Professional Documents
Culture Documents
Gibson, S., & Dembo, M. H. (1984) - Teacher Efficacy A Construct Validation. Journal of Educational Psychology, 76 (4), 569-582.
Gibson, S., & Dembo, M. H. (1984) - Teacher Efficacy A Construct Validation. Journal of Educational Psychology, 76 (4), 569-582.
There is evidence that teachers' beliefs achieving schools spent longer proportions
in their abilities to instruct students may of time in instruction and demonstrated
account for individual differences in effec- greater concern and commitment to their
tiveness (Armor et al., 1976; Herman & students' achievement. In addition, al-
McLaughlin, 1977; Brookover et al., 1978; though the authors did not specifically refer
Brophy & Bvertson, 1977). Berman and to teacher efficacy, Brophy and Evertson
McLaughlin (1977), in their evaluation of (1977) reported that teachers who were
100 Title III projects of the 1965 Elementary successful in producing student learning
and Secondary Education Act, found that gains in the Texas Teacher Effectiveness
the most important characteristic deter- Study tended to have higher expectations
mining the effectiveness of change-agent and assumed personal responsibility for
projects was teachers' sense of efficacy—a making sure that students learned. If these
belief that teachers can help even the most teachers encountered difficulties, they
difficult or unmotivated students. Armor viewed them as obstacles to be overcome by
et al. (1976) reached a similar conclusion in discovering appropriate teaching methods,
evaluating the effectiveness of the School not as indicators that the students could not
Preferred Reading Program in Los Angeles. learn.
These researchers reported that the greater Although the importance of teachers'
the teachers' efficacy, the more their stu- sense of efficacy has been identified, re-
dents advanced in reading achievement. searchers are not certain how to conceptu-
The measure of teachers' sense of efficacy in alize and adequately measure the construct.
both of these studies was derived from two Denham and Michael (1981) and Ashton and
questions based on Rotter's (1966) locus of Webb (1982) developed multidimensional
control construct. models of teacher efficacy that were in-
Brookover et al. (1978), in their investi- fluenced by Bandura's (1977, 1978) con-
gation of school climate variables influencing ceptualization of self-efficacy. Bandura
achievement, found that teachers in high- (1977) argued that although locus of control
is primarily concerned with causal beliefs
Requests for reprints should be sent to Sherri Gibson, about action-outcome contingencies or a
Director, Auxiliary Education, Clovis, California person's estimate that a given behavior will
93612. lead to certain outcomes, personal efficacy
569
570 SHERRI GIBSON AND MYRON H. DEMBO
is concerned with the conviction that one can more likely to have a stronger academic focus
successfully execute the behavior required in their classrooms. There is evidence from
to produce the outcomes. Outcome and ef- teacher effectiveness research that the
ficacy expectations are differentiated be- amount of time spent directly on instruction
cause individuals can believe that certain is related to gains in student achievement,
behaviors will produce certain outcomes, but whereas time spent in noninstructional ac-
if they do not believe that they can perform tivities (e.g., games and art) is related to
the necessary activities, they will not initiate negative gains (Stallings & Kaskowitz, 1974).
the relevant behaviors, or if they do, they will In addition, efficacious teachers may tend to
not persist. use elements of direct instruction that in-
If we apply Bandura's theory to the con- clude a pattern of behavior used by effective
struct of teacher efficacy, outcome expec- elementary teachers: structured academic
tancy would essentially reflect the degree to activities supervised by the teacher, exten-
which teachers believed the environment sive content coverage, monitoring of student
could be controlled, that is, the extent to performance, specific questioning of stu-
which students can be taught given such dents with immediate feedback, and use of
factors as family background, IQ, and school large group instruction. There may also be
conditions. Self-efficacy beliefs would in- differences in teacher feedback patterns
dicate teachers' evaluation of their abilities between high- and low-efficacy teachers.
to bring about positive student change. For example, Good (personal communica-
Bandura's theoretical predictions of initia- tion, January 22, 1982) maintained that a
tion and persistence of coping behavior teacher's willingness to stay with a student
suggest that persons high on both variables in a failure situation is indicative of a
will respond with active, assured respon- teacher's confidence in his or her teaching
siveness and persons low on both variables ability and/or the student's ability to
will give up readily if they do not get results. learn.
