Attribution Feedback in the Elementary Classroom.

Abstract. This study investigates the types of feedback statements utilized by 3rd-grade teachers during mathematics instruction. It considers how these statements might affect numerous education variables from the perspective of attribution theory (Weiner et al., 1971). The goal of the study was to create and test a coding scheme by which attribution feedback can be reliably identified. Twenty teachers were observed, and feedback statements were analyzed. The results support a conceptual relationship between feedback and attribution theory, and indicate that teachers rarely provide students with attributionally informative feedback. Implications for motivation, achievement, and teacher training are discussed.

Teacher beliefs and expectations regarding individual students are thought to hold great influence over student motivation and achievement. A number of models have been proposed that attempt to explain how teachers can influence student motivation. Brophy and Good (1970), for example, suggest that teachers behave differently toward various students, based on preliminary expectations. This differential treatment informs students about how they are expected to perform. If the treatment persists, Brophy and Good suggest that it will likely affect self-concept, achievement motivation, levels of aspiration, classroom conduct, and interactions with the teacher. This, in turn, reinforces the teacher's initial expectations, which ultimately affect student achievement.

One way in which teachers might have a particular impact on motivation is in the types of feedback statements they make to students during instruction. These statements can directly affect students' self-perceptions. A number of studies have examined teacher feedback in the classroom (i.e., Irvine 1986; Meyer & Thompson, 1956; Simpson & Erickson, 1983). These studies, however, lack a theoretical framework that would explain the relationship between feedback and student motivation or achievement. This study examines the types of feedback statements teachers use during mathematics instruction, and considers the implications for student motivation and achievement from the perspective of attribution theory. Assertions are currently being made by education reformists to include attribution-type feedback in the classroom (Richardson, 1995). These suggestions need to be tested before widespread utilization is warranted. The goal of this study was to create and test a coding scheme by which attribution feedback can b e reliably identified.

Attribution Theory

Weiner and his colleagues (1971) proposed that students' perceptions of the cause for success or failure in the classroom were much more important influences on future task performance than actual experiences of success or failure. According to attribution theory, the cause that an individual decides upon can take one of four types: ability, effort, task difficulty, or luck (Weiner, 1980, 1984, 1990). In addition, these elements can be categorized by their levels of stability, internality, and control (Frieze & Snyder, 1980). If a child perceives an outcome to be due to ability level or difficulty of the task, then the perception can be categorized as stable. Conversely, outcomes attributed to effort and luck are categorized as unstable. Causes that originate from within the individual, such as ability and effort, are categorized separately from external causes, such as task difficulty or mood of the teacher. The control category differentiates between causes such as effort, which is within the power of the i ndividual, and ability, which is not controllable (see Table 1).

If a child perceives ability (an internal, stable, and uncontrollable factor) to be the cause of a particular success, then he or she is more likely to attempt and persist at future similar tasks. If a failure is attributed to lack of ability, then the child will likely avoid subsequent situations, in order to preserve self-worth. A child who perceives effort (an unstable, internal, and controllable factor) as the cause of success may attempt to perform the task in the future, expecting to succeed with sustained effort. However, this child is less likely to attempt the task than the child whose success is attributed to ability and who, in turn, perceives that less effort is needed to achieve success. If failure is attributed to lack of effort, the child does not feel incapable, but anticipates future success with increased effort and is probably less avoidant of the situation than the child whose failure is attributed to lack of ability (see Weiner, 1979).

Research suggests that at least in the laboratory setting, statements that infer a student's level of ability or effort directly affect motivation and achievement (Schunk, 1983, 1984, 1989). In this series of studies, students were provided positive feedback that either suggested their level of ability ("You're good at this"), level of effort ("You've been working hard"), a combination of ability and effort ("You're good at this and you've been working hard"), or no attribution feedback ("okay"). In general, students who received positive ability and effort feedback had higher levels of self-efficacy and mathematical skill than did students who received no attribution feedback. In addition, students whose feedback emphasized ability had higher levels of self-efficacy and mathematical skill than students whose feedback emphasized effort, a finding consistent with attribution theory.

