In J. Kuljis, L. Baldwin & R. Scoble (Eds). Proc.

PPIG 14 Pages 45-57

Programming aptitude testing as a prediction of learning to program

Markku Tukiainen and Eero Mönkkönen

Department of Computer Science,
University of Joensuu, Finland
[email protected]
[email protected]

Keywords: POP-I.B. preprogramming knowledge, POP-II.A. individual differences

Abstract
Identifying variables that predict programming aptitude may help educators and employers select
potential students and employees. The local polytechnic school wanted to have a test that would
predict success in learning programming concepts. In this paper we present an empirical study that
attempts to evaluate how predictive the developed testing of programming aptitude was.
Introduction
Programming aptitude testing has coexisted with the programming profession and education for
decades. Probably the best known early test used to predict programming aptitude was the IBM
Programmer Aptitude Test (PAT). The test consisted of three parts. The first part required the
participant to figure out the next number in series, the second part to figure out analogies represented
in figures, and the third part to solve arithmetic problems. The test was widely used in industry (in
1966, 349 American and Canadian companies used IBM-PAT to screen job candidates), but the
relationship between the test and the actual job performance of programmers were often quite low
(Mayer & Stalnaker, 1968).
Wolfe (1971) developed a series of tests to overcome some of the limitations of earlier tests. His tests
required participants to manipulate numbers according to a complicated set of procedures. These tests
primarily assessed the participant’s ability to follow a detailed procedural instructions. The progress
of programming aptitude testing continued to add new features to test battery, the most recent tests
include at least a basic set of tasks that measure, for example logical ability, interpretation of
specifications, attention to detail, accuracy and reasoning with symbols (see e.g. PAAT or I-PAT,
Winrow, 1999).
The admission to educational institutes requires scholastic aptitude testing in some countries. In the
U.S. the SAT tests are used on a normal basis for college admissions, although at the present time
some schools, for example University of California, are contemplating the possibility of scrapping the
SAT requirements altogether (Organ, 2001). Like all tests, the SAT tests are imperfect. It is argued
that they account for no more than about 20% of the variance in first-year college grade point
averages (Ransdell, 2001). It is suggested that some of the tests used for academic assessment
purposes, e.g. English for Specific Academic Purposes, are suggested to be abandoned altogether
because of the difficulties inherent in testing (Clapham, 2000). The basic conclusion of these student
performance studies is that aptitude, which is the potential to learn a task, is not necessarily a good
indicator of performance on the task per se.
This paper reports a study, which was conducted to find out whether the programming aptitude of
Finnish polytechnic students can be predicted using the results of scholastic aptitude tests that
measure one’s ability to make logical conclusions, learning ability and verbal ability.
Method
We invited 33 volunteers to participate in the experiment. They were administered the scholastic
aptitude test before entering the study program, a pre-test of programming aptitude before the first
programming course, and a post test as the final exam of the course.

14th Workshop of the Psychology of Programming Interest Group, Brunel University, June 2002 www.ppig.org
Tukiainen & Eero Mönkkönen ii

Subjects
The subjects in the experiment were students of the Business Information Degree programme at North
Karelia Polytechnic. All of them were first-year students and all volunteered for the experiment. They
received no pay for their participation nor was any other incentive used. Their background education
consisted of a normal Finnish elementary education with 12 years of public schooling, ending with a
matriculation examination.

Materials
The pre-test and the post-test materials are presented in appendices A and B respectively at the end of
this paper.

Data collected
Students’ mathematics and Finnish high school grades (MATH = mathematics grade, FINN = Finnish
language grade, scale 1-10), results of scholastic aptitude tests (LOG = logical aptitude, VERB =
verbal aptitude, LEARN = learning aptitude, scale 1-5) and results of the programming aptitude test
(scale 1-5) and results of the final exam (scale 1-5) of students’ first programming course at the
polytechnic were used in this study.

Design
All of the subjects participated at each of the tests. There were no differences made in the treatment of
the participants. Of the original 33 subjects, 7 were removed from the later analysis because they did
not for various reasons attend the final exam.

Procedure
The high school marks were collected from the students’ applications to the Polytechnic. The
scholastic aptitude tests were carried out by a professional company. In these tests the applicants’
verbal ability (VERB) was tested by having them follow a given set of instructions. Their logical
ability (LOG) was tested through tasks in which they had to handle relations between different words.
Their learning ability (LEARN) was tested through tasks where the applicants had to convert words
into numbers according to a given set of instructions as fast as possible. The scholastic aptitude tests
were carried out during the entrance exams to the polytechnic. The programming aptitude test was
carried out after the students had been admitted to the school and before the start of their first
programming course. The pre-test (a programming aptitude test) (Huoman, 1989) was one designed
by M.Sc. J. Huoman, a graduate from the Department of Computer Science, University of Joensuu.
The test consists of a collection of programming-like tasks that do not require previous programming
experience. The post test was the final exam of the students’ first programming course, which lasted
one semester with 6 hours a week lectures and exercises The tool used during the course was Ms
Visual Basic.
Results
We equated the students’ understanding of programming concepts in the final exam with the mastery
of their first course materials. The questions in the final exam concerned the mastery of Ms Visual
Basic concepts, control flow constructs, understanding of the data and control flow and two simple
programming tasks.