One would predict that teachers who believe In attempting to validate the construct of
student learning can be influenced by ef- teacher efficacy, the construct should also be
fective teaching, and who also have confi- distinguished from other individual attrib-
dence in their own teaching abilities, should utes identified in more effective teachers.
persist longer, provide a greater academic Because verbal ability (Berman &
focus in the classroom, and exhibit different McLaughlin, 1977; Bowles & Levin, 1968;
types of feedback than teachers who have Coleman et al., 1966; Ekstrom, 1975b; Han-
lower expectations concerning their ability usek, 1970) and flexibility (Ekstrom, 1975b)
to influence student learning. are related to teacher behavior and student
Teacher interview and correlational data outcomes, it is important to investigate the
provide support for at least two different relationship between these variables and
efficacy dimensions—teaching efficacy and teacher efficacy.
personal teaching efficacy (Ashton & Webb, The present investigation was separated
1982; Webb, 1982). However, attempts to into three distinct phases investigating the
develop more useful measures of teacher following questions: Phase 1 (factor anal-
efficacy than the two items used in the ysis): What are the dimensions of teacher
studies by Berman and McLaughlin (1977), efficacy? How do these dimensions relate
and Armor et al. (1976) have been unsuc- to Bandura's theory of self-efficacy? What
cessful (Ashton, Olejnik, Crocker, & is the internal consistency of the teacher ef-
MacAuliffe, 1982). ficacy measure? Phase 2 (multitrait-mul-
Another area that needs to be addressed timethod analysis): Does evidence of
is the relationship between teacher efficacy teacher efficacy gathered from different
and classroom behavior. Research on sources in different ways converge? Can
teacher use of time and direct instruction teacher efficacy be differentiated from other
may provide some possible linkage between constructs? Phase 3 (classroom observa-
teacher efficacy and student learning tion): Do high- and low-efficacy teachers
(Rosenshine, 1979). The description of ef- exhibit differential patterns of teacher be-
ficacious teachers indicates that they may be haviors in the classroom related to academic
TEACHER EFFICACY 571
focus, feedback, and persistence in failure thogonal factor structure was utilized as the final solu-
situations? tion. A relatively rigorous level for significance of factor
loadings (> .45) was selected as a criterion for inclusion
of individual items in the factor structure.
Method
Phase 2: Multitrait-Multimethod
Preliminary Scale Development Analysis
The development of the Teacher Efficacy Scale began
in a pilot study where 53 sample items were adminis- Subjects and instrumentation. Fifty-five teachers
enrolled in graduate education courses at a state uni-
tered to 90 teachers. The initial item pool of 53 items
versity in California completed two teacher efficacy,
was based on teacher interviews and an analysis of the verbal ability, and flexibility measures. The measures
literature that reported characteristics of teachers of teacher efficacy consisted of the Teacher Efficacy
identified by previous researchers as having a sense of Scale (Phase 1) and a more open-ended measure of
efficacy. Preliminary data analysis of pilot items in-
teacher efficacy in which the teachers were asked to
volved principal factor analysis, elimination of items check 10 of 20 variables contributing most to a student's
with poor variability, and maintenance only of those
items that loaded clearly on one of the substantial fac- success or failure in school. Among the 20 items were
10 teacher-related variables such as teacher rapport,
tors (Gibson & Brown, 1982). These remaining items with students, teacher ability to individualize instruc-
were revised to clarify ambiguities and assure proper
item construction. The revised Teacher Efficacy Scale tion, and teacher management skills, as well as 10 ex-
consisted of 30 items in Likert format. ternally related variables such as student intelligence,
parental support, and student home environment. The
measures of verbal ability and flexibility were compo-
Phase 1: Factor Analysis nents of the Teacher Aptitude battery of the Beginning
Teacher Evaluation Study, Phase 2,1973-76 (BTES;
Subjects and instrumentation. Subjects in Phase Ekstrom, 1975a). The measures of verbal ability were
1 were 208 elementary school teachers selected from 13 the Verbal Facility Test (Coleman et al., 1966) and
elementary schools (kindergarten through sixth grade) Controlled Associations Teat (French, Ekstrom, &
within two neighboring unified school districts. Their Price, 1963). The measures of flexibility were the
teaching experience ranged from 1 to 39 years. Finding Useful Parts and the Planning Test The latter
Twenty-six percent had 1 to 5 years of experience, 25% three tests were adaptations of instruments from the
had 6 to 10 years of experience, 23.7% had 11 to 20 years Educational Testing Service, Kit of Reference Tests for
of experience, and the remaining 16.3% ranged from 21 Cognitive Factors (French et al., 1963).
