and ≤ 5 is considered as a moderate level (Çokluk, Şekercioğlu and Büyüköztürk, 2010).When these values are examined, for success attribution scale χ ² / df = 1232.6/ 389 = 3.16 ≤ 5 mid -level competence, and for failure attribution scale for χ² / df = 756.97/281= 2.69 ≤ 3 is expressed as excellent competence since χ²/sd rate is below 3.
The other goodness of fit index statistics produced and frequently used apart from this competence suitability index (AGFI), controversial competence suitability index (CFI), mean square root of the predicted error (RMSEA), residual mean square root (RMR) and standardized fit index (NFI) (Şimşek, 2007).Being below .05 of RMSEA is expressed as excellent competence (Kline, 2005;Çokluk et al., 2010), and between .05 and .08 is as good competence (Browne and Cudeck, 1993;Sümer 2000) and between .05 and .100 is as acceptable competence (Weston and Gore, 2006;Hair et al., 1998) and between .08 and .100 is medium level competence MacCallum et al. 1996, Quoted by: Byrne, 2001) and 100 and above is as poor competence (Browne ve Cudeck, 1993;Sümer 2000).Accordingly, when RMSEA is investigated, it can be said that (.07) for success attribution scale and (.06) for failure attribution scale are in good competence.
Being .90 and above of AGFI is accepted as the indicators of good fit (Hu and Bentler , 1995;Sümer, 2000;Hooper et al.,2008), .95 is a perfect fit ( Sümer, 2000;Schumacker and Lomax, 2004;Hooper et al. 2008, multiplicity, Şekercioğlu andBüyüköztürk, 2010).Being bigger than .90 of GFI and CFI is accepted as the indicator of good fit, being bigger than .95 is an excellent fit value (Çokluk et al., 2010).Accordingly, for success attributions GFI = .82, AGFI = .78,CFI = .87were calculated.GFI, AGFI and CFI indexes were not found sufficient.However when the explanations above are taken into consideration, the model can be interpreted since the statistic show on acceptable fit.

Internal reliability
In order to investigate the reliability of the dimensions investigated with DFA, item total correlations were calculated with alpha internal consistency reliability coefficients.For this, the difference between item mean scores of sub-group with 27% and top groups with 27% that was made up of due to the total scores of the scale which were tested by using unbound t-test.That the observed differences between the groups in the desired direction were significant is evaluated as an indicator of internal consistency of the test.Analysis results indicate that to what extent the item differentiate the individuals in terms of measured behavior (Büyüköztürk, 2005).In this context, primarily for success attributions scale, it is identified as N = 559 total points based on 27% (n = 151) and bottom 27% (n = 151) parent group , the failure load for N = 369 for the total score compared to 27% (n = 100) and lower 27% (n = 100).The study is based on the statistical significance of .05 .The analysis results are shown in Table 3 and Table 4. Accordingly, corrected item-total correlation values in success attribution scale are between .21 and .77,corrected item-total correlation values in failure attribution scale ranged from .36 to .67.The t-test results made between the scores of top group with 27% and sub-group with 27% show there is a difference for total scores of all the items and sub-scale in both success attribution scale and failure attribution scale.
Looking at the internal consistency coefficient of success attribution scale that consist of 30 items and three sub factors, reliability predictions were obtained and the total internal consistency coefficient of the scale was found as .956.The internal consistency coefficient of the scale are as follows: for the sub-factor of "attribution to the teacher and course" it was founded as .96,for the sub-factor of "attribution to the family" as .88,for the subfactor of "attribution to himself" as .68.These values are seen as acceptable values for the reliability level of success attribution scale.The total internal consistency coefficient of failure attribution scale that consists of 26 items, founded to be .92.The internal consistency coefficient of sub-dimensions of the scale was founded as; for the sub-factor of "attribution to the family" as .90, for the sub-factor of "attribution to the teacher and course" as .88,and for the sub-factor of "attribution to himself" as .75.These values are seen as acceptable values for the reliability level of failure attribution scale.

Second level conformatory factor analysis of success/failure attribution subscales
In order to identify what the students attribute their success and failures to, that is the third sub-problem of the research, structural equation modeling was performed.For this, Kolmogorov-Smirnov test was performed in order to prove the normal distribution of the scores first of all since it will be studied in a parametric pattern.In this test, p-value = 0.00 <  = .05and none of the variables in both success and failure scale are normally distributed.Since the assumption of normality could not be maintained Maximum Likelihood method was used while making analysis in LISREL program.Therefore asymptotic covariance matrixes were used in the analysis.In order to make structural equation modeling, multiple linear scale test of success and failure attributions scale is required.In this case, multiple linear connection problem among data was eliminated by making factor analyzing to the data.The results of the factor analysis yielded following three factors for the Success Attributions Scale:

Variable comparison between the variances
The size of the difference between the variances of all the variables were controlled, but since the number of variables is too much, and the binary combinations are more, F-test and Levene test results were given due to the way of example in terms of b1 and b2 variables variance comparison.According to this result, F-test (pvalue = .838)and Levene's test (p-value = .47)were calculated for success attributions, and F-test (p-value = .838)and Levene's test (p-value = .47)were calculated for failure attributions.No significant evidence was found about being unequal of the both variables variance for both success and failure attributions scales.Considering the combination of all the variables, the same results were obtained.

