ERIC ED422925: Beyond Course Availability: An Investigation into Order and Concurrency Effects of Undergraduate Programming Courses on Learning.

8 Pages·1997·0.17 MB·English

by  ERIC

DOCUMENT RESUME
IR 057 083
ED 422 925
Urbaczewski, Andrew; Urbaczewski, Lise
AUTHOR
Beyond Course Availability: An Investigation into Order and
TITLE
Concurrency Effects of Undergraduate Programming Courses on
Learning.
1997-00-00
PUB DATE
7p.; In: Proceedings of the International Academy for
NOTE
Information Management Annual Conference (12th, Atlanta, GA,
December 12-14, 1997); see IR 057 067.
Research (143) -- Speeches/Meeting Papers (150)
Reports
PUB TYPE
MF01/PC01 Plus Postage.
EDRS PRICE
Academic Achievement; Higher Education; Information Science
DESCRIPTORS
Education; Instructional Effectiveness; Introductory
Courses; *Programming; *Programming Languages; Student
Attitudes; Student Surveys; Teaching Methods
ABSTRACT
The objective of this study was to find the answers to two
primary research questions: "Do students learn programming languages better
when they are offered in a particular order, such as 4th generation languages
before 3rd generation languages?"; and "Do students learn programming
languages better when they are taken in separate semesters as opposed to
simultaneously?" Students from nine introductory programming classes over two
semesters at a large Midwestern university were used as subjects for this
experiment; 275 students responded to a survey at the end of the semester.
Subject responses were divided into three groups, depending on the class
being rated. These classes were: introduction to Visual Programming,
introduction to COBOL programming, and introduction to C programming. To test
hypotheses, linear regression was used, running the data against two separate
dependent variables, grade and comfort factor. Mixed results were found for
the different types of classes.
(AEF)
********************************************************************************
Reproductions supplied by EDRS are the best that can be made
from the original document.
********************************************************************************
BEYOND COURSE AVAILABILITY:
AN INVESTIGATION INTO ORDER AND
CONCURRENCY EFFECTS OF UNDERGRADUATE
PROGRAMMING COURSES ON LEARNING
U.S. DEPARTMENT OF EDUCATION
Office of Educational Research
"PERMISSION TO REPRODUCE THIS
and Improvement
Andrew Urbaczewski
EDUCATIONAL RESOURCES
fNFORMATION
MATERIAL HAS BEEN GRANTED BY
CENTER (ERIC)
Indiana University
D1 This document has been
T. Case
reproduced as
received from the person
or organization
originating it.
O Minor changes have been
made to
improve reproduction quality.
Lise Urbaczewski
Indiana University
Points of view or opinions stated
in this
TO THE EDUCATIONAL RESOURCES
document do not necessarily
represent
INFORMATION CENTER (ERIC)."
official OERI position or policy.
claims for first learning the object oriented
INTRODUCTION
language (Currid 1992) and for first learning the
3GL (Powell 1997).
The process of undergraduate MIS education is
As educators, we find
constantly evolving.
Moreover, students often find themselves in a
ourselves in a constant state of curriculum
redesign, often to meet the demands of the
crunch to get registered for the required courses.
With the undergraduate MIS major's increasing
various recruiters that visit our campuses each
popularity, it is getting harder for the students to
Fall looking for new talent in the expanding field.
It is not
arrange their schedules' optimally
are
these
technologies
Quite
new
often,
uncommon to find students taking two or three
programming languages. It is common today to
different programming languages during the
hear recruiters ask for students skilled in
same semester. The authors find it difficult to
C++,Visual Basic, Powerbuilder, or Java before
believe that a student can perform optimally
In accordance with
entering the job market.
under these conditions, as this would be like a
Association for Computing Machinery (ACM)
student trying to learn two spoken foreign
curriculum design guidelines (ACM 1991), we
languages, like French and German, during the
into
our
these
technologies
incorporate
same semester.
curriculums, keeping the content as current as
However, in the rush to design our
possible.
