Sie sind auf Seite 1von 6

Session S1E

Effects of Peer Tutoring and Attitude on Academic


Performance of First Year Introductory Programming
Students
Lisa Facey-Shaw and Paul Golding

School of Computing & Information Technology, University of Technology, Jamaica


237 Old Hope Road, Kingston 6, Jamaica

Abstract - Previous research has estimated failure rates in and Information Technology (SCIT) at the University of
introductory programming courses to be as high as 50% in Technology, Jamaica (UTech), Introduction to Programming
universities around the world. These high incidences of (ITP) is the foundation introductory programming course
failure have also led to negative attitudes towards taught in the first semester of the first year of the programme.
programming. This paper takes a look at the Introduction It is important to note that the ITP course is language-
to Programming course offered at the University of independent and traditionally does not require the use of
Technology, Jamaica. Pre-and post test programming computers to test algorithms developed.
scores were used to measure academic performance, while Records show that failure rates for ITP for the past four
a survey derived from a modified Fennema-Sherman years range from approximately 30% to 48%, that is, those
mathematics attitudes scale was used to measure student BSCIT students who fail ITP on their first attempt. Various
attitude towards programming. Three sub-categories were strategies have been undertaken to come up with answers as to
used for the measurement of attitudes, namely, personal why the situation continues to exist and how best it can be
confidence in learning programming, usefulness of resolved. These include an action research project to identify
programming and the perception of teacher motivation. factors contributing to poor student performance in an
Results indicate little evidence of a positive effect of peer introductory C Programming course [2], reviews of aptitude
tutoring on the academic performance of students. It also tests, the results of which are used as a part of the selection
reveals that attitude, and in particular, confidence in criteria for students entering the programme, as well as other
learning programming, plays an important role in strategies. Also, a study is being undertaken to examine the
students’ academic performance. This paper discusses the admission criteria for students entering the BSCIT programme
design of the study and analyzes the results gathered. in SCIT [3].
The knowledge to be generated by the study will interest
Index Terms - attitude to programming, introductory programming instructors, not just from UTech but from other
programming, peer tutoring. universities having similar experiences of high failure rates.
The findings of the study will be quite useful in determining
INTRODUCTION how educational practices in the area of introductory
Introductory programming is a foundation first year course in programming can be improved and whether peer tutoring
many Computer Science degree programmes. Its role is to should be considered as an effective instructional method for
teach general principles that will enable the student to develop this course.
logical, reasoning and problem-solving skills. These skills are LITERATURE REVIEW
difficult to teach and consequently, many students find it
difficult to grasp the general concepts of programming, often The instructional method used to deliver an introductory
resulting in high dropout or failure rates. Universities around programming course impacts significantly on success in the
the world have described undesirable failure rates in course. The use of conceptual methods of instruction to foster
introductory programming courses, with ‘reported withdrawal, knowledge of the computer and the programming language is
failure, and D-grade rates approaching 50 percent’ [1]. This required for learning programming skills and it is believed that
research examines the effects of peer tutoring, as well as novices differ from experts in their degree of understanding of
attitude to programming, on the academic performance of first the concepts underlying programming [4]. Conventional or
year students taking the introductory programming course. traditional teaching involves a teacher-centred strategy in
which the teacher performs most of the activities, presenting
BACKGROUND structured knowledge to the student and playing a significant
In the Bachelor of Science in Computing and Information role in the delivery of the instruction. Tutorials are used to
Technology (BSCIT) programme in the School of Computing help the learner use knowledge and skills, learn interpersonal