While training studies such as these are promising, it is unknown if similar results will occur outside the laboratory. With so many other variables at work in the classroom, the effects of attribution feedback from the teacher may be minimal. These studies also are limited by the narrow range of positive feedback statements used. A broader range of statements are used by teachers in the natural setting, and these statements include both positive and negative forms of feedback.

Studies that examine teacher feedback statements in the classroom typically assess the degree of teacher approval or disapproval, and they do not utilize a theoretical framework such as attribution theory classroom (e.g., Irvine, 1986; Meyer & Thompson, 1956; Simpson & Erickson, 1983). Although informative at some level, these studies do not reveal the underlying relationships between feedback and motivation or achievement. Before attribution feedback can be evaluated for its relationship to motivation and achievement, it is necessary to establish a valid method of gathering evidence of its use.

In this study, a system was developed to observe teacher feedback in the classroom setting with respect to its influence on student attributions for academic success or failure. The technique examined teacher feedback as either ability-oriented, effort-oriented, conduct-oriented, or general (see Appendix A for examples of each type of feedback).

The ability and effort feedback statements were examined because they are assumed to influence directly the students' ability and effort attributions. Conduct feedback was selected because it is a particularly common form of feedback in the classroom (especially negative conduct). Also, it is thought that conduct feedback is attributionally informative. Negative conduct implies that the student is not providing the appropriate effort, whereas positive conduct is related to positive ability and effort. Students who are perceived as having positive ability are more likely to be viewed as performing the supplemental actions required to demonstrate their ability, while students who are concerned with providing positive effort are likely to be viewed as presenting positive conduct.

The general feedback statements were included not so much because they represent attribution theory, but rather because they lack the informative nature of attributional feedback. These statements are frequently used in the classroom, and present feedback to students concerning the degree of correctness. They are of concern, however, because they do not provide the student with specific information regarding the cause of the outcome. Thus, such statements leave student attributions open to inference and to the possibility that a less motivating attribution will be made.

Attribution theory would suggest that external feedback factors, such as task difficulty and luck, also should be included as categories for observation. It was decided to exclude these external categories, based on the relative infrequency of luck feedback (teachers should not reasonably inform a student that they succeeded or failed on an academic task based on luck), and because task difficulty feedback mapped onto ability. If a teacher informed a student that a particular task was too difficult, then it was inferred that the student lacked ability. In contrast, if the student was informed that the task was very easy, it was inferred that the student possessed positive ability. Task difficulty feedback made by a teacher, therefore, was coded as if the teacher had made an ability comment. This is not a vast leap, because the feedback comments were made to individual students rather than to the class as a whole. Task difficulty determinations are thought to be made as part of a social comparison. Students fe el that a task is easy if everyone performs well. On the other hand, if the group performs poorly, the task is perceived to be difficult. Task difficulty feedback to an individual reflects the ability of that individual.

In this study, a feedback coding scheme was analyzed for inter-rater reliability and consistency. These tests were required to assess the validity of the technique. If coders were unable to agree on the types of feedback presented within the class, or if teachers were inconsistent in their feedback, then it would be difficult to make later associations between feedback statements and student motivation or achievement. Second, means and standard deviations for each type of feedback were examined to determine the frequency with which teachers used the various types of feedback. A measure of level of use was necessary to make inferences regarding the type of feedback and motivation. Third, correlational and factor analyses were expected to reveal a relationship between the conceptual model of attribution theory and the empirical nature of the feedback statements. That is, certain types of feedback were expected to be positively related to other types of feedback with the same sort of motivational characteristics proposed by the theory (i.e., positive ability and positive effort feedback).

Method

Subjects

Twenty 3rd-grade math teachers in classrooms of heterogeneous ability were recruited for this study from 10 different elementary schools (8 public, 1 private, 1 parochial) in central and western New York State. Two schools were rural, 2 were urban, and 6 were suburban. The sample included teachers with an average of 14.9 years of teaching experience (ranging from 1-37 years) and an average of 7.8 years of experience at the 3rd-grade level (ranging from 1-25 years). All of the teachers had earned at least a bachelor's degree in education, and 16 had earned a master's degree. The average number of students in each class was 19.3 (ranging from 11-24 students). All classrooms were co-ed. The cultural background of the students was diverse, although students in two classrooms were entirely of Caucasian background, while one classroom was entirely composed of students of African American descent. Students in the remaining classrooms were of various backgrounds.