Given the apparent mentioned conflicts, the
curriculums to give students maximum exposure
objective of our study was to find the answers to
to required technologies, perhaps we overlooked
2 primary research questions:
factors to maximize learning efficiency in our
students.
1) Do students learn languages better when
they are offered in a particular order, such as
A debate exists today over the proper method of
4th Generation Languages (4GL) before
Research has been
programmer instruction.
conducted in the past to find the optimal
3GL's or vice versa or is there no effect?
sequence for offering programming instruction.
2) Do students learn programming languages
Veteran programmers learned the older second
better when they are taken in separate
and third generation languages (2GLs and 3GLs)
before they learned 4GLs because no 4GLs
semesters as opposed to simultaneously?
Students now are often afforded the
existed.
In an attempt to find the answers to these
opportunity to learn a 4GL or object-oriented
questions, we decided to ask the students
programming language without ever learning a
In the classroom, programming
themselves.
3GL. Empirical research in the past has found
instructors hear a variety of complaints from
mixed results (Manns and Carlson 1992, Rosson
students regarding their difficulty or inability to
and Alpert 1990), while authors have made
131
Proceedings of the 12th Annual Conference of the International Academy for Information Management
2
learn the language at hand. We divide these
H2b.
Students who have higher grade point
comments in two groups. The first group is from
perform better
averages
will
in
a
students who have learned a prior language and
prograniming course than those who have
are having difficulty learning a second language.
lower grade point averages.
Often they are
attempting to
learn two
programming languages at the same fune. This
H2c.
Students who have progressed further in
is the group we call "But there's an easier
will have higher grades
way
school
in
"
The other group are students that have never had
programming courses than those who
any programming courses before. They tend to
have progressed less in school.
report in the classroom that they feel inferior to
other students in the class who may know more
about computers or be more experienced with
We would still like to get another dependent
They are worried that their
programming.
measure of mastery of the skill. While objective
objective performance in the class will suffer
measures are important in science, a subjective
because they are being compared to this other
measure of mastery may also be important.
group of students.
This group we call "It's not
Students may make a high grade in a course but
fair
not have any mastery of the material, and vice
"
versa. Therefore we propose level of comfort as a
Based on experiences noted in the classroom,
dependent measure for student mastery of
we
formulate the following hypotheses:
programming material.
H1a.
Students that have a prior experience
H3a.
Students that have a prior experience
with a programming language will have
with a programming language will feel
higher grades in an introductory course
more comfortable with that language
in that language than those who have no
than those who have no prior experience
prior experience with that programming
with that programming language.
language.
that
H3b.
Students
have
taken
any
Hlb.
that
Students
have
taken
any
programming class
will
feel
more
programming class will have better
comfortable with the language than those
grades in a course than those who have
who have not taken a programming class.
not taken a programming class.
H3c.
Students who are taking more than one
Hie.
Students who are taking more than one
programming language simultaneously
programming language simultaneously
less comfortable with the
will
feel
will have worse grades than those who
languages than those who have taken the
have taken the same programming
same programming languages in any
in any non-simultaneous
languages
non-simultaneous order.
order.
METHODOLOGY
We use grades in H1 because they are intended to
be an objective measure
of performance.
Students from nine introductory programming
However, there are often other factors than
classes over two semesters at a large Midwestern
mastery of the skill which figure into the
university were used as subjects
this
for
awarding of grades.
These often include
experiment. 275 students responded to a survey
attendance, year of progression through school,
at the end of the semester.
This was done
and some
factor
general
of
intelligence.
without compensation to the students, requiring
Therefore, we also propose:
a few minutes of their time at the beginning of
the class.
One response was determined to be
H2a.
Students who attend class more regularly
unusable and it was discarded.
Subjects were
will have higher grades in programming
also asked that if they had completed this survey
courses than those who do not attend
in another class to indicate this at the top of the
class regularly.
page, and 19 surveys were eliminated through
this method.