to 39 years of experience. Approximately 75% of the The Verbal Facility Test consists of 30 items in which
sample was female. The teachers completed the pre- the subject is asked to select from five options the best
viously piloted Teacher Efficacy Scale (see the Ap- word to be used in a blank to complete a sentence.
pendix), consisting of 30 items presented in a Likert Ekstrom (1975b) suggested that this instrument was
scale format in which the teachers selected a number (1 more nearly a measure of verbal reasoning than a simple
= strongly disagree to 6 = strongly agree) to indicate vocabulary test. The reliability in the BTES for the
their level of agreement with each individual state- Verbal Facility Test was found to be .72. The Con-
ment. trolled Association Test consisted of eight items. The
Procedure. All subjects were assured of anonymity subject was asked to write as many synonyms as possible
and were asked at respective faculty meetings to com- for each of these stimulus words. Communalities
plete the scale. Most subjects completed the scale (which represent lower boundary reliability estimates)
within 15 min. for this test for two groups of teachers in the BTES were
Data collection and analysis. We used a factor so- .67 and .54. Both the Verbal Facility Test and Con-
lution of the Statistical Package of the Social Sciences trolled Association Test had major loadings on the
(SPSS) of principal factoring to analyze the underlying verbal fluency factor and correlated .35 with each other
factor structure of teacher responses to the 30-item (Ekstrom, 1975a, 1975b).
Teacher Efficacy Scale. We used squared multiple The Finding Useful Parts Test consisted of 10 items.
correlations in the main diagonal of the correlation The subjects was asked to select from five options the
matrix as communality estimates and used an iteration one object that could be used as a makeshift substitute
procedure for improving estimates of communality. for a specified purpose when the object usually used is
Two factors were extracted based on Catell's screen unavailable. Commonalities established for this test
test as well as theoretical concurrence with Banduras were .70 and .63 (Ekstrom, 1975a).
two-factor model of self-efficacy. As suggested by The Planning Test was a test of sensitivity to prob-
Rummel (1970), both oblique and orthogonal rotations lems. The subject is asked to indicate what is wrong
were used to compare item loadings and degree of cor- with each of several plans presented for solving a variety
relations between factors. With delta value set at zero, of practical problems. Communalities established for
the oblique rotation revealed that the.two factors were this test were .84 and .64 (Ekstrom, 1975a). Both the
only moderately correlated (r = -.19), suggesting that Finding Useful Parts Test and the Planning Test loaded
the two factors represent related, but relatively inde- heavily on the flexibility factor in the BTES.
