Begle Revisited: Teacher Knowledge and Student Achievement in Algebra

Author(s): Theodore A. Eisenberg

Source: Journal for Research in Mathematics Education, Vol. 8, No. 3 (May, 1977), pp. 216-
222
Published by: National Council of Teachers of Mathematics
Stable URL: http://www.jstor.org/stable/748523
Accessed: 08-11-2017 04:57 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/748523?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

National Council of Teachers of Mathematics is collaborating with JSTOR to digitize,

preserve and extend access to Journal for Research in Mathematics Education

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
BEGLE REVISITED: TEACHER KNOWLEDGE AND
STUDENT ACHIEVEMENT IN ALGEBRA

THEODORE A. EISENBERG

Northern Michigan University

Begle (1972) investigated the extent to which a teacher's knowledge

algebra influences student performance. His results showed no educa
ally significant correlation between teacher knowledge and student perfo
ance. Moreover, Begle's findings were in accord with other studies (C
land & Doyle, 1973; Hunkler, 1968; Hurst, 1967; Keeves, 1972; No
1968; Rouse, 1967).
Teacher-training programs are built on the assumption that teacher
influence student learning. Grading procedures, curriculum committ
and certifying agencies are just a few of the safeguards built into the sys
to ensure that teachers have measured up to a set of standards. Hence, th
is a fundamental belief that teachers influence learning, but the st
above seem to indicate otherwise.
One can ask questions regarding each of the studies about the experimen-
tal design that might subsequently cast a shadow on the findings. For
example, Begle started off with 492 teachers; he ended up with 308 teachers.
All were National Science Foundation Institute participants who volun-
teered to participate in the study. Thus, Begle worked with an extremely
select and highly motivated group. Begle was aware of this and consequently
argued that the overall nonsignificant correlations he obtained between
teacher knowledge and student performance were probably attributable to
the existence of a lower bound of knowledge teachers must possess, below
which the relationship does hold. All teachers in his study were far above
that lower bound. (Fitzgerald (1973), however, in reviewing the Begle study,
did not cite this inherent selection bias as a potential threat to the negative
finding.) Similar threats can be cited for each of the other studies listed.
The Begle study represents, with the exception of the selection bias, a
carefully planned and educationally sound study. But the selection bias
cannot be ignored. Because of this, the Begle study was replicated (and
subsequently refined) on a sample chosen without selection bias. Thus, the
hypothesis was that for a "normal" audience, significant positive correla-
tions between teacher knowledge and student performance would appear.

The author wishes to thank Alan R. Osborne for the advice and work he contrib-
uted to the initial parts of this study. This project was undertaken as part of a
postdoctoral program at The Ohio State University and was supported in part by
OSU Faculty Research Grant 1200-121299-0401.

216 Journal for Research in Mathematics Education

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
 Procedures

Teacher Selection

A "normal" audience was defined to be the mathematics teachers in the

field. Each of the 26 junior high schools in Columbus, Ohio, offered at least
one section of Algebra I. All schools were contacted, and of the 52 algebra
teachers, 42 were asked to participate. The 10 teachers not asked to partici-
pate taught in schools that this investigator was not permitted to enter; their
administrators refused to have their school participate in a study that would
measure a teacher's knowledge of subject matter. Of the remaining 42
teachers, 28 participated. Five of the nonparticipating teachers were in-
eligible because they were involved in a team-teaching situation. The re-
maining 9 nonparticipating teachers simply refused.
It must be argued that the nine nonparticipating teachers were similar to
the participating ones. Four of the nonparticipating teachers felt they would
like to participate but could not give up four class sessions for the testing the
study required. Another of the nonparticipating teachers also wanted to
take part in the study, but her class refused! This teacher taught in a junior
high school located near the Ohio State University campus. Apparently, the
students wanted relief from the constant testing by researchers from OSU.
The remaining four nonparticipating teachers seemed to indicate that re-
search was a waste of time and that they had more important things to do.
Also, it was believed that had permission been granted to talk with the 10
teachers with recalcitrant administrators, at least eight would have partici-
pated in the study. Hence, it was assumed that the nonparticipating teachers
were not unlike participating ones (hereinafter called Columbus teachers).

