1

Bibliothèque Form@PEx
(accès libre)

Recent Research on Teaching Behaviors and

Student Achievement 1

Barak Rosenshine
(University of Illinois at Urbana-Champaign)

Although studies on teacher behaviors and teacher effects have been reported since
1940, the modern era of this research began in 1957 with the work of investigators such as
Flanders, Medley, and Mitzel. Even since then, the number of studies has been small. Fewer
than 25 studies have been conducted on any specific variable such as teacher praise or
teacher questions, and these studies are spread across all grade levels, subject areas, and
student backgrounds.

The number of investigators in this field is also small. There are no more than 12
researchers 2 or groups of researchers currently studying the relationship between class-
room instruction and student achievement. These 12 are spread out across the different
instructional contexts and variables being studied. The number of researchers actively col-
lecting data in any given year is much smaller than 12. Furthermore, these researchers
have met only twice, as a group, to discuss common findings, inconsistencies, and prob-
lems.

Although the recent studies, summarized in this issue, represent methodological and
conceptual expansion of previous work, research on observed teaching behavior is new,
sparse, and not always consistent in results. What we have learned to date is offered more
as hypotheses for future study than as validated variables for the training and evaluation of
teachers. Although practitioners can easily amass a large number of questions on
teaching methods for which they would like clear answers, at the rate we are going it
will be years before many of these questions are even studied.

1 . Journal of Teacher Education, Spring 1976, vol. 27 n° 1, pp. 61-64.

2 . The 12 or so researchers, groups of researchers, and their organizations are:
- David Berliner, Charlie Fisher, and Len Cahen – Far West Laboratory;
- Walter Borg – Utah State University;
- Jere Brophy and Carolyn Evertson – Texas Research and Development Center in Teacher Education;
- Homer Coker – Carrolton State College, Georgia, and Don Medley – University of Virginia;
- N.L. Gage – Stanford Research and Development Center in Teaching;
- Meredith Gall, William Tikunoff, and Betty Ward – Far West Laboratory;
- Tom Good – University of Missouri;
- Gene Hall, Sue Loucks, Gary Borich, and Robert Peck – Texas Research and Development Center
in Teaching;
- Gaea Leinhardt, Margaret Wang, and William Cooley – Learning Research and Development Center, University
of Pittsburg;
- Fred McDonald – Educational Testing Service;
- Robert Soar – University of Florida;
- Jane Stallings – Stanford Research Institute.

www.formapex.com Bibliothèque Form@PEx

 2

In 1975, a number of these researchers met in San Diego and again in Austin to pre-
sent and discuss their recent research, and a number of these papers are summarized be-
low. Unfortunately, these papers do not fit into one piece. Each investigator worked on a
problem individually, and there has been little communication between them. Hopefully,
now that people have been brought together to share ideas, there can be more coordination
and cooperation.

Summary of Recent Studies

The topics of these papers are too different to permit easy grouping. Yet, I intuitively
grouped them into categories: experimental studies of questioning, tutoring, program im-
plementation, generic and specific teaching skills, and teaching low SES students.

Experimental Studies of Questioning

Teacher questions have been a major area in all teacher training programs, yet the
research base for our practice is woefully thin. The two experimental studies of questioning -
– by Ward and Tickunoff, 1975 3 and Stanford Program on Teaching Effectiveness 4 – are
exciting because they represent serious and well-designed attempts to overcome design defi-
ciencies of previous studies.

Overall, both studies found that classes where students were asked more recall
questions did slightly better on the recall tests, whereas all classes did equally well on inte-
grative questions, no matter what percent were asked in the class. In an additional analysis
reported in Ward and Tickunoff, it was found that low ability students did best with factual
questions, and without probing and redirection, whereas high ability students did best with
probing and redirection.

Thus, conclusions emerge from these studies. First, factual questions appear to be
functional for low ability students, a finding which is also supported by recent correlational
studies 5, and are not necessarily dysfunctional for high ability students. Second, we are as
yet unaware of the optimal types and sequencing of questions and responses for teaching
the ability to draw inferences or apply learning. Finally, the optimal sequences may be dif-
ferent for different types of students and outcomes, but we still do not know what these se-
quences are. Put another way, the continual bromides that factual questions are bad
and higher level questions are good, were not supported by well-designed research.