pendent, constructs. Because of this moderate corre- Procedure. Data were collected at the beginning of
lation as well as the fact that the same items yielded four consecutive class sessions. The Teacher Efficacy
significant item loadings for both solutions, the or- Scale was administered during the first session with
572 SHERRI GIBSON AND MYRON H. DBMBO
subjects given the same general instructions as in Phase personal teaching efficacy). These 16 teachers were
1. The second, more open-ended efficacy measure and asked to identify themselves by their social security
the Verbal Facility Test were administered in the sec- number and to grant permission for classroom obser-
ond session. The Controlled Associations Test and vations. Twelve of the 16 teachers granted permission
Finding Useful Parts Test were administered in the for observation, and the 8 teachers with the most dis-
third session, and the Planning Test was administered crepant and equally distributed scores were selected
in the fourth class session. The verbal ability and from among these 12. In order to examine the combined
flexibility measures were administered according to effects of both factors, a composite teacher efficacy score
given standardized procedures. was calculated for each teacher by subtracting their
Data collection and analysis. Data from the three Factor 2 score from their Factor 1 score. Although we
traits (teacher efficacy, verbal ability, and flexibility) recognize the problem of lowered reliability for differ-
were analyzed across two methods of measurement ence scores, the reliability coefficients established in
(closed-ended and open-ended), using multitrait- Phase 1 for both factors and total scale (.78, .75, and .79,
multimethod matrix method of analysis delineated by respectively) as well as theoretical concerns support the
Campbell and Fiske (1959). Correlations of variables use of composite scores in this initial study.
within and between methods were computed for pre- Seven observers were trained to collect data utilizing
sentation to examine convergence of the construct videotaped lessons. Interrater reliability ranged from
across methods and to determine discriminability of .73 to .91. Each of the 8 teachers was observed for ap-
teacher efficacy from the other constructs. proximately 7.5 hr during morning "academic" class-
room time and was observed by three different ob-
servers.
Phase 3: Classroom Observation Data collection and analysis. Two types of mea-
sures were derived from the raw coding sheets. Total
Subjects and instrumentation. A subsample of 8 minutes allocated by the teachers to each of the obser-
teachers (4 high efficacy and 4 low efficacy) were se- vation categories were used to analyze the data from the
lected from the 208 teachers who participated in Phase teacher-use-of-time instrument. The second set of
1 of the study to serve as subjects in a pilot investigation measures was percentage figures derived from the
of the relationship between teacher efficacy and teacher question-answer-feedback sequence instrument in
classroom behaviors. All teachers were from 2 of the which absolute frequency differences across teachers
13 schools to minimize the influence of organizational were statistically controlled in order to .allow a com-
variables on teacher efficacy (Fuller, Wood, Rapaport, parison of relative differences between groups (Brophy
& Bornbusch, 1982). & Good, 1970; Cooper & Baron, 1977).
The observation instruments used to code classroom The teacher-use-of-time instrument yielded time
behavior were a teacher-use-of-time measure and a allocation measures of high- and low-efficacy teachers
question-answer-feedback sequence measure adapted for the following categories: (a) daily rituals, (b) tran-
from Good and Brophy (1973). The teacher-use-of- sitions, (c) whole class, (d) small group academic, (e)
time instrument codes a teacher's behavior whenever checking seatwork, (f) preparation or paperwork, (g)
activities are introduced or changed by the teacher. It intellectual games, (h) unfocused small talk, and (i)
enables one to measure the proportion of time a teacher recess. Total academic time was also computed by
spends on activities related to teaching and academic summing the total minutes of categories c, d, and e for
learning. The question-answer-feedback sequence each group. The remaining categories observed were
instrument codes teacher and student dyadic behavior summed to yield a total nonacademic time variable.
during question and answer interchanges. The quality Measures of teacher praise and criticism were con-
of a student's response and the nature of the teacher's trolled for absolute frequency differences in students'
feedback to the student are coded. correct and incorrect responses. The measure of
Procedure. Factor scores were computed for all teacher praise was derived by dividing the number of
teachers who participated in Phase 1. A high score on teacher praises by the number of correct student re-
Factor 1 (Personal Teaching Efficacy) indicated high sponses. This praise-per-successful interaction mea-
personal efficacy (or self-efficacy), whereas a low score sure indicates the frequency of rewarding behavior with
on Factor 2 (Teaching Efficacy) indicated high teaching the number of correct answers held constant. Similarly,
efficacy (or outcome expectancy). Bandura (1978) the number of teacher criticisms was divided by the
stated that observable behavior is predicted by con- number of incorrect student responses.