Procedure

School district and teacher permission for participation were solicited through letters and conversations with administrators. Teachers who agreed to participate were informed that their feedback statements to students would be observed. Over a 10-week period, each teacher was observed and videotaped three times during formal mathematics instruction by one of three individuals-the author or two graduate students trained by the author. To standardize lesson observation time, each lesson was only taped for 20 minutes, and one hour of videotape for each teacher was used in the analyses. For confidentiality reasons, students were not visible on the tapes. Because students could not be observed, there are concerns that perhaps only a few select students received the majority of feedback. Teachers may have been different in the types of feedback they provided to these select students, which would have influenced the generalizability of the results.

For this reason, the camera person noted when a student was addressed who had not previously been addressed by the teacher during the lesson. This was done by quietly saying "new" into the camera microphone following the teacher's statement to a new student. Using this method, the observational technique could be evaluated in two ways: one in which all feedback was examined regardless of the student to whom it was directed (i.e., a few select students could influence the data), and another in which only the first feedback statement directed to a student during a single lesson was examined (a contrived, but perhaps more informative, observation, since feedback was examined based on a more even distribution to students). After each taping, the camera person reviewed the observation tape for feedback coding purposes.

In pilot testings of feedback observation and coding, it was determined that, at times, within the same lesson, teacher feedback was very frequent and other times more sporadic. Coding every instance of feedback became very difficult and inter-rater reliability was very low. If one coder disagreed early on in the observation, all other inter-rater codings from the same tape might be influenced. The following example serves to illustrate this problem: A teacher assigns a math problem that each student is required to answer. The students perform the problem and the teacher circulates throughout the room to check the students' accuracy. As she does this, she provides at least one feedback statement to each student in the class in a period of less than one minute. If two coders were observing the teacher and categorizing the feedback, and one coder failed to identify one instance of feedback, each of the coder's remaining categorization codings would be off by the one missed statement. The correlation with the o ther coder then could be greatly affected.

In order to minimize confusion and to standardize coding, it was decided to code the statements according to a timed behavior checklist. This required that the coder identify only the first feedback statement in each 20-second interval throughout the observation. The coder used the checklist to identify the gender of the student to whom feedback statements were made, and to note if the statement was made to a new student or to one who previously had been addressed. After the camera person had performed the initial coding, a copy of the tape was passed to another coder (the author or one of the two graduate students). This coder viewed a random 5-minute portion of the tape to assess the inter-rater reliability.

Eight undergraduate psychology students were trained by the author in classroom observation methods to assess and categorize the types of feedback directed toward students. Six training videos of sample classes were created, and trainees spent 10 hours viewing these videos, recording types of feedback, and discussing questions that arose about categorizing feedback. A final video was prepared in which feedback previously had been scripted to include designated levels of each type of feedback. The observers were tested on inter-rater reliability using this video, and the four most reliable observers were retained for actual classroom observations.

These four observers were paired randomly for videotaped classroom observations over a week-long period. At the beginning of each week, the observers were reassigned to new pairings. Each observer received a videotape of lessons and a copy of a tally sheet from the feedback codings, which denoted the times where feedback occurred, as well as instances of students not previously addressed. This procedure was necessary because the students were not identifiable from the tapes; as mentioned previously, the tapes only focused on the teacher. One observer within the pair tallied the types of feedback directed toward the students from the teacher, as identified by the feedback sheet for the entire 20-minute lesson. The second observer, at a later time, viewed a random 5-minute portion of that video, tallying only the feedback presented within that interval. Observers noted frequencies of positive and negative ability, effort, conduct, and general feedback. If a feedback statement was considered to characterize more than one category it was coded as both (for example, "You are very smart (ability) and you tried hard (effort)"). Analysis of inter-rater reliability between the two observers for the shared 5-minute interval was made.

Coding Procedures

Fifty-eight classroom tapes were retained, to be coded for the identification and categorization of feedback. The observations for each of the 20 classes were combined to obtain proportions of each type of feedback by the number of students in the class. Two tapes were inadvertently damaged prior to completion of coding. Outcome measures for the two teachers for whom only two observations were available were made based on the proportions for the two tapes, rather than all three.