0-7803-9077-6/05/$20.00 © 2005 IEEE October 19 – 22, 2005, Indianapolis, IN


35th ASEE/IEEE Frontiers in Education Conference
S1E-1
Session S1E
skills and improve self-understanding [5]. Cognitive theories programming concepts. It required them to analyze algorithms
of learning attempt to explain how instructors can use and answer questions on their execution as well as to evaluate
information to improve student learning. For complex topics problem-based mathematical expressions. The objective of the
such as programming, the instructional approach should survey was to measure the attitudes of students towards the
emphasize student responsibility for their own learning and introductory programming course. It was carried out using an
utilize conceptual models so that students can better apply instrument derived from a modified Fennema-Sherman
their learning to creatively solve problems [6]. Constructivist mathematics attitudes scale, a scale used extensively in other
theories also promote the concept of student interactions with studies [10, 12]. This instrument was modified to reflect
advanced peers, suggesting that this is a highly effective introductory programming, rather than mathematics and
method of teaching and learning. consisted of three rather than ten subscales measuring self-
Peer tutoring allows students to reap special learning confidence in learning programming, usefulness of
benefits from reviewing and restructuring knowledge as they programming, and perception of teacher motivation. Each
present it to other students. It has been operating successfully scale consisted of twelve statements, six positively worded
in universities in the United Kingdom and ‘has the potential to and six negatively worded. These items were rated using a
facilitate learning in ways unlikely to be found in conventional Likert scale with responses ranging from 1 to 5 of ‘Strongly
teaching’ [7]. Many peer tutoring strategies take the form of Agree’, ‘Agree’, ‘Neither Agree nor Disagree’, ‘Disagree’,
more advanced students helping less advanced students with and ‘Strongly Disagree’, respectively. Each item was given a
their studies. In the United Kingdom this form of peer tutoring score from 1 to 5, with 5 indicating a more positive attitude.
is referred to as 'proctoring'. Reference [8] defines peer For each scale, the scores range from a minimum of 12 to a
tutoring as the ‘structuring of the environment so that pairs of maximum of 60.
same-age students team up to practice basic skills’. While The instrument also captured students’ demographic
much empirical research on peer tutoring strategies has been items regarding age, gender, highest educational level
done at the elementary and secondary levels, little has been attained, highest mathematical level attained and prior
done at the university level [9]. The use of pair programming, programming experience.
a style of programming in which two programmers work The sample, purposively selected, consisted of two
together at a computer, is becoming popular as a collaborative BSCIT first year classes of students, each consisting of
technique in programming courses. It is believed that this approximately fifteen to twenty students. The study was
technique leads to improvements in the ability of students to conducted during the normal tutorial time for the ITP course –
design programs, their overall performance in programming a two-hour period, held once per week. Both tutorial periods
and their attitude towards programming [10]. were held in the afternoon. One group was randomly
designated to be the experimental group. In the experimental
METHODOLOGY group, students were randomly assigned in pairs by the
Design instructor to avoid close friends and strong programmers
working together.
A quantitative approach utilizing a pretest-posttest control The pre-test was administered during the beginning weeks
group quasi-experimental design was adopted for the research. of the first semester in the academic year 2004-2005. It
This design was chosen due to the existence of naturally consisted of general multiple-choice based questions
assembled classes for which random assignments to the measuring the students’ ability to evaluate mathematically-
experimental and control groups could not be applied [11]. based expressions, interpret problem-based questions and
Attempts were made however to match the characteristics of determine results after the execution of an algorithm. Results
the experimental and control groups as closely as possible to from the pre-test were not included in the final grades
minimize differences that existed between them. This method obtained by the students in the course. The post-test,
allowed the researcher to measure the impact of the treatment administered at the end of the semester, was of a more
on the academic performance of the experimental group by practical nature and measured the students’ ability to create
assessing the degree of change in academic performance, tools such as pseudocode, flowcharts and decision tables to
when compared to the academic performance of the control solve programming problems based on problem-oriented
group, who will be receiving conventional instruction. The questions. Participants were informed that participation in the
study was carried out over a period of one semester study was voluntary and that their test scores and results from
(approximately 14 weeks during the period September - the attitude survey would be kept confidential.
December). To minimize the effects of any differences in
teaching style, the same instructor taught both the control and Group Composition
experimental groups. The gender composition for the sample was predominantly
Surveys male - approximately 75%. This also reflects the gender
makeup of SCIT. The age of almost 90% of the sample was 19
Students were given a survey constituting a brief or younger and almost half of the group had had no form of
programming pre-test which measured their knowledge of programming experience. The educational levels of both