sidering both a person's self-efficacy and his or her Data were also analyzed to determine teacher per-
outcome expectancies. In an attempt to address both sistence in a failure situation (when student answered
factors simultaneously, 16 teachers from the two se- unsuccessfully). Teacher persistence was defined as
lected schools were selected whose Factor 1 and Factor the ratio of feedback interactions in which a teacher
2 scores fell in opposite ends of the frequency distri- either repeated the queston, provided a clue, or asked
bution. Eight high-efficacy teachers were selected a new question following a student's incorrect response.
whose Factor 1 scores fell within the top 6% of the fre- The number of teacher persistent behaviors was divided
quency distribution (high personal teaching efficacy) by the total number of student incorrect responses.
and whose Factor 2 scores fell within the bottom 22% Conversely, lack of persistence in failure situations was
of the distribution (high teaching efficacy). Eight defined as the ratio of feedback interactions in which
low-efficacy teachers were selected whose Factor 2 a teacher either gave the answer, asked another student,
scores fell within the top 27% of the frequency distri- or allowed another student to call out the answers when
bution (low teaching efficacy) and whose Factor 1 scores a student failed to respond or responded unsuccessfully.
fell in the bottom 45% of frequency distribution (low The number of teacher nonpersistent behaviors was
TEACHER EFFICACY 573
divided by the total number of student incorrect re- counting for 10.6% of the total variance,
sponses. One-tailedj tests with the teacher as the unit Each of the remaining factors accounted for
of analysis were used to analyze differences between , e9Sthan
,, „„, ,.,,hetota
6% of
, , , . m 1.1 i
high- and low-efficacy teachers in teacher behaviors } * * variance. Table 1
related to academic focus, teacher feedback, and teacher includes these items that loaded on each
persistence. factor.
Factor 1 appears to represent a teacher's
Phase 1: Factor Analysis gense of personaiteaching efficacy, or belief
Results • that one has the skills and abilities to bring
about student learning. This dimension
Three questions were the focus of Phase corresponds to the specific item used in
1: What are the dimensions of Teacher Ef- previous research (Herman & McLaughlin,
ficacy? How do these dimensions relate to 1977; Armor et al., 1976), "If I really try hard,
Bandura's theory of self-efficacy? What is I can get through to even the most difficult
the internal consistency of the teacher eff i- or unmotivated students." All of the items
cacy measure? included in Factor 1 reflect the teacher's
Two substantial factors emerged from the sense of personal responsibility in student
factor analysis, with Factor 1 accounting for learning and/or behavior and correspond to
18.2% of the total variance and Factor 2 ac- Bandura's self-efficacy dimension.
Table 1
Factor Item Loading
Factor
Item no. loading Item
The second dimension that is reflected in necessary tasks to bring about positive stu-
Factor 2 represents a teacher's sense of dent change and is clearly represented by the
teaching efficacy, or belief that any teacher's first factor, Personal Teaching Efficacy.
ability to bring about change is significantly Internal consistency reliability of the
limited by factors external to the teacher, Teacher Efficacy Scale is evident in the
such as the home environment, family current study and is essential to the suc-
background, and parental influences. This cessful development of a scale of teacher
dimension reflects the teacher's belief about efficacy. Not only is reliability of the scale
the general relationship between teaching important in and of itself but it is crucial to
and learning and is represented by the sec- the establishment of construct validity.
ond item used in previous research (Herman Results indicate that teacher efficacy, as
& McLaughlin, 1977; Armor et al., 1976), measured by the Teacher Efficacy Scale, is
"When it comes right down to it, a teacher multidimensional and comprises at least two
really can't do much because most of a stu- clearly distinguishable factors. These fac-
dent's motivation and performance depends tors are only moderately correlated, which
on his or her home environment." This further emphasizes the need to investigate
second factor clearly corresponds to Ban- teacher efficacy from a multidimensional
dura's outcome expectancy dimension. approach. The two scales that resulted from
Analysis of internal consistency reliabili- the current study appear to measure reliably
ties yielded Cronbach's alpha coefficients of these constructs in terms of internal consis-
.78 for the Personal Teaching Efficacy factor, tency and lend support to the applicability
.75 for the Teaching Efficacy factor, and .79 of Bandura's conceptualization of self-effi-
for the total 16 items. cacy in research on teacher efficacy.