Inter-rater agreement for the identification of feedback statements was assessed for each observation in which at least one feedback statement was observed by at least one coder during the random 5-minute interval in which the two coders viewed the tape. The kappa-statistic measure of agreement was used because of the limited number of categorical ratings that could be coded during the observations (Landis & Koch, 1977). The average inter-rater reliability for the identification of feedback statements was [Z.sub.fisher]=2.15, or an average of k=.97. This average reliability was derived by transforming the kappas for the 5-minute intervals observed by two coders for each tape into the corresponding [Z.sub.fisher]' deriving the average [Z.sub.fisher] for all of the tapes, and transforming this average back to the corresponding kappa correlation.

Inter-rater agreement for the categorization of feedback statements was derived for each category on each tape in which at least one coder rated at least one feedback statement of that type during that 5-minute interval. The kappa-statistic for agreement was again used, because there were only two categorical ratings that could be coded during the observation (presence or absence). Inter-rater agreement for positive ability was obtained from 13 of the observations, with an average [Z.sub.fisher]=2.06, corresponding to an average k=.97. Fifteen observations contained at least one instance of positive effort coding with an average [Z.sub.fisher]=1.99, corresponding to an average k=.96. There were 13 instances of positive conduct observations, with an average [Z.sub.fisher]=2.76, corresponding to an average k=.99. Positive general feedback was quite prevalent and appeared on 105 of the coding sheets with an average [Z.sub.fisher]=2.70, corresponding to an average k=.99. Negative ability feedback was coded on onl y five of the observations with an average [Z.sub.fisher]=2.39, corresponding to an average k=.98. Negative effort feedback appeared in 13 observations with an average [Z.sub.fisher]= 1.44, corresponding to an average k=.89. Negative conduct appeared in 39 observations with an average [Z.sub.fisher]=2.54, corresponding to an average k=.99. Finally, negative general feedback appeared in 49 observations [Z.sub.fisher]=1.95, corresponding to an average k=.96.

Consistency of the teacher use of feedback categories was estimated by computing correlations on the first and third observation for each teacher. The average correlation was .98 with a range of .81-1.00 (average Z=2.29). The means and standard deviations of the proportions for each type of feedback for each teacher were derived to assess the variability between teachers in the patterns of feedback used.

Results

The first analysis of teacher feedback statements was performed, ignoring the presence of the "new" coding, which addressed whether or not a student previously had been called upon during the class session. Basic descriptive statistics were performed on the proportions of each type of feedback for the 20 classrooms (see Table 2).

The second analysis of teacher feedback was performed using only feedback that was coded as "new." This was done so that students who were repeatedly addressed (i.e., the particularly vocal or disruptive students) would not skew the interpretations regarding the use of feedback in the classroom. Basic descriptive statistics were performed on the proportions of each type of feedback for the 20 classrooms (see Table 3).

The next set of analyses examines the degree to which the hypothesized conceptual relationships among the feedback categories are empirically supported. In the first analysis, an 8 X 8 correlation matrix involving all of the feedback categories was formulated; this is summarized in Table 4. Significant correlations between positive ability and positive effort, positive conduct, and negative effort were found, but not between positive ability and positive general feedback. Positive effort was significantly correlated with positive ability and negative effort. Negative ability was significantly related to positive general, negative effort, and negative conduct feedback. Negative general feedback was not related to any of the other categories.

A factor analysis of the raw data on feedback by category, using the centroid method with a varimax normalized rotation, revealed two oblique hierarchical factors. The oblique option was used because the categories are not considered to be orthogonal (see Cooper & Baron, 1979). Factor 1 included significant loadings among positive general, negative ability, negative conduct, and negative general feedback. Factor 2 included significant loadings among positive ability, positive effort, positive conduct, and negative effort (see Table 5). There was a significant relationship between the two factors, suggesting that the feedback may be related to a second-order factor.