Testing Teachers
Each of the 28 teachers in the study took the Algebra Inventory Form B
examination (Begle, 1972). This was a 50-minute multiple-choice test of 34
questions designed to measure the teacher's understanding of the real num-
ber system and other related algebraic structures. A complete listing of test
questions and test characteristics can be found in Begle (1972). The mean
scores are listed in Table 1. A z test comparing the mean scores of the Begle
teachers with the Columbus teachers was not significant (p = .05).

Table 1
Algebra Inventory Form B

Columbus teachers Begle teachers

Number of cases 28 375
Number of items 34 34
Mean total score 16.93 18.43
Standard deviation 6.39 5.83

z test (difference of two means) z

May 1977 217

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
Testing Students
Testing of the students took place during the fall of 1973 and spring of
1974. By 16 October, all students (N = 807) had taken two examinations: an
examination on logic authored by the investigator and designed to measure
the ability to handle logical syllogisms; and the Mathematics Inventory I
examination (Begle, 1972).
The spring testing was done during the last 2 weeks of the school year.
Mathematics Inventories III and IV (Begle, 1972) were given to 715 stu-
dents. (Three teachers dropped out of the study.) Mathematics Inventory III
was designed to measure the students' algebraic computational skills (AS).
Mathematics Inventory IV measured algebraic concepts (AC).
During the winter of 1973-74 other data were collected and used to factor
out differences among students in mathematics achievement and basic men-
tal ability. These data were taken from student records. Data from teacher
personnel files were also collected. (See Table 2 for a list of the variables.)

Table 2
Variables

Variable name

Teachers
1. Algebraic Inventory Form B X0
2. Number of years of experience X18
3. Number of postcalculus mathematics courses X19
4. Collegiate mathematics grade point average X20
Students
1. Logic Test (authored by investigator) X1
2. Mathematics Inventory I (Begle) X2
3. Grade earned in last English course (A = 4, B = 3, etc.) X3
4. Grade earned in last mathematics course X4
5. Short Form Test of Academic Aptitude (SFTAA-Leve
language) X5
6. STFAA (Level 3, Language) X6
7. SFTAA (Level 3, Composite (Variables X4, X5)) X7
8. California Comprehensive Test of Basic Skills (Level 4, Form Q)
(CCTBS-Reading Vocabulary) X8
9. CCTBS (Reading Comprehension) X9
10. CCTBS (Arithmetic Computation) XIO0
11. CCTBS (Arithmetic Concepts) XI1
12. CCTBS (Arithmetic Applications) X12
13. CCTBS (Language Mechanics) X13
14. CCTBS (Language Expression) X14
15. CCTBS (Language Spelling) X15
16. Mathematics Inventory III-Algebraic Skills (Begle) X16
17. Mathematics Inventory IV-Algebraic Concepts (Begle) X17

Design
Each teacher was assigned the mean score for his class on variables X1,
X2, X3, ... , X17. Hence, each teacher had the associated vector (XO, X18,
X19, X20, XI, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11,... , X17).
218 Journal for Research in Mathematics Education

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
Regression equations were determined for the variables X1, X2, . . . , X15 on
X16 and XI, X2, ... ,X15 on X . The r egression equations were then used
to predict an X16 (X16) and an X17 (X17) for eachteacher. The difference
between the predicted X16 and the observed X16 (X16 - X16) was called the
effect of the teacher with respect to algebraic skills, EAS. Similarly defined
was the effect of the teacher with respect to algebraic concepts, EAC. Appro-
priate correlations (biserial or Pearson product-moment) between EAS and
variables XO, X18, X19, X20 and EAC and XO, X18, X19, X20 were com-
puted.

Findings
The "best" eight-variable model (using the maximum R2 improvement
procedure) for predicting X16 from X, X2, ... , X15 is listed in Table 3,
along with the regression coefficients and the amount of variance in X16
attributable to these variables. An eight-variable model was arbitrarily
chosen in order to account for a good proportion of the variance in X16.