Experimental Studies of Tutoring

The Far West Laboratory studies of tutoring 6 were attempts to validate their mini-
course on mathematics tutoring. The results suggested that training in tutoring, or tutoring
itself, may not be as important as expected. If students are given an extra half hour a day of
mathematics instruction, they learn more, regardless of whether this instruction takes place
in regular class or in tutoring sessions, and regardless of whether the teachers are trained

3 . B.A. Ward and W.J. Tikunoff, Application of Research to Teaching, Report A75-2 (San Francisco, Calif.: Far West
Laboratory for Educational Research, 1975).
4 . Stanford Program on Teaching Effectiveness, A Factorially Designed Experiment on Teacher Structuring, Soliciting,

and Reacting (Stanford, Calif.: Stanford Center for Research and Development in Teaching, 1976).
5 . J.A. Stallings and D.H. Kaskowitz, “A Study of Follow Through Implementation”, Paper presented to the annual

meeting of the American Educational Research Association, April 1975 (Menlo Park, Stanford Research Institute).
J.E. Brophy and C.M. Evertson, Process-Product Correlations in the Texas Teacher Effectiveness Study: Final Re-
port (Austin, Tex.: The University of Texas, 1974).
R.S. Soar, Follow Through Classroom Process Measurement and Pupil Growth (1970-71): Final Report (Gainesville,
Fla.: College of Education, University of Florida, 1973).
6 . Ward and Tikunoff, op. cit.

B.J. Williams et al., Math Tutoring Study (San Francisco, Calif.: Far West Laboratory for Educational Research
and Development, 1975).

www.formapex.com Bibliothèque Form@PEx

 3

or untrained in tutoring. A similar study, but one which only involved 30 minutes of teach-
ing or tutoring found that tutoring yielded significantly greater achievement than classroom
teaching 7. Although this study raises questions about common practice, it remains for fu-
ture research to tell us effective procedures for improving tutoring or teaching.

Levels of Implementation
As research is extended and findings become incorporated into curriculum packages,
the monitoring of curriculum implementation becomes important. Comparisons of curricu-
lum packages and regular programs are not particularly useful if one does not know how
well the curriculum is being implemented. Both Stallings and Kaskowitz and Soar used ob-
servational data to measure curriculum implementation. Hall and his associates 8 comple-
ment this work with their development of a teacher checklist (Stages of Concern) and an
interview schedule (Levels of Use) to obtain information on the thought and practice of
teachers who are implementing an innovation. The Levels of Use Instrument 9 demonstrated
that, as teachers gain experience, they shift from mechanical use of an innovation to stabi-
lized and refined use. This instrument might be useful for monitoring implementation.

Generic and Specific Skills

Two studies – the ethnographic study of the Far West Laboratory (FWL) 10 and the
observational study by Educational Testing Service (ETS) 11 – are examples in which re-
searchers developed their variables after they had recorded the events which occurred in
observed classrooms. A second similarity was that both studied teaching in four contexts:
reading and mathematics in both the second and fifth grades. Finally, in both studies, no
differentiation was made for student SES background.

One question which the FWL study addressed is whether teaching skills are generic
or specific. Overall, the FWL study found that over a third of the variables, or dimensions,
which they studied were generic, that is, these 21 dimensions were significant in each of the
four contexts. In addition, these researchers found that reading and mathematics were more
similar at second grade than at fifth, and that across the grades reading was more similar
than math. There were no inconsistent dimensions, that is, dimensions which were signifi-
cantly positive in one context but negative in another.

In sharp contrast, the ETS study 12 found no dimensions which were generic, found
that reading and math were most similar at fifth grade, and found that across the grades
math was more similar than reading. Finally, there were almost as many inconsistent di-
mensions as consistent dimensions.

Thus, these two studies give different pictures of generic and specific skills. The FWL
study argues for a mixture of generic and specific skills, with skills in second-grade reading
most general and skills in fifth-grade mathematics most specific. In contrast, the ETS study
presents the case that effective skills are highly unique in each context.

7 . R.B. Bausell, W.B. Moody, and F.N. Walzl, “A Factorial Study of Tutoring Versus Classroom Instruction”, Ameri-
can Educational Research journal 9(1972): 591-99.
8 . G.E. Hall, The Effects of “Change” on Teachers and Professors — Theory, Research, and Implications for Decision

Makers (Austin, Tex.: Research and Development Center for Teacher Education, the University of Texas, 1975).
9 . Gene E. Hall et al., “Levels of Use of the Innovation: A Framework for Analyzing Innovation Adoption”, ]ournal of

Teacher Education XXVI, no. 1 (Spring 1975): 52-56.

10 . W.J. Tikunoff, D. Berliner, and R. Rist, An Ethnographic Study of Forty Classrooms, Technical Report no. 75-10-

5 (San Francisco, Calif.: Far West Laboratory for Educational Research, 1975).
11 . F.J. McDonald, et al., Beginning Teacher Education Study, Phase II: Final Report (Princeton, N.J.: Educational

Testing Service, 1975).