Because acceptable reliability coefficients
resulted from only 16 of the original 30 items, Phase 2: Multitrait-Multimethod
further research is suggested with a revised Analysis
scale of 16-20 items. Analyses in all three
phases of the present study were based on Results
responses to the 16 of the original 30 items
that yielded significant loadings on one of The following questions directed Phase 2
the two factors. of the research study: Does evidence of
teacher efficacy gathered from different
Discussion sources in different ways converge? Can
teacher efficacy be differentiated from other
The two resulting dimensions clearly constructs?
conform to Bandura's conceptualization of Intercorrelations between three traits
self-efficacy and support Ashton and Webb's (verbal ability, flexibility, and teacher effi-
(1982) model of teacher efficacy. As men- cacy) across two methods (closed-ended and
tioned previously, Bandura proposed that open-ended format) are presented in a
one's behavior is determined by both a gen- multitrait-multimethod matrix in Table 2.
eral outcome expectancy (belief that be- This matrix can help one identify categories
havior will lead to desirable outcomes) as that pass specified tests of convergent and
well as a sense of self-efficacy (belief that one discriminant validity.
has the requisite skills to bring about the Evidence of teacher efficacy gathered
outcome). When applied to the construct through both a closed-ended additive scale
of teacher efficacy, outcome expectancy format as well as a more open-ended format
would essentially reflect the degree to which converge, as indicated by a positive correla-
students can be taught given their family tion of .42 (p < .001). All three traits-(i.e.,
background, socioeconomic status (SES), verbal ability, flexibility, and teacher effi-
and school conditions. This dimension is cacy) pass the criteria for convergent validity
clearly represented by the second factor, because the validity diagonal values of all
Teaching Efficacy. Bandura's self-efficacy three traits were found to be significant be-
dimension would indicate a teacher's rating yond the .05 level (.30, .39, and .42, respec-
of his or her own abilities to perform the tively). Because all three traits pass the test
TEACHER EFFICACY 575
Table 2
Multitrait-Multimethod Matrix
Method 1 Method 2
(closed-ended) (open-ended)
Verbal Verbal
Method ability Flexibility Efficacy ability Flexibility Efficacy
1 (closed-ended) (.72)
Verbal ability
Flexibility
Efficacy .25
.27 ,22
for convergent validity, the next step is to between verbal ability and teacher efficacy
examine for discriminant validity. not having method in common (r = .08, r =
Discriminant validity must be assessed in .09) and between verbal ability and flexi-
two steps. First, each validity value must be bility not having method in common (r = .12,
compared with all values in its row and col- r = .26). The correlation of flexibility when
umn in the heterotrait-heteromethod measured by two methods (r = .39) exceeds
(broken line) block to determine whether the both the correlation between flexibility and
correlations between different methods teacher efficacy not having method in com-
(heteromethod) of measuring the same trait mon (r = .21, r = -.06) and between flexi-
exceed correlations between that trait and bility and verbal ability not having method
other traits (heterotrait) not having method in common (r = .12, r = .26).
in common. The second step in determining discrim-
The validity value for teacher efficacy inant validity is completed by comparing
meets this first criterion, lending support to each trait's validity value with values in the
discriminant validity. The correlation of heterotrait-monomethod (solid line) trian-
teacher efficacy when measured by two gles in which that trait is involved. This
methods (r = .42) exceeds both the correla- step determines whether the correlation
tions between teacher efficacy and verbal between different methods (heteromethod)
ability not having method in common (r = of measuring the same trait (monotrait) ex-
.08, r = .09) and between teacher efficacy ceeds correlations between that trait and
and flexibility not having method in com- other traits that have method in common.