Table 3
"Best" Variable Regression Models for Predicting Algebraic Skills (X16)

Number in
model R2 Variables in model

1 .556 X14
2 .720 n', 3TR
3 .743 X3, X13IX14
4 .754 2, X X X13 . X14
5 .774 X- X8 X13 XL4
6 .780 ,2,X : ," ,-TX.-14
7 .824 73, X6, X7, X8,T 1 XT
8 .836 n3T, T,7, I, kiX ? f, X1l1, X14
Source B values t p Std err B Std B values
meap -24.72
X14 1.66 3.88 0.00 0.43 1.18
,n -9.29 -4.17 0.00 2.23 -0.51
X8 -1.32 -2.19 0.04 0.60 -0.98
X-6 2.72 3.05 0.01 0.89 2.01
X7 -2.33 -2.59 0.02 0.90 -1.94
X7 0.50 1.08 0.30 0.48 0.41
XT1 -0.86 - 1.90 0.07 0.45 -0.70
XTT 1.03 1.69 0.11 0.61 0.94

The "best" eight-variable mode

X15, along with appropriate stat
The eight-variable models were
correlation matrix for EAS, EAC, X

Discussion

If one had to choose, for example, the "best" four-variable model from
Table 3 for predicting student competency with algebraic skills, one would

May 1977 219

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
 Table 4
"Best" Variab

Number in
model R2 Variables in model

1 .694 1
2 .747 F13, _.
3 .775 If "XO X13
4 .835 - , X5, W, xi1
5 .863 , T5r, II, I
6 .8865, M, 17, -- 1-5
7 .899 14, TX' X"MTT 7T" XT' "J
8 .912 " , 1 , , , , X"',
Source B values t p Std err B Std B values
mean -9.69
Xt3 0.88 4.91 0.00 0.18 1.00
X-T 0.73 3.70 0.00 0.20 1.10
-0.26 -1.48 0.15 0.17 -0.38
0.67 3.76 0.00 0.18 0.73
-0.25 -2.09 0.05 0.12 -0.29
XTT -0.44 -1.97 0.06 0.22 -0.75
MfT -0.25 2.04 0.06 0.12 -0.43
XT -2.14 -1.96 0.07 1.09 -0.22

be forced to choose variables X2

for more than 75% of the varia
X3, X13, and X14 do not measu
mathematical attributes. These
pression, yet they had more influ
arithmetic concepts. A similar s
model for predicting a score on
cepts). It is therefore suggested
language skills and mathematics
The only significant correlation
Table 5) was between the teacher
examination and the teacher's n
expect this, but it is surprising t
not mastery as measured by gr

Table 5
Correlation Matrix for EAS, EAC, XO, X18, X19, X20a

N = 25 EAC EAS X0 X18 X19 X20

EAC 1.00 .05 -.18 -.20 -.25 -.22

EAS 1.00 -.02 -.12 .18 -.24
X0 1.00 -.18 .57* .15
X18 1.00 -.23 -.33
X19 1.00 .40
X20 1.00

" Numbers in
*p < .005.

220 Jou

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
college mathematics grade-point averages correlate meaningfully with their
knowledge of the real number system and other related structures? It is
frightening to think that exposure to material, rather than mastery of
material, is of prime issue.
The lack of a meaningful correlation between the teacher effects (EAS and
EAC) and all other variables was disheartening. Perhaps Begle's inter-
pretation is correct. There is a lower bound of knowledge, below which the
relationship between teacher knowledge and student performance does
hold. But this lower bound is so low that it is not worth worrying about, at
least with respect to algebra teachers and their knowledge of algebra. With
"exposure" to courses, one can pick up an adequate amount of knowledge
to be a "successful" teacher. Whether or not this holds for other grade levels
is a subject for further study. But assuming that it does, and there is no
reason to assume otherwise, then we, as teacher educators, should reexam-
ine the role of grading and certifying agencies that currently guard the
entrance to our profession.
An unexpected finding was made when comparing student scores of
Columbus teachers with student scores of Begle teachers. The results are
summarized in Table 6.