12 . The results of generic and specific variables in the ETS study came from my reading of a draft of their final

report. McDonald (1975) did not discuss generic and specific results directly, but the lack of generic results is clear
in that paper.

www.formapex.com Bibliothèque Form@PEx

 4

Such differing results are puzzling. One can argue that differences in coding proce-
dures, selection of sample, and length of instruction are so large that comparing these stud-
ies is meaningless. At any rate, additional studies are essential before we are clear about
which skills are generic, grade level specific, or subject area specific.

Studies of Primary Grade Low SES Students

Three of the investigations 13 covered the same context – primary grade reading and
mathematics for children from low SES backgrounds. All three conducted correlational
studies between prespecified instructional variables and student achievement. There was an
exciting convergence of significant results across these three studies. Consistent positive
correlations were obtained for direct time, narrow questions, teacher positive feedback, stu-
dent attention to task, and supervised student study in small or large groups. The results
on eight common variables in these studies are summarized in Table 1.

TIME
In these studies, the amount of time spent directly on instruction was significantly related to
student achievement. For example, Stallings and Kaskowitz coded the observed time spent
on reading and mathematics activities, and obtained significant and positive results. In ad-
dition, all three investigators coded time spent on nonacademic activities (such as dramatic
play, games, or questions about home and family) and all three obtained consistent negative
correlations.
Other recent studies also support the importance of direct time; Wiley and Harnischfeger 14
found that the average number of hours of schooling per year was positively and significant-
ly related to achievement in reading and mathematics directly or indirectly. In other studies,
the coding of content covered has yielded significant positive correlations with achieve-
ment 15. Similarly, in a review of studies comparing different curriculum programs, Walker
and Schaffarzick 16 concluded that the outcomes of different curriculum programs mirrored
the content that was taught. That is, new curriculums were more effective when the post-
test measured content relevant to them, whereas traditional curriculums were slightly supe-
rior when traditional post-tests were used.
Overall, then, both direct instructional time and content covered have been positively related
to achievement, whereas time on nonacademic activities has been negatively related.

QUESTIONS
In all three studies, the frequency of factual, single-answer questions was correlated posi-
tively and significantly with achievement, whereas the frequency of more complex, difficult,
or divergent questions had negative correlations. Brophy and Evertson also found that for
low SES students, the percentage of correct answers was positively and significantly related
to achievement. The results on questions suggest that for this context (primary grade stu-
dents from low SES backgrounds) it may be preferable to proceed in small steps and ask
factual questions at the child's level.
The above results, on the functional value of factual questions and undemonstrated value
for higher order questions, parallel those obtained in the two cited experimental studies.
The lack of significant results for complex or higher level questions has puzzled all the
researchers, and led us to conclude that we need to rethink what is meant by types of
questions and their effects.

13 . Brophy and Evertson, op. cit. ; Soar, op. cit. ; Stallings and Kaskowitz, op. cit.
14 . D.E. Wiley and A. Harnischfeger, “Explosion of a Myth: Quantity of Schooling and Exposure to Instruction,
Major Educational Vehicles”, Educational Researcher 3 (1974): 7-12.
15 . Rosenshine and Furst, op .cit.
16 . D. Walker and J. Schaffarzick, “Comparing Curricula”, Review of Educational Research 44(1974): 83-111.

www.formapex.com Bibliothèque Form@PEx

 5

TABLE 1
SUMMARY OF RESULTS ON COMMON VARIABLES IN RECENT STUDIES ON PRIMARY
GRADE LOW SES IN READING & MATH

Kaskowitz

Evertson
Variable

Stalling

Brophy

Others
Soar
Direct time on academic activities + +
Time on noncurricular activities — — —
Allotted time for school or instruction o +
Content covered +
Direct, narrow questions + + + +
Higher order, open questions — — — o

Student attention to task + o, + o o, +

Student inattention, misbehavior — — —
Student in large group + +
Students in independent study without teacher — —
Students in supervised independent study +

Praise, & adult positive feedback + + o, + o

Criticism, & adult negative feedback + — o, — —
Accepting student comments + +

Student comments-relevant o o +
Student comments-irrelevant — o o
Student questions-relevant — o o
Student questions-irrelevant — o —

Teacher initiation —,+ —, o

Teacher management requests, requests for order o —

+ = positive & significant correlations

O = nonsignificant and mixed correlations
— = negative and significant correlations

STUDENT INATTENTION
Student inattention was consistently, significantly, and negatively related to achievement in
all three studies. Similar negative results have usually also been obtained by McDonald 17
and other investigators. The results for student attention or on-task behavior were positive,
but the correlations were not as high or consistent as those for inattention; again, similar
results were obtained in other studies.