Discussion

The coding scheme for identifying and classifying feedback statements appears to be reliable. Inter-rater reliability for both the feedback identification and categorization processes was excellent. The lowest kappa-statistic (k=.89 for negative effort) between raters was well within the .81-1.00 range designated by Landis and Koch (1977) as "almost perfect." Individual teacher feedback also appeared to be stable across days, indicating that teachers are consistent in their feedback use. That is, the teachers who had a high or low rate of use of certain forms of feedback on one day presented a similar pattern on a later date. Each category of feedback demonstrated variability among teachers, which suggests that there are individual differences in the levels of the various forms of feedback used by teachers. The finding that individual teachers are consistent in their patterns of feedback use across several days and that different teachers present different patterns supports the view that these forms of feedba ck might contribute to differences in student motivation or achievement.

The ecological nature of the feedback statements cannot be determined from this study. The degree to which students influence teacher feedback or feedback emits as a teacher style or personality variable is unknown. To argue, however, that individual student differences on motivation and achievement constructs are related to teacher feedback statements, it is necessary that different teachers vary in their use of feedback, but be somewhat consistent in their own patterns of use.

Internal validity is evident in the correlations among feedback categories across teachers. As expected, significant correlations were found for categories that hold similar characteristics. Positive ability and positive effort are related because they are both internal forms of academically informative feedback that indicate a correct response. Positive conduct was significantly related to positive ability and weakly related to positive effort. This result may have occurred for the following reasons: 1) all three forms of feedback suggest to students that they performed the correct action, 2) all are informative of the reason why the praise occurred (e.g., "I like the way you're writing your numbers neatly"), and 3) all inform that the behavior was related to a personal action (a level of internality). Negative effort was significantly related to positive ability and positive effort, perhaps because it informs the student that success is possible despite the failure and because it informs the student that p ersonal action (internal controllable factors) can lead to future success. Positive general feedback was not significantly related to the other types of feedback in this group. This finding might be expected, because positive general feedback shares only the single characteristic that the student was correct with positive ability, effort, and conduct feedback. Positive general feedback differs from the other positive forms of feedback, however, because it fails to inform the student of the reason for the praise. Correlations between positive general feedback and the other positive types were, at best, weak.

Correlations also appeared in an expected manner on the theoretically less motivating forms of feedback. Negative ability feedback was significantly related to negative conduct feedback and weakly related to negative general feedback. These categories were expected to correlate, based on the fact that they each lacked at least two characteristics leading to motivation. Negative ability feedback informs students that they were incorrect, and the cause of their failure was internal, stable, and uncontrollable. Negative conduct feedback was academically irrelevant, and informed the student that actions were not held in high regard by the teacher. Negative general feedback suggested to the student that the behavior was incorrect, yet failed to inform the student of the reasons for the failure. Interestingly, positive general feedback was significantly related to negative ability feedback. This finding suggests that the uninformative nature of positive general feedback somehow shares characteristics with negative ability feedback, perhaps the uncontrollable factor apparent in both variables. Negative ability and negative effort feedback were significantly correlated. While not predicted, this finding is not surprising, since both types of feedback inform students that they are incorrect and both provide academically relevant information as to why they were incorrect. Negative conduct feedback also was related to negative effort, again not surprisingly, since negative conduct infers that the appropriate effort is not being provided and because both suggest that student action is incorrect yet controllable. Negative general feedback was not significantly related to any of the other categories of feedback. This finding suggests that negative general feedback has little effect on motivation.

The factor analysis further supports the predicted relationships among feedback categories. Two oblique factors were found; one in which attribution theory might claim variables that are motivationally detracting or neutral loaded, and the other in which more theoretically motivational variables were loaded. This finding implies that such feedback categories as positive ability, positive effort, and positive conduct, as well as negative effort, are interrelated and different from other categories such as negative ability, negative conduct, and negative general feedback, as well as positive general feedback. The modest, but significant, correlation (r=.44) among the factors suggests that the feedback variables examined in this study may be related to a second order factor. This second order factor can be explained by a number of constructs. Variables in one factor tend to contribute to higher self-efficacy, perceived control, outcome expectancy, and self-concept, while decreasing feelings of learned helplessn ess. Meanwhile, variables in the other factor tend to lead to lower self-efficacy, perceived control, outcome expectancy, and self-concept, and they increase learned helplessness.

Overall, the internal validation of correlations among feedback categories that have similar characteristics and the factor analysis present evidence for the expected empirical relationship among feedback types. These analyses also suggest that certain forms of feedback are related on some higher level. To determine the nature of this higher level construct, an external examination/validation of the relationship between feedback and student behavior is necessary.