Table 6
Students' Scores on Mathematics Inventories I, III, IV

Inventory Inventory Inventory

I III IV

Students of Begle teachers na = 625 na = 625 na = 625

TX= 17.70 X= 19.21 X=15.22
SD = 5.32 SD = 8.01 SD = 4.96

Students of Columbus nb = 807 ne = 361 nd = 265

teachers X = 17.63 X = 17.23 T = 14.04
SD = 5.49 SD = 7.26 SD = 4.85

z test (difference
of two means)
Inventory I z = 1.31
Iventory II z = 3.97 (p < .001)
Inventory III z = 3.31 (p < .001)
" This n represents the scores of at least 2 students randomly chosen from each classroom of each teacher in
b This n represents all students who took Mathematics Inventory I.
"c Of the 25 teachers who completed the study, 13 were randomly selected. This n represents the scores of all studen
classrooms.
d Of the 25 teachers who completed the study, 10 were randomly selected. This n represents the scores of all students in
their classrooms.

These data indicate that students of Begle teachers were ahead of Colum-
bus teachers in the fall and that this difference was accentuated by the
spring. The main question is, "Why?" Why such a hiatus between the two
groups of students? What happened in the intervening months to cause such
discrepancies in performance? Could it be that the tests were not equally

May 1977 221

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms
valid achievement measures for the two groups? Or could it be that the
training received in the institutes really helped the teachers do a better job of
teaching? Or maybe only better, more highly motivated teachers applied to
the institutes or were selected to participate in them. If we adhere to the
premise that teachers do effect learning, then it is obvious from Table 6 that
something went on in the classrooms of Begle teachers that was absent from
the classrooms of Columbus teachers. What caused such drastic differences?

Conclusion

The conclusion was that teacher knowledge of subject matter had little
effect on student performance. There is a growing body of evidence that
comes to the same conclusion. An unexpected finding was that students of
teachers who had attended NSF Institutes in 1970 had more initial knowl-
edge of mathematics than students of noninstitute teachers. By the end of
the school year, these students were significantly superior to their "or-
dinary" counterparts in algebraic skills and concepts. Why the hiatus?

REFERENCES

Begle, E. G. Teacher knowledge and student achievement in algebra. SMSG Reports, No. 9.
Stanford: School Mathematics Study Group, 1972.
Copeland, W. D., & Doyle, W. Laboratory skill training and student teacher classroom
performance. Journal of Experimental Education, 1973, 42, 16.
Fitzgerald, W. M. Review of Teacher knowledge and student achievement in algebra, SMSG
Reports, No. 9. Investigations in Mathematics Education, 1973, 6, 13-15.
Hunkler, R. F. Achievement of sixth-grade pupils in modern mathematics as related to their
teacher's mathematics preparation (Doctoral dissertation, Texas A & M University, 1968).
Dissertation Abstracts, 1968, 29, 3897A.
Hurst, D. The relationship between certain teacher-related variables and student achievement
in third grade arithmetic (Doctoral dissertation, Oklahoma State University, 1967). Dis-
sertation Abstracts, 1968, 28, 4935A.
Keeves, J. P. A multivariate study of the contributions of the home, the school and the peer group
to educational environment and student achievement. Stockholm: Almqvist and Wiksell, 1972.
Norris, F. R. Pupil achievement as a function of an inservice training program on mathematics
concepts for sixth grade teachers (Doctoral dissertation, George Peabody College for Teach-
ers, 1968). Dissertation Abstracts International, 1969, 30, 1054A.
Rouse, W. M., Jr. A study of the correlation between the academic preparation of teachers of
mathematics and the mathematics achievement of their students in kindergarten through
grade eight (Doctoral dissertation, Michigan State University, 1967). Dissertation Abstracts,
1968, 28, 4031 A.
[Received October 1974; revised November 1976]

222 Journal for Research in Mathematics Education

This content downloaded from 177.228.81.90 on Wed, 08 Nov 2017 04:57:31 UTC
All use subject to http://about.jstor.org/terms