0-7803-9077-6/05/$20.00 © 2005 IEEE October 19 – 22, 2005, Indianapolis, IN


35th ASEE/IEEE Frontiers in Education Conference
S1E-2
Session S1E
groups were also found to be similar. Forty-six percent had TABLE I
MEAN PRE-TEST AND POST-TEST SCORES BY GROUP
attained education at the General Cambridge Examinations
Type of Group Pre-Test Score Post-Test Score
(GCE) Advanced Level or the Caribbean Advanced
Proficiency Examinations (CAPE) Level. Both groups were Control Group Mean 43.13 40.89
more or less equal in terms of the number of students who had Std. 14.48 17.83
Deviation
attained qualifications at the GCE/CAPE Advanced Level.
Experimental Mean 37.14 42.86
The minimum matriculation requirements for entering the Group
BSCIT programme is achievement at the GCE Ordinary Level Std. 16.37 24.16
or the Caribbean Examination Council (CXC) Ordinary Level. Deviation
The control group had approximately three times as many Total Mean 40.33 41.73
students attaining qualifications up to the CXC/GCE Ordinary Std. 15.42 20.42
Level than the experimental group. Thirty percent had also Deviation
attained up to an Advanced Level Proficiency in Mathematics.
Tasks A one-way ANOVA (Table II) done for the pre-test
scores indicates that there was enough similarities between the
Similar learning materials and resources were adopted for both groups at the beginning of the instruction (F-value = 1.129).
groups. Group work included problem-oriented questions The p value, however, of 0.297 suggests it is not significant,
based on the lecture previously held on that topic. In the quite likely due to the small sample size.
experimental group, students were encouraged to work
together with their peer and to try to answer the questions their TABLE II
peer had, however if they felt they didn’t understand a topic ANOVA FOR PRE-TEST SCORES
they were still allowed to consult with the tutor, who would Sum of df Mean F Sig.
work with both students. For the control group, students Squares Square
Between 267.202381 1 267.202381 1.129 0.297
generally worked alone on problems. After being given the Groups
opportunity to work on the problem for some time, students Within Groups 6629.464286 28 236.766582
would then present the solution on the board. In most Total 6896.666667 29
instances, both the control and experimental groups were
assigned the same problems, but problems would occasionally
differ. All tests were taken individually. The one-way ANOVA done for the post-test scores
(Table III) also indicates that there was no significant
FINDINGS difference between the results of the control and experimental
groups (F-value=.072). The p-value of .790 again suggests
Although the sample size was relatively small, (19 persons in that the result may be due to the small sample size.
the control group and 14 in the experimental group), there
were some interesting results obtained from the data collected, TABLE III
which could form the basis for further study in this area with a ANOVA FOR POST-TEST SCORES
larger sample size. Due to the nature of the research design, Sum of df Mean F Sig.
the results were interpreted cautiously [11]. Squares Square
Between 31.04169515 1 31.0416951 0.072 0.79
Peer Tutoring and Academic Performance Groups
Within Groups 13309.50376 31 429.338831
Table I contains the pre-test and post-test score means and Total 13340.54545 32
standard deviations for the experimental and control groups. It
reveals that the mean pre-test scores were higher for the
control group than for the experimental group, however the A Mann Whitney U test was used to further examine the
mean post-test scores were higher for the experimental group differences between the results of the control and experimental
than for the control group. Of note also is the fact that for the group in terms of their pre- and post-test scores (Table IV). In
control group there was a decline in the mean score from the terms of the pre-test scores, the control group (Mean Rank =
pre-test to the post-test, while the experimental group showed 17.53) was found to score higher than the experimental group
an increase in mean scores from pre-test to post-test. (Mean Rank = 13.18), however this difference was not
significant (U=79.5, p>.05). No significant difference (U =
133, p = 1) was found between the control group (Mean Rank
= 17.00) and the experimental group (Mean Rank = 17.00) in
terms of the post-test scores.

0-7803-9077-6/05/$20.00 © 2005 IEEE October 19 – 22, 2005, Indianapolis, IN


35th ASEE/IEEE Frontiers in Education Conference
S1E-3
Session S1E
TABLE IV performance. A visual illustration of this relationship can be
MANN-WHITNEY U TEST FOR PRE & POST-TEST SCORES
found in Figure 1. This however contradicts the results
Pre-Test Score Post-Test Score
obtained for the control group, which as mentioned before,
though they appeared to be more confident in their ability to
Mean Rank – Control Group 17.53 17.00 learn programming, achieved a lower performance than the
experimental group.
Mean Rank – Experimental Group 13.18 17.00