Based on the factor analysis and attribution theory, a model of the level of motivation for each type of feedback can be made for both positive and negative feedback (see Figure 1). Although the factor loading for positive ability feedback is lower than that of positive effort feedback, positive ability feedback following success, according to the theory, should be considered the most motivating, because it indicates that the student should attribute an internal stable cause for success. Past research on the relationship between teacher expectancy and student behavior suggests that the teacher is more likely to provide effort-oriented feedback, based on the teacher's perception of a student's level of effort (e.g., Silverstein, 1979). However, this research is not experimental and, therefore, does not manipulate the level of ability feedback that a student receives. Laboratory studies of the relationship between feedback and motivation (e.g., Schunk, 1983, 1984, 1989) indicate that positive ability feedback is somewhat more motivating than positive effort feedback. Therefore, there is no reason to suggest that attribution theory is incorrect in assuming that positive ability feedback is more motivational than positive effort feedback. Theoretically, the next level of feedback would be positive effort feedback, because this form of feedback loads the highest on the motivating factor and it is academically informative of the internal nature of success. Positive conduct feedback would appear next on the motivational hierarchy because it also loads on the motivating factor, and because it is informative, albeit non-academically, of the reasons for student success. The last form of positive feedback and the least motivating would then be positive general feedback. Positive general feedback did not load with the other positive categories (see Table 5). It is uninformative, both academically and non-academically, and allows the chance for students to make less motivating external attributions (see Table 1).

Of the negative feedback categories, negative effort feedback is the most motivating, because it did not load with the other less motivating categories, and because it indicates to the student that future success is possible. Negative conduct feedback is likely to be the next most motivating of these categories because it loads least on the less motivating factor, and because it is mappable onto negative effort feedback. As stated earlier, negative conduct infers negative effort. Negative general feedback would follow negative conduct, because it informs the student of the failure but does not directly state that the student is incapable of future success. Also, negative general feedback loads on the less motivating factor slightly higher than negative conduct. Negative ability, then, is the least motivating of the feedback types, because it indicates that the student is not capable of actions that would lead to future success. Negative ability feedback also carries the highest load on the less motivating fa ctor.

Although this depiction of the levels of motivation for each type of feedback suggests the types of feedback that would be most desirable in the classroom, further testing is necessary to determine if this hierarchy is empirically accurate. The actual observations of classroom feedback demonstrated that the so-called more motivating forms of feedback rarely were used.

The proportion of positive attribution feedback directed to students was disappointingly low, indicating that teachers rarely provided students with feedback attributing success to the internal factors of ability and effort. Positive conduct feedback was also very rare, indicating that teachers were not likely to use verbal reinforcement as a classroom management technique. Positive general feedback was the most frequently used category of feedback. This is unfortunate, because positive general feedback is less informative to students than ability or effort feedback, and as stated previously, it is less motivating because it is highly inferential.

Negative ability feedback was the rarest type of feedback observed. According to attribution theory and the hierarchy presented, negative ability feedback is the least motivating type of feedback. However, negative effort feedback, which is generally considered to be the most motivating feedback following failure, was also rare. Negative conduct feedback was the third most prevalent type of feedback observed, following only positive and negative general feedback. This finding suggests that teachers are expending a great deal of energy reprimanding students for poor conduct. The hierarchy of negative motivation feedback, however, suggests that negative conduct and negative effort feedback are more motivating than negative general feedback; thus, it is unfortunate that negative general feedback occurs more often than do more motivating negative feedback categories. Negative general feedback, although the second most prevalent type of feedback, was much rarer than positive general feedback. Teachers were theref ore more likely to provide feedback indicating success than failure. However, just as positive general feedback is uninformative of the reasons for success, negative general feedback fails to indicate why the student did not succeed. Although negative general feedback is more motivating than negative ability feedback, it is much less motivating than negative effort feedback, because of its inferential nature.