132
Proceedings of the 12th Annual Conference of the International Academy for Information
Management
3
Dependent Variable = Grade
The survey asked students to respond to several
1.
A. Class Rated = Introduction to Visual
items concerning their academic performance and
Programming using Visual Basic
relative comfort with the language.
Students
responded anonymously to remove threats to
internal validity (Campbell and Stanley 1963).
Unstandardized Standardized
These items covered objective performance levels,
Sig.
Coefficients
t
Coefficients
such as grading, and subjective performance,
Beta
S.E.
B
including comfort with the language and desire to
1.043 .299
.560
.537
(Constant)
learn more about the language. Subjects reported
-.856
.393
-.062
-.359
.420
W-COBOL
not only on the class they were currently
.140
1.483
.115
.176
.119
W-C
finishing but also on all other programming
.750 .454
.069
.126
<.001
H-COBOL
classes they had taken for a grade at the
.429
.793
.075
<.001
.121
H-C
This was done to
collegiate level or above.
3.053 .003
.223
.050
Attendance
.151
data about
their
for
classes
all
capture
2.822
.005
.213
.035
PhorExp
<.001
determination of ordering effects.
3.242
.234
.123
.001
.397
GPA
.539
.615
.048
.065
<.001
Y rl nColl
Subject responses were divided into three groups,
depending on the class being rated. These classes
were introduction to Visual Programming,
B. Class Rated = Introduction to COBOL
introduction to COBOL programming, and
programming
This of course
introduction to C programming
meant that the same subject may have been in
Unstandardized Standardized
each of the three pools if she or he had taken all
t Sig.
Coefficients
Coefficients
These groups
of the courses being examined.
Beta
S.E.
B
were analyzed individually and then compared to
1.315 .195
1.006
1.322
(Constant)
the others for significance. Linear regression was
.675
.050
.464
W-VB
.195
.421
used with the same covariates against two
-.746
.459
-.092
-.164
.220
W-C
different dependent variables, grade in the course
-.750
.457
-.099
.374
-.281
H-VB
and comfort level with the language. Reported
-.677
-.095
.501
-.159 .235
H-C
prior experience was also captured as a Likert
-.525
-.066
.602
.098
Attendance
-<.001
scale and used as a covariate.
.468
-.092
-.731
.089
PriorExp
-<.001
4.876
.000
.605
.243
1.183
GPA
RESULTS AND INTERPRETATIONS
-1.890
-.246
.065
-.315 .167
YrInColl
To test our hypotheses, we used linear regression,
Class Rated = Introduction to
running the data against two separate dependent
C.
C programming
As noted
variables, grade and comfort factor.
above, grade was the self reported letter grade for
the student for prior classes and expected grade
Unstandardized Standardized
This was then converted
for the current class.
Coefficients
Sig.
Coefficients
t
into the numeric equivalents at that university.
Beta
S.E.
B
The data set was also divided into three
1.229
.447
.549
.221
(Constant)
partitions, depending on the class being rated.
2.635
.009
.170
.105
.276
W-VB
Since attendance was expected to have only a
-1.834
.068
-.297
-.117
.162
W-COBOL
significant influence on grade rather than
.458
.046
.743
.128
<.001
H-VB
comfort level, they were only suggested as
.845
.013
.196
.103
H-COBOL
<.001
hypotheses when grade was the dependent
.086
1.724
.045
.112
Attendance
<.001
variable. We inserted the variables for H2 into
.004
.024 .185
<.001
2.901
PhorExp
those regressions anyway for the reader's benefit.
7.356
.000
.484
.753 .102
GPA
-.334
-.023
.739
.056
-<.001
Y rl nCol I
133
Proceedings of the 12th Annual Conference of the International Academy for Information Management
It is interesting that we find different results for
PriorExp
.158
.067
.178
2.350
.020
the different types of classes. This would support
GPA
<.001
.237
.016
.220
.826
the mixed results that we have seen before. For
YrInColl
.142
.125
.089
.255
1.142
the visual program.ming courses, we find grades
are in no way related to any other programming
language work.