mon (r = ,21, r = -.06). Teacher efficacy as well as verbal ability
The validity values for verbal ability and and flexibility meet this criteria, although
flexibility also meet this first criterion for the validity coefficient for verbal ability (r
discriminant validity. The correlation of = .30) only slightly exceeds some of the other
verbal ability when measured by two meth- comparative values (r = .27, r = .25). The
ods (r = .30) exceeds both the correlation values in the monomethod (closed-ended),
576 SHERRI GIBSON AND MYRON H. DEMBO
allocate similar proportions of time, with the repeated the question, provided a clue, or
exception of the use of intellectual games asked a new question), the mean difference
and preparation or paperwork. It is inter- was again in favor of the high-efficacy
esting to note that although low-efficacy teachers. A significant difference did result
teachers spent a relatively small portion (2%) in lack of persistence, i(6) - 3.29, p < .01.
of their time (M = 10.5 min) in intellectual Low-efficacy teachers were more likely to go
games, no observations of this type of ac- on by giving the answer, asking another
tivity occurred in rooms of high-efficacy student, or allowing another student to call
teachers, t(6) = 2.01, p < .05. High-efficacy out before a student gave the correct re-
teachers also tended to spend more time (M sponse. It appears that although both high-
= 39.5 min) in preparation or paperwork and low-efficacy teachers did provide stu-
than low-efficacy teachers (M = 16.3 min). dents further opportunities to correct their
Table 4 presents the means for praise per responses (persistence), high-efficacy
correct student answer and criticism, per- teachers were more effective in leading stu-
sistence, and lack of persistence per incorrect dents to correct responses through their
student answer. Analysis of these teacher questioning, whereas low-efficacy teachers
feedback patterns yielded a significant dif- would go on to other students or another
ference between groups on teacher criticism question (lack of persistence) before the
following a student's incorrect response, t(6) student arrived at the correct response.
= 5.17, p < .01. When students gave an in-
correct response to low-efficacy teacher Discussion
questions, 4% of these interactions resulted
in teacher feedback in the form of criticism. Academic focus. A number of studies
However, no observations of criticism oc- have revealed that successful teachers
curred in any of the high-efficacy teachers' maintain a strong academic focus and spend
rooms. Although the difference in praise less time in nonacademic activities. At-
per correct answer is small, the mean dif- tempts to look at global academic time by
ference is in favor of high-efficacy collapsing academic and nonacademic
teachers. categories of the teacher-use-of-time mea-
Analysis of teacher persistence and lack of sure failed to yield significant results. As
persistence indicates that although there was noted by other researchers, the amount of
not a significant difference in persistence time a teacher allocates for academic in-
(defined as the ratio of feedback interactions struction, the time a teacher is actually en-
to student failures in which a teacher either gaged in that instruction, and the time a
student is engaged may all be quite different
Table 4 amounts of time. Rosenshine (1979) re-
Means and Standard Deviations of Praise per ported that in studies considering only al-
Correct Answer and Criticism, Persistence, located time most of the results tend to be
and Lack of Persistence per Incorrect Answer nonsignificant. It appeared that academic
time coding in the present study was more a
High efficacy Low efficacy reflection of allocated time and organization
Variable M SD M SD than of academically engaged time or aca-
demic focus. Although the teacher-use-
Praise per correct of-time measure coded the activity that the
answer 0,03 0.03 0.01 0.02 teacher had allocated time for and was en-
Criticism per gaged in, it did not reflect in any way the
incorrect answer 0.00* 0.00 0.04* 0.02 students' engagement rates, and anecdotal
Persistence per
incorrect answer 0.75 0.37 0.66 0.34 observation data as well as observed group-
Lack of persistence ing differences suggested that differences in
per incorrect student engagement rates may have existed
answer 0.38* 0.11 0.67* 0.12 between high- and low-efficacy teachers.
Note. Means and standard deviations are presented In the present study, low-efficacy teachers
as ratio figures. spent an average of 2.4% of their time (M =
* p < .01, one4ailed. 10.5 min) in intellectual games. Although
578 SHERRI GIBSON AND MYRON H. DEMBO
as those inherent within the construct of factors of the scale to the two items previ-
teacher efficacy may influence feedback ously discussed (Berman & McLaughlin,
behaviors and teacher persistence. Those 1977; Armor et al, 1976). Also, construct
teache'rs who in general expect students to validation should continue to be investigated
learn and who have confidence in their across different populations and settings.