This study attempted to create and test an observational coding scheme for examining teacher feedback in the classroom. Before any widespread evaluation of the effects of attribution feedback on motivation or achievement can be performed, a reliable form of observation is necessary. This study presented a coding scheme that can be a reliable means of assessing feedback. The study is, however, limited by its design. Teachers, rather than students, were the unit of analysis. To obtain an adequate sample size, 20 classrooms were recruited. This restricted the ability of the researchers to identify and track individual children. Had this been feasible, it would have been necessary to obtain permission slips from the parent of each child in each classroom to relate student behaviors to the teacher feedback statements. Because the observations were performed in the classroom as a part of regular instruction, any student without parental permission would have needed to miss three days of mathematics instruction, wh ich would have been, of course, inappropriate. Therefore, the coding scheme has only been subjected to internal validation.

Schunk (1983, 1984, 1989) provides evidence that attribution feedback affects student motivation and achievement in the laboratory. However, the study presented here demonstrates that attribution feedback is rarely used in the classroom. Before major assertions can be made that suggest the need to train teachers to use certain forms of attribution feedback, it is necessary to determine, in a well-controlled study, if teachers trained in these feedback methods actually implement them in the classroom; and if utilized, whether this feedback leads to increases in motivational variables such as self-efficacy; and finally, if motivation indeed increases, whether it leads to greater academic achievement.

Future research in this area also should include studies of teacher responses to individual students. It is probable that feedback to students differs according to race, gender, ability, or socioecomomic status. However it is unlikely that students varying on these dimensions differ in the way that feedback motivates them. If teachers vary in their types of feedback to individual students, and this feedback is found to influence motivation and achievement in the classroom, they must be made aware of the potential detrimental effects.

Acknowledgements. This research was performed as part of a doctoral program requirement at the Syracuse University Department of Psychology. The assistance of Dr. William J. Meyer, Dr. D. Bruce Carter, and Dr. Vernon C. Hall is greatly appreciated. A portion of this research was presented at the 1996 annual meeting of the Northeastern Educational Research Association as research in progress.

References

Brophy, J., & Good, T. (1970). Teachers' communication of differential expectations for children's classroom performance: Some behavioral data. Journal of Educational Psychology. 61, 365-374.

Cooper, H., & Baron, R. (1979). Academic expectations, attributed responsibility, and teachers' reinforcement behavior: A suggested integration of conflicting literatures. Journal of Educational Psychology, 71, 274-277.

Frieze, I., & Snyder, H. (1980). Children's beliefs about the causes of success and failure in school settings. Journal of Educational Psychology, 72, 186-196.

Irvine, J. (1986). Teacher-student interactions: Effects of student race, sex, and grade level. Journal of Educational Psychology, 78, 14-21.

Landis, J.R., & Koch, G.G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159-174.

Meyer, W., & Thompson, G. (1956). Sex differences in the distribution of teacher approval and disapproval among sixth-grade children. Journal of Educational Psychology, 47, 385-396.

Richardson, L. (1995). Dr. Crew's prescription: 'Efficacy' looms in New York's next education philosophy. The New York Times, November 26 (pg. 39, 42).

Schunk, D. (1983). Ability versus effort attribution feedback: Differential effects on self-efficacy and achievement. Journal of Educational Psychology, 75, 848-856.

Schunk, D. (1984). Sequential attribution feedback and children's achievement behavior. Journal of Educational Psychology, 78, 1159-1169.

Schunk, D. (1989). Self-efficacy and cognitive skill learning. In C. Ames & R. Ames (Eds.), Research on motivation in education (Vol. 3, pp. 13-44). New York: Academic Press.

Silverstein, M. (1979). An attributional analysis of teachers' evaluative judgments. Dissertation Abstracts International, 39, 1055-1056c.

Simpson, A., & Erickson, M. (1983). Teachers' verbal and nonverbal communication patterns as a function of teacher race, student gender, and student race. American Educational Research Journal, 20, 183-198.

Weiner, B. (1979). A theory of motivation for some classroom experiences. Journal of Educational Psychology, 71, 3-25.

Weiner, B. (1980). Human motivation. New York: Holt, Rinehart and Winsten.

Weiner, B. (1984). Principles for a theory of student motivation and their application within an attribution framework. In C. Ames & R. Ames (Eds.), Research on motivation in education (Vol. 1, pp. 1536). Orlando, FL: Academic Press.

Weiner, B. (1990). History of motivational research in education. Journal of Educational Psychology, 82, 616-622.