Mann-Whitney U 79.500 133.000

Wilcoxon W 184.500 238.000


Z -1.387 .000

Asymp. Sig. (2-tailed) .165 1.000

Exact Sig. [2*(1-tailed Sig.)] .179 1.000(

Attitude and Academic Performance


After obtaining scores for each category in the attitude survey,
mean values were calculated as shown in Table V. A mean
value of 3 suggests that there is no difference in attitude to
programming. Scores for both groups for all the categories
ranged between 3.49 and 3.98 indicating that while students
did not have a negative attitude towards programming, the
attitude was not very positive. Looking at the various sub-
categories, we find that students in the control group were
more confident in their ability to learn programming than
students in the experimental group even though it was the FIGURE 1
experimental group that achieved more in terms of academic GRAPH SHOWING RELATIONSHIP BETWEEN CONFIDENCE IN LEARNING
PROGRAMMING AND POST-TEST SCORES
performance at the end of the semester. This leads to a
discussion on whether or not confidence necessarily leads to
Results also indicate that the experimental group had a
ability, an issue which has been raised in previous research
greater belief in the usefulness of programming, as well as
studies [13].
how they perceived the motivation of their teacher towards
TABLE V
their performance, which could be possible reasons for their
MEAN ATTITUDE SCORES BY GROUP better performance, but statistical analysis revealed no
Type of Personal Usefulness Perception significant influence of these factors.
Group confidence in of of
learning programming teacher Other Factors
programming motivation
Though not the primary focus of our study, we find it quite
Control Mean 3.73 3.86 3.49
Group useful to analyze our results in terms of gender. The results,
Std. 1.01 0.92 0.69 illustrated in Table VI, are very interesting when the overall
Deviation mean scores for both groups are compared by gender. The
Experimental Mean 3.65 3.98 3.86 data shows that there is minimal improvement in the mean
Group scores for males from the pre-test (40.87) to the post-test
Std. 1 0.66 0.64 (40.40), while the mean scores for the females improved
Deviation
significantly from 38.57 to 45.88, suggesting that female
Total Mean 3.70 3.91 3.64
overall performance was somewhat better than the male
Std. 0.99 0.82 0.68
overall performance. Again, due to the size of the sample, this
Deviation
result has to be interpreted cautiously.

While confidence in a course does not necessarily lead to


high performance, the highly positive correlation yielded
indicates that there is a strong relationship between confidence
and ability. Stepwise linear regression indicates that personal
confidence (r2 = .549) contributes significantly to academic
0-7803-9077-6/05/$20.00 © 2005 IEEE October 19 – 22, 2005, Indianapolis, IN
35th ASEE/IEEE Frontiers in Education Conference
S1E-4
Session S1E
TABLE VI worked together trying to figure out the approach to the
MEAN PRE-TEST AND POST-TEST SCORES BY GENDER
solution and explained to their peers areas that the other did
Gender Pre-Test Score Post-Test Score
not understand.
Male Mean 40.87 40.40
Std. 15.64 20.34 CONCLUSIONS AND FUTURE WORK
Deviation
Female Mean 38.57 45.88 Based on our study, the effects of peer tutoring cannot be
Std. 15.74 21.47 generalized due to the non-random assignment of students to
Deviation the experimental and control groups. While the data suggests
Total Mean 40.33 41.73 some evidence of a positive effect of peer tutoring on the
Std. 15.42 20.42 academic performance of students, as measured by the pre-test
Deviation and post-test [15], these results should be accepted with
caution due to the small sample size used in this study. This is,
however, a preliminary study and further investigations will
Previous research has shown that prior experience in
be done with larger sample sizes to see if the effect of peer
programming leads to better performance than those with no
tutoring may yield more positive results.
experience [14]. The results from this study contradicts this
The study also revealed that various aspects of a person’s
finding but could be due to the fact that the ITP course was
attitude contribute to academic performance. Personal
language-independent and more problem-oriented, while prior
confidence in learning programming contributed more
experience in programming would most likely have involved
significantly than the usefulness of programming and the
using a particular programming language.
perception of teacher motivation; however these factors
In Table VI where the results are compared by gender, we
cannot be ignored. This area would need further research.
find there was no significant observation in the mean scores
The solution to the problem of high failure rates may not
for females, which were constant (3.75) across all sub-
be any one silver bullet but a combination of factors and
categories of attitude, when compared with the mean scores
further investigation would be required to determine what this
for males, which varied minimally for the same sub-
combination is. Evaluation of the ITP course at UTech will
categories.
continue as we seek to determine the best strategies,
TABLE VI instructional and otherwise, that will result in improved
MEAN ATTITUDE SCORES BY GENDER success rates.
Gender Personal Usefulness Perception
confidence in of of ACKNOWLEDGMENT
learning programming teacher
programming motivation The authors would like to thank the students who participated
in this study.
Male Mean 3.68 3.96 3.61
Std. 1.09 0.76 0.67
Deviation
REFERENCES

Female Mean 3.75 3.75 3.75 [1] Mahmoud, Q. H., Wlodek D. & Swayne, D., " Making Computer
Std. 0.56 1.03 0.79 Programming Fun and Accessible ", Computer (IEEE), 2004, pp. 108,
Deviation 106-107.