Attendance, Prior experience,
B.
Class Rated = Introduction to
and GPA are the only significant factors affecting
COBOL programming
grade, supporting H1 a, H2a, and H2b. We fail to
reject the null hypothesis for Hib, Hic, and H2c.
Unstandardized Standardized
Coefficients
t Sig.
Coefficients
With the course in COBOL, however,
we get
S.E.
B
Beta
different results. The only significant predictor of
(Constant)
4.572
2.074
2.204
.032
grade is GPA, supporting H2b. It is interesting
W-VB
.479
.798
.081
.601
.551
that Hia is not supported, perhaps giving
W-C
.234 .458
.073
.510
.612
credibility to the notion that grades and mastery
H-VB
-.281
.779
-.055
-.360
.720
are not perfectly related.
Perhaps even more
H-C
.632
.218
1.369
.461
.177
significant is that year in school is marginally
Attendance
.245 .200
.175
1.225
.226
significant (p=.065), in the opposite direction,
PriorExp
.186
-<.001
-.077
.939
-.011
suggesting that students who are further along in
GPA
.498
-<.001
-.015
-.103
.919
school will perform worse in COBOL. This
may
YrInColl
-.346
.347
-.150
-.998
.323
be evidence for the term "senioritis", suggesting
that students may slack off on their studies
as
they get closer to graduation.
Class Rated = Introduction to
C.
C programming
Finally, the Introduction to C programming
Hla is
course gives even different results.
Unstandardized Standardized
supported, confirming that those who have prior
Coefficients
Coefficients
Sig.
t
experience will earn a better grade. Taking the
B
Beta
S.E.
course with the visual programming course
(Constant)
.560
.537
1.043
.299
improves their grade in C, completely opposite of
W-VB
.674
.194
3.476
.231
.001
Mc. However, taking the course with COBOL is
W-Cobol
-.725
.325
-.148
-2.232
.027
marginally significant (p=.068), supporting Hic.
H-VB
.686
.160
2.442
.281
.015
This is especially interesting given that C is
more
H-Cobol
<.001
.017
.807
.221
.244
closely related to COBOL than it is related to
Attendance
.097
<.001
.028
.406
.685
Visual Basic. H2b (GPA) is also supported, and
PriorExp
.183
.053
.230
3.449
.001
there is mild support (p=.086) for attendance
GPA
.299
.212
.096
1.408
.161
improving grades.
YrInColl
-.103
-.065
-.927
.112
.355
Dependent Variable = Comfort Level
2.
Again in examining the Comfort variable,
we get
Class Rated = Introduction to Visual
A.
mixed results.
These results also tend to be
Programming using Visual Basic
different from those using grade as a dependent
variable, lending more credibility to the notion of
grades and mastery being mildly correlated. In
Unstandardized Standardized
the Visual Programming class, H3a is supported,
Coefficients
Coefficients
Sig.
t
S.E.
which may suggest that one semester of a visual
B
Beta
programming course is not enough to bring
3.223
(Constant)
1.038
3.106
.002
novices on a par with more experienced
W-Cobol
.664
.983
.107
users.
1.481
.140
W-C
We were again surprised to find that H3c
.536
.223
.188
2.407
.017
was
supported in the opposite direction than
H-Cobol
.244
.387
.146
1.586
.115
we
suspected, as increased comfort level
H-C
.169
.234
.069
.472 was
.721
Attendance
reported from those who took the two classes
.096
<.001
.052
.705
.482
together.
134
Proceedings of the 12 h Annual Conference of the International Academy for
Information Management
5
Thus we can suggest that students
the COBOL course, we find
nothing
reverse.
In
should take visual programming first, and then a
significant. None of the variables we captured
course in COBOL.
were predictors of a person's comfort level with
COBOL.
FIGURE 1
The lack of support for many variables in the
COBOL course in both situations may be due to a
lack of power. There were far less students who
had taken the COBOL course (56) than the C
course (191) or the Visual Programming course
(173).