ability to teach may communicate higher Further factor analytic studies, including use
expectations by providing less criticism to of LISREL procedures, should be used to
students and persisting with students until confirm further the trait and factor struc-
they respond correctly rather than going on ture.
to another student or another question. We Third, investigation of the relationships
must emphasize that caution must be taken between teacher characteristics (i.e., sex,
in reaching any conclusion regarding the years of teaching experience, grade levels,
classroom behaviors of high- and low-effi- and personal attributes) and sense of efficacy
cacy teachers until larger samples of teachers is needed. Relationships with situational
are studied. The present data do suggest and organizational variables should also be
several important hypotheses for future re- investigated because teacher efficacy is likely
search. to be situation specific and may not gener-
alize from one setting to another (e.g., high
General Discussion SES school to low SES school).
Fourth, the relationship between teacher
General conclusions can be made within efficacy and teacher task persistence should
the limitations of the present study. be expanded beyond analysis of teacher
Teacher efficacy is multidimensional, con- feedback responses. Other task persistence
sisting of at least two dimensions that cor- measures, such as use of mastery teaching
respond to Bandura's two-component model techniques, criteria for advancement in
of self-efficacy. The measures of teacher curricular materials, and requests that stu-
efficacy identified through different meth- dents make corrections, should be ex-
ods converge, while at the same time they plored.
can be differentiated from verbal ability and Fifth, any subsequent investigations of the
flexibility. Finally, preliminary classroom relationship between teacher efficacy and
observation data suggest that teacher effi- observable classroom process variables
cacy may influence certain patterns of should include measures of student en-
classroom behavior known to yield achieve- gagement rates or on-task behaviors.
ment gains. Teacher behavior variables that can increase
A number of research issues should be student engagement rates should be ex-
addressed in future investigations: First, plored such as how feedback is provided to
further elements of Bandura's theory of students who are working independently and
self-efficacy should be explored as they re- physical proximity of teacher to students.
late to teacher efficacy. For example, Ban- Sixth, the relationship between teacher
dura speaks of dimensions of generality, efficacy (and each of its components) and
magnitude, and strength of self-efficacy. student achievement should be examined.
When applied to teachers, it may be that It must be determined whether measures of
generality relates to the extent to which a teacher efficacy predict student achievement
teacher feels efficacious in a variety of in different settings and type of courses.
teaching situations rather than a narrowly Finally, study of the relationship between
defined range of situations. Magnitude may teacher efficacy and teacher decision mak-
be reflected in the degree of difficulty of the ing, particularly in the area of classroom
task for which a teacher feels efficacious, and organization and management, is needed.
strength may be manifest in the relative ease Perhaps decisions such as student grouping
or difficulty with which it may be modi- decisions are based on the sense of confi-
fied. dence a teacher feels in achieving instruc-
Second, further research on the validation tional goals, being able to manage the be-
and refinement of the Teacher Efficacy Scale havior of students, or being in control of the
is needed, especially comparing the two instructional setting. Intervention and/or
580 SHERRI GIBSON AND MYRON H. DEMBO
10). St. Louis, MO: Central Midwestern Regional Park, CA: Stanford Research Institute.
Educational Lab. (ERIC Document Reproduction Webb, R. (1982, March). Teaching and the domains
Service No. ED 185 165) of efficacy. Paper presented at the annual meeting
Stallings, J., & Kaskowitz, D. (1974). Follow through of the American Educational Research Association,
classroom observation evaluation, 1972-73. Menlo New York.
Appendix
Please indicate the degree to which you agree or disagree with each
statement below by circling the appropriate numeral to the right of each
statement.
18. If students are particularly disruptive one day, I ask myself what I have
been doing differently. 1 2 3 4 5 6
21. If a student masters a new math concept quickly, this might be because
I knew t h e necessary steps i n teaching that concept. 1 2 3 4 5 6
22. Parent conferences can help a teacher judge how much to expect from a
student by giving the teacher an idea of the parents' values toward
education, discipline, etc. - 1 2 3 4 5 6
30. Even a teacher with good teaching abilities may not reach many
students. 1 " 2 3 4 5 6