Total Mean 3.70 3.91 3.64 [2] McNamarah, S., Pyne, R., " Teaching a First Level Programming
Course: Strategies for Improving Students’ Performance ", Journal of
Std. 0.99 0.82 0.68 Art Science & Technology, Volume 1, 2004, pp. 42-49.
Deviation
[3] McNamarah, S., Golding, P., " Assessment of Admission Criteria for
Predicting Students’ Academic Performance in SCIT ", (Unpublished).
The attitude survey was only administered at the end of
[4] Shih, Y., Alessi, S. M., " Mental Models and Transfer of Learning in
the course; however it would have been useful to examine the Computer Programming ", Journal of Research on Computing in
change in attitude from the beginning to the end of the course. Education, Vol. 26, Issue 2, 1993.

Teacher Observations [5] Cotton, J., " The Theory of Learning Strategies: An Introduction ",
Kogan Page Limited: Great Britain, 1995.
In the experimental group, it was observed that there was
[6] Slavin, R., " Educational Psychology: Theory and Practice", Allyn and
reluctance on the part of a few students to work with their peer Bacon: USA, 2000.
when the pair was someone they seemed to seldom interact
[7] Magin, D. J., Churches, A.E., " Peer tutoring in Engineering Design: A
with outside of the class. A few students actually commented Case Study ", Studies in Higher Education, Vol. 20, Issue 1, 1995.
that they preferred to work alone than with partners, however
they were encouraged to and did work with their peers for the [8] Schloss, P. J., Kobza, S.A., “The Use of Peer Tutoring for the
Acquisition of Functional Math Skills Among Students with Moderate
entire semester. Others however seemed very positive about Retardation ", Education & Treatment of Children, Vol. 20, Issue 2,
the peer tutoring experience. When given the problems, they 1997.

0-7803-9077-6/05/$20.00 © 2005 IEEE October 19 – 22, 2005, Indianapolis, IN


35th ASEE/IEEE Frontiers in Education Conference
S1E-5
Session S1E
[9] George, P. G., " The Effectiveness of Cooperative Learning Strategies in
Multicultural University Classrooms ", Journal on Excellence in College
Teaching, 1994, 5(1), 21.
[10] Williams, L. W., Wiebe, E., Yang, K., Ferzli, M., Miller, C., " In
Support of Pair Programming in the Introductory Computer Science
Course ", Computer Science Education, Vol. 12, No. 3, pp. 197-212,
2002.
[11] Best, J.W., Kahn, J.V., Research in Education 9th Ed., 2003, p. 178.
[12] Doepken, D., Lawsky, E., Padwa, L., Modified Fennema-Sherman
Attitude Scale. Retrieved August 16, 2004 from
http://www.woodrow.org/teachers/math/gender/08scale.html
[13] Thomas, L., Ratcliffe, M., Robertson, A., " Code Warriors and Code-A-
Phobes: A Study in Attitude and Pair Programming ", 34th SIGCSE
Technical Symposium on Computer Science Education (Reno, Nevada,
2003), pp. 363-367.
[14] Byrne, P., Lyons, G., " The Effects of Student Attributes on Success in
Programming ", Proceedings of the 6th Annual Conference on
Innovation and Technology in Computer Science Education
(Canterbury, United Kingdom),pp. 49-52.
[15] Sabin, R. E., Sabin, E.P., " Collaborative Learning in an Introductory
Computer Science Course ", 25th SIGCSE Technical Symposium on
Computer Science Education (Phoenix, Arizona,1994), pp. 304-308.

0-7803-9077-6/05/$20.00 © 2005 IEEE October 19 – 22, 2005, Indianapolis, IN


35th ASEE/IEEE Frontiers in Education Conference
S1E-6

Das könnte Ihnen auch gefallen