Finally, in the C course, we have a third set of
In line with the Visual Programming
results.
course, students reported a higher comfort level if
they were taking the course concurrently with
visual programming, providing opposite support
More research should be done to try to discover
for H3c. The more expected result was found
the reasons for the contrasting results.
For
from students who were taking C with the
example, it is not clear to the authors why
COBOL class, supporting H3c. H3b was also
students seem to perform better when taking C
supported with the Visual Programming course,
concurrently with visual programming but worse
but it was not with the COBOL course. This was
when taking it with COBOL programming. We
the only instance where either Hlb or H3b was
cannot explain these effects and would like to try
significant. Prior experience was also significant,
Exploration into
to investigate them further.
supporting H3a.
student experiences with taking concurrent
programming languages would also help, at least
CONCLUSIONS AND FUTURE
on a exploratory level. Perhaps comparing them
RESEARCH DIRECTIONS
to students who have attempted to learn two
foreign languages simultaneously would help
As we noted in our results section, the findings
discover learning and memorization patterns.
we had with the data were very mixed. No single
hypothesis was supported across all classes when
Moreover, we are unsure how this research
Comfort was the dependent variable, and only
It would be
applies outside university settings.
GPA was supported consistently as a predictor of
interesting to survey professional programmers
Grade. These mixed results are not necessarily
Professional training
in the same manner.
bad, however. They help us in curriculum design,
courses could be used to find programmers of
which was the goal of this paper.
different ability and style.
The Introduction to COBOL course was the one
which had the least predictors for either
Overall, this research is of interest to researchers
in one of our forms of patronage to society,
dependent variable. GPA was the only predictor
educating students. We owe it our students (and
of Grade, and there were no predictors of comfort
often our taxpayers) to help them as much as
This can be interpreted to mean that
level.
Finding the
possible in the education process.
nothing gives anyone an undue advantage in the
optimal way to deliver that education is part of
COBOL course. Beginners are just as likely to do
our responsibility.
well in the course as are experts. Thus we can
suggest Introduction to COBOL as the first
PARTIAL REFERENCES
programming
the
in
programming
course
sequence.
ACM Curricula Recommendations Volume II:
Information Systems, Association for Computing
Secondly, we noticed that students seemed to
Machinery, 1991.
perform better in the C course when they had
already taken or were currently taking the visual
This did not hold true in
programming course.
135
Proceedings of the 12th Annual Conference of the International Academy for Information Management
6
Currid, C. "IS Curriculum Continues to Evolve at
Powell, A. "Programming Course Sequence and
the Corporate Level", Infoworld, June 8, 1992,
p.
Prior Knowledge of Programming Languages: Do
S64.
they Affect Students' Grades?" Proceedings of the
MS Americas
Information
Conference
on
Manns,
and
Carlson,
M.
"Retraining
D.
Systems, 1997.
Procedural Programmers: A Case Against
"Unlearning", Object Oriented Systems and
Rosson, M. and S. Alpert. "The Cognitive
Language Applications Conference, 1992,
pp. 131-
Consequences
of
Object-Oriented
Design",
133.
Human Computer Interaction, 1990, pp. 360-378.
Webster's New Collegiate Dictionary, 1993.
136
Proceedings of the 12' Annual Conference of the International Academy for
Information Management
U.S. DEPARTMENT OF EDUCATION
ERIC
Office of Educational Research and improvement (QERI)
Educational Resources Information Center (ERIC)
NOTICE
REPRODUCTION BASIS
This document is covered by a signed "Reproduction Release
(Blanket)" form (on file within the ERIC system), encompassing all
or classes of documents from its source organization and, therefore,
does not require a "Specific Document" Release form.
This document is Federally-funded, or carries its own permission to
reproduce, or is otherwise in the public domain and, therefore, may
be .reproduced by ERIC without a signed Reprod.ucdon Release
form (either "Specific Document" or "Blanket")..