The second level confirmative factor analysis
The attribution to teachers and course, the attribution to family and the attribution to himself are the different components of success attribution variable that is a higher level structure.Figure 3 gives a path diagram for the three factors of the success attribution scale.The findings obtained as a result of second level confirmatory factor analysis are given in Table 6.When Figure 3 and Table 6 are examined, the success attribution seems to be associated mostly with the attribution to himself.Variability is described mostly the attribution to himself (.99), at least for the attribution to family (.33).In this case, it can be expressed that the success attributions to the family has weak relationships with success attributions variables.Failure attributions variable is a structure at a higher level having the components of family-oriented attribution, attribution to the teacher and the course and the attribution to himself.Figure 4 gives a path diagram for the three factors of the failure attribution scale.The findings obtained as a result of Second Level Confirmatory factor analysis are given in Table 7.When Figure 4 and Table 7 are examined, the success attribution seems to be associated mostly with the attribution to himself.Variability is described mostly to the attribution to himself (.75), at least for the attribution to family (.41).As a result, when students fail, they tend to attribute this failure to themselves first, then to the teacher and the course and finally to their family.

DISCUSSION AND SUGGESTIONS
The aim of this study is to determine the distribution of the expressions to sub-dimensions in success / failure Attribution scale that helps to detect the perception of the students related to success / failure reasons and at what level it explains success and failure attribution of the subdimensions by using exploratory and confirmatory factor.In the study, success and failure attributions scale was discussed as two separate scales and analyzed separately.EFA performed in the scope of validity studies of the scale has revealed three-factor structure that explain 58,79 % of the total variance for success attributions scale and three-factor structure that explain 53,13 % of the total variance for failure attributions scale.In order to test the three-factor structure obtained as a result of EFA, CFA was applied and for success attributions, it was calculated as GFI = .82;AGFI = .78;CFI = .87ve RMSEA = .077.These results are in acceptable compliance for success attributions.As for the failure attributions, it was calculated as GFI = .86,AGFI = .83,CFI = .90and RMSEA = .068.Since it is CFI .90,model is compatible to be accepted.The internal consistency reliability coefficient of success attributions scale was found to be .95and for failure attribution scale, it was .92.These values are said to be acceptable values for the reliability of the scales.

CONCLUSION
In the study, it was also revealed what results the subdimensions of success attribution and failure attribution predicted in success / failure attributions scale as a result of the analysis comes out.Variability is described mostly the attribution to himself (.99), at least for the attribution to family (.33).In this case, it can be expressed that the success attributions to the family has weak relationship with success attributions variables.In the same way, when we look at the failure attributions, the variability was observed for the attribution the himself at most (.75 ) and at least for family attribution (.41) in the variable.It can be said that the the relationship between attribution variable and failure attribution top variable is strong and its relationship with the attribution variable is at moderate level.
According to these results, it can be stated that students hold firstly themselves responsible for both success and failure cases.The students tend to attribute their success to chance ("My chance is going well") and to getting help from others ("There is the one who can help my lessons").The students attribute their failures to incompetency, not being sociable, bad background, not studying hard and to believing to be unsuccessful.These results are consistent with the classification of Weiner.If a person thinks the reason of his failure as a lack of effort (not studying hard), the reason will be seen as intrinsic.The attributions such as lack of ability, believing to be unsuccessful are    related to the individual himself, meanly they are intrinsic (Schunk, 2011;Slavin, 2012).The results attributed to stable and uncontrollable reasons such as ability weaken the motivation and lead to learned help-lessness at extreme situations (Abramson et al., 1989;Abramson et al., 1978;Gibb andAlloy, 2006, Quoted by: McClure et al., 2011).For this reason, it is required to investigate whether these students experience learned helplessness or not and whether they internalize their failure or not.In the new researches to be done, the cognitive and emotional responses that come out in success or failure should be determined.
It is revealed that the students attribute their success and failure to teacher and the course after themselves and their tendency to attribute them to their parents comes last.The reason of their attribution to their parents last can be connected to the fact that the students are the university students and live apart from their parents.These results are consistent with previous studies.In studies with younger age groups (Painsi and Parncutt, 2004) it was found that in general children attributed their musical success to their teachers more than their families.It was found that the students attributed their success to primarily effort, chance and work force and their failure to the lack of ability and to their parents or teachers.
Özmenteş (2012) also got the similar results in his study.Students attribute their success in the courses to studying exams, their affection to music and music courses, teacher features and their attitudes towards music courses, their skill and effort.They attribute their failure to the lack of ability, the negative behavior of their classmates, not having affection to the songs in the lessons and not studying for music courses.
It was found that McClure et al. (2011) attributed good marks (the success) of the students to their chance, family and friends in their study.While the students see effort as a strong indicator of their success, they attribute their low marks to the teacher.On the other hand, in their study, Turk and Bry (2005) investigated the attribution of the adolescent children and their parents and it was found that the attribution style of the parents had passed to their children.Thus, to make the investigation about if there is an effect of parents and teachers' attribution on the success and failure of the children is recommended.

Factor 1 :
Attribution to the teacher and the course Factor 2: Attribution to the family Factor 3: Attribution to himself.The results of the factor analysis yielded following three factors for the Failure Attribution Scale Factor 1: Attribution to the family

Factor 2 :
Attribution to the teacher and the course Factor 3: Attribution to himself.

Figure 3 .
Figure 3. Path diagram related to second level confirmatory factor analysis of success attribution

Figure 4 .
Figure 4. Path diagram related to second level confirmatory factor analysis of failure attribution

Table 1 .
Different dimensions to explain success and failure in an examination (a student's causal attributions from the words of this student)

Table 2 .
The distribution of students into sections and classes

Table 3 .
Eigenvalues line chart on success attributions scale.Factor Factor analysis related to success attribution scale, item total scale correlations and t-test results * p<.05

Table 4 .
Factor analysis related to Failure Attribution Scale, Item Total Scale Correlations and t Test Results * p<.05

Table 6 .
Second level factor analysis results about success attributions

Table 7 .
Second level factor analysis results about failure attributions