Weiner, B., Frieze, L, Kulda, A., Reed, L., Rest, S., & Rosenbaum, R. (1971). Perceiving the causes of success and failure. New York: General Learning Press.

Table 1
Causes of Successs and Failure, Classified According
to Locus, Stability, and Controllability
(from Weiner, 1979)
                 Internal             External
Controllability  Stable    Unstable   Stable           Unstable
Uncontrollable   ability   mood       task difficulty  luck
Controllable     typical   immediate  teacher bias     unusual help
                 effort    effort                      from others
Table 2
Descriptive Statistics of Overall Classroom Feedback
(results are presented as a proportion of feedback by the number of
students in the class)
Type of Feedback    Mean  Minimum  Maximum  Std. Dev.
 Positive Ability   .023   .000       .083    .024
 Positive Effort    .041   .000       .208    .053
 Positive Conduct   .029   .000       .063    .023
 Positive General  1.167   .593      2.225    .448
 Negative Ability   .016   .000       .091    .024
 Negative Effort    .031   .000       .182    .044
 Negative Conduct   .173   .017       .485    .129
 Negative General   .193   .042       .404    .095
Table 3
Descriptive Statistics of "New" Classroom Fedback
(results are presented as a proportion of feedback by the number of
students in the class)
Type of Feedback   Mean  Minimum  Maximum  Std. Dev.
 Positive Ability  .005     .000     .018       .001
 Positive Effort   .015     .000     .083       .021
 Positive Conduct  .008     .000     .028       .010
 Positive General  .451     .183     .667       .127
 Negative Ability  .004     .000     .025       .008
 Negative Effort   .008     .000     .042       .013
 Negative Conduct  .070     .000     .214       .060
 Negative General  .076     .014     .173       .047
Table 4
Correlations Among Feedback Categories
Feedback  Positive  Positive  Positive  Positive  Negative
  Type     Ability    Effort   Conduct   General   Ability
PA            1.00     .64 *     .51 *       .17       .23
PE                    1.00       .39         .18       .14
PC                              1.00         .33       .27
PG                                          1.00       .50 *
NA                                                    1.00
NE
NC
NG
Feedback  Negative    Negative  Negative
  Type      Effort     Conduct   General
PA             .53 *     .36        -.07
PE             .54 *     .19        -.25
PC             .37       .36        -.05
PG             .29       .17         .36
NA             .68 *     .68 *       .36
NE            1.00       .65 *       .14
NC                      1.00         .23
NG                                  1.00
(*)Marked correlations are significant atp[less than].05.
Table 5
Feedback Factor Loadings (Varimax normalized)
Variable           Factor 1  Factor 2
 Positive Ability    .089      .811 *
 Positive Effort    -.078      .853 *
 Positive Conduct    .219      .569 *
 Positive General    .666 *    .164
 Negative Ability    .841 *    .237
 Negative Effort     .517      .648 *
 Negative Conduct    .598 *    .386
 Negative General    .608 *   -.236
(*)Correlation Between the Oblique Factorsis .44
Figure 1
Motivational Hierarchy of Feedback
                  Positive  Negative
Most Motivating   Ability   Effort
                  Effort    Conduct
                  Conduct   General
Least Motivating  General   Ability
Appendix A
Examples of Feedback Statements, By Category
Positive Ability                    Negative Ability
You're a natural at this.           Math is not your best subject.
You're very bright.                 Your talents lie outside of math.
You show true talent.               I don't know if you can do this
                                    part.
You are very good in math.          I'll have to help you when it comes
                                    to that part.
Positive Effort                     Negative Effort
You put a lot of time into this.    You're not trying your hardest.
You have worked hard.               I know you can do better than that.
Good effort.                        Maybe you should have studied.
You are a hard worker.              You need to start doing your
                                    homework.
Positive Conduct                    Negative Conduct
Thank you for sitting so quietly.   Be quiet.
Look how neatly (name)'s paper is.  You're late.
(name) has already gotten her       Sit down.
pencil out.
Thank you for walking quietly.      Sit still.
Positive General                    Negative General
That's very good.                   No.
Yes.                                That's not correct.
Correct.                            That's not what I'm looking for.
Well done.                          You didn't do very well.