17 . McDonald, et al., op. cit.

www.formapex.com Bibliothèque Form@PEx

 6

WORK GROUPINGS
For both Stallings and Kaskowitz, and for Soar, positive and significant correlations were
obtained for students working in groups or doing seatwork under supervision. Both investi-
gators also found negative correlations for children working independently without supervi-
sion. The researchers agreed that independent study without supervision does not ap-
pear successful unless students are first taught how to work independently.

ADULT FEEDBACK. PRAISE AND CRITICISM

Overall, teacher praise showed consistent, positive, but low correlations with student
achievement. Praise of student academic responses had higher correlations than praise for
student behavior. However, the results were not consistent for academic criticism — criti-
cism following a student answer. Stallings and Kaskowitz found that criticism following an
incorrect answer has positive correlations with achievement, whereas Brophy and Evertson
found significant negative correlations for the same variable, and Soar obtained negative-
but-low correlations.

STUDENT-INITIATED AND TEACHER-INITIATED COMMENTS

With one exception all types of student-initiated talk, whether academic or nonacademic,
yielded negative or low correlations. In discussing this, researchers concluded that student-
initiated talk does not appear to be as important for this type of achievement as once
thought.
The frequency of teacher-initiated comments did not show consistent results in these stud-
ies – sometimes yielding significant positive and sometimes significant negative correlations.
In discussing this finding, the researchers suggested that coding the probable cause of a
teacher's comment (e.g., management, clarification, feedback, or motivation) might be help-
ful in future research.

MANAGEMENT AND CONTROL REQUESTS

There were no consistent results for teacher management or control statement. Brophy and
Evertson found that such statements (e.g., “silence” “get to work”) were negatively related to
achievement. However, Stallings and Kaskowitz found significant positive correlations (in
third grade) both for adult punishment of children and for negative corrective feedback on
academic matters, while Soar found nonsignificant relationships between control statements
and achievement.

Direct Instruction for Low SES Pupils

In the three studies on teaching low SES students in primary grades, there was con-
vergence on an optimal pattern for this instruction, which might be labeled as direct instruc-
tion. In direct instruction a great deal of time is spent on academic activities, with a predom-
inance of seatwork using structured materials. Teacher and workbook questions are narrow
and direct, usually with a single correct answer. Teachers or materials provide immediate
feedback using praise and acknowledgement of student answers. Students work in groups
supervised by the teacher with little free time or unsupervised activity, resulting in less off-
task student behavior.
The results of some studies also suggest that in direct instruction, the teacher is the
dominant leader who decides which activities will take place. The learning is approached in
a direct business-like manner and is organized around questions posed by the teacher or
the materials. Materials and instruction are systematically organized, proceeding within
small steps. Goals are clear and known to the students. Yet, within this task setting the
teacher is warm and convivial, frequently giving praise and encouragement to the students
for academic work. Future investigations will be needed to generalize these findings, both for
low SES and middle SES students.

www.formapex.com Bibliothèque Form@PEx

 7

Overall

The recent studies covered a number of topics: questioning, tutoring, implementing

innovations, generic and specific skills, and teaching low SES students. The studies on
questioning and tutoring serve mainly to raise questions about current practice while leav-
ing answers to future research. The studies on generic teaching skills provide a fascinating
but unresolved puzzle. The studies of teaching low SES students yield a convergence on a
pattern of variables which was labeled direct instruction and such a pattern awaits replica-
tion and modification. One is cautiously optimistic about some of the results; yet, the rate at
which this research is conducted is very slow and answers may not be immediately forth-
coming.

BIBLIOGRAPHY
Flanders, N.A. Analyzing Classroom Behavior, New York: Addison-Wesley, 1970.
Gall, M. et al. The Effects of Teacher Use of Questioning Techniques on Student Achievement and Atti-
tudes. San Francisco, Calif.: Far West Laboratory for Educational Research and Development,
1975.
Good, T.L. and Grouws, D.A. Process Product Relationships in 4th Grade Mathematics Classes. Co-
lumbia, Mo.: College of Education, University of Missouri, 1975.
Rosenshine, B. and Martin, M. “Teacher Education and Teacher Behavior”. Educational Researcher 3
(1974): 11-14.
Stallings, J.A. and Kaskowitz, D.H. Follow Through Classroom Observation Evaluation (1972-73).
Menlo Park, Calif.: Stanford Research Institute, 1974.
Tallmadge, K. The Development of Project Information Packages for Effective Approaches in Compensa-
tory Education. Technical Report UR-254. Mountain View, Calif: RMC Research Corp., 1974.

www.formapex.com Bibliothèque Form@PEx