Number of ABILITY Total

students 2 3 4
Final 1 2 3 2 7
exam 2 1 3 1 5
grade 3 2 4 1 7
4 4 1 1 6

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen iii

5 1 1
Total 9 12 5 26

Figure 1: Cross-tabulation of scholastic aptitude and the final exam grades

Figure 1 shows the distribution of the final exam grades and the grades from the scholastic aptitude
tests. The attributes in the scholastic aptitude test (LOG, VERB, LEARN) have been unified in a
single variable entitled ABILITY (the average of the original attributes). The justification for the
unification was tested using Cronbach’s test (alpha = 0.8413).

There were no significant correlations between the grades from scholastic aptitude and the grades
from the final exams. Neither were there any significant correlations between the programming
aptitude test and the mathematics grade, the Finnish language grade and the attributes in the scholastic
aptitude test (See Table 1).

If we take a closer look at grades from the programming aptitude test and the grades from the final
test, we realize that there seems to be no interaction between them (See Figure 2).

MATH FINN LOG VERB LEARN HUOM

FINN P ,912**
Sp ,892**
LOG P ,364* ,320
Sp ,321 ,317
VERB P ,370* ,377* ,485**
Sp ,351* ,377* ,428*
LEARN P ,378* ,412* ,362* ,384*
Sp ,365* ,411* ,378* ,398*
HUOM P -,209 -,177 ,201 ,081 ,023
Sp -,203 -,107 ,237 ,095 ,061
EXAM P -,280 -,362* ,241 -,186 -,220 ,513**
Sp -,275 -,304 ,255 -,167 -,247 ,518**
Table 1 – . Cross-tabulation of variables and Pearson and Spearman correlations. * Significant
correlation (p < 0.05). ** Very significant correlation (p < 0.01).
Number of Huoman test grade Total
students 2 3 4 5
Final 1 5 1 1 7
exam 2 4 1 5
grade 3 5 2 7
4 1 4 1 6
5 1 1
Total 15 8 2 1 26
Figure 2: Cross-tabulation of Huoman aptitude test and the final exam grades

Table 1 shows that there is either a significant or very significant correlation between the attributes of
scholastic aptitude tests. This seems quite natural. When the final exam grades are compared to the
attributes in the scholastic aptitude test, the other attributes, with the exception of the logical attribute,
have a negative correlation to the grades. So it seems that the final exam could emphasize the logic
ability of the students.

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen iv

There is a very significant correlation between the Huoman programming aptitude test and the final
exam. So it seems that the Huoman test and the exam measure the same abilities of the students.
Because no other dependencies to the measured background variables were found, this made us
wonder if there are any other background variables that could better explain the interaction. We
decided to ask prior programming experience from the subjects. They all answered the email
questionnaire concerning their experience.
Thus the group was divided into two study groups. The first group (n = 11) had no background
knowledge of programming, whereas the second group (n = 15) had some experience of programming
(on an average they had experience on one programming language with 1 to 5 programs with a length
of 1-100 lines). Their background was controlled by their final mathematics and Finnish language
marks on their high school diploma and the scholastic aptitude test. There were no significant
differences between the groups.
When we take into consideration the background knowledge of the subjects, we see a difference in the
distribution of Huoman aptitude test results and the grades from the final exam (See Figures 3 and 4).
In Figure 3 (the group with no prior programming knowledge), it seems that the Huoman
programming aptitude test gives no evidence of success in the programming exam. The only
noticeable trend could be that if a student is very successful in the Huoman test (grades 4 to 5), then
she will be successful in the programming test as well.
In Figure 4 (the group with prior programming knowledge), there seems to be a more systematic
arrangement. The Huoman aptitude test seems to provide a floor limit to the success in the
programming test. This is not a very clear effect, but seems to be coherent.

6
Final exam grade

0
1 2 3 4 5 6

Huoman test grade

Figure 3: Group with no prior programming knowledge

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen v

Final exam grade

4

0
1 2 3 4 5

Huoman test grade

Figure 4: Group with prior programming knowledge

Discussion
The immediate question that raises from the study is then what to measure as a prediction for
programming aptitude. Some research (see e.g. Whitley 1996) has suggested, for example the
importance of cognitive style or psychological type to predict one’s aptitude. Using the Myers-Briggs
type indicator it has been determined that in the dimension SN (sensing-intuiting), intuitive persons
are more prone to succeed in programming than sensitive persons. This seems to be in contrast with
the stereotypic ISTJ-personality (introvert, sensing, thinking, and judging) associated with
programmers (Lyons, 1985).
On the other hand, if we consider the skills portfolio needed in the IT industry, there is a demand for
very different sets of talent other than just pure programming or coding (Nakayama & Sutcliffe,
2001). What the industry needs more is not the aptitude tests for programming but the establishment
of credentials that accurately represent the ability of software engineering (Shaw, 2000). In
programming we will probably continue to see a ratio of 20:1 differences in performance as has been
customary. The aptitude just presents the potential to learn to program not the performance in
programming. If we think about the factors affecting the performance in programming one cannot
deny the role of native talent. Yet one’s psychological style is important. Great programmers are also
great tinkerers (Curtis, 1991), they have the intrinsic motivation to see how a program or a system
works.
Conclusion
The ability to predict an individual’s potential to learn programming concepts is important for many
reasons (Weinberg, 1998). Our immediate concern in this study was to find tests other than the
scholastic aptitude tests to help in the selection of the students to the polytechnic programming degree
program. Although the Huoman programming aptitude test predicted no more than 25 percent of the
final exam results (which is the common finding of almost any of the measured predictors, see e.g.
Evans & Simkin, 1989), it was considered to predict the floor limit of performance in the first
programming course for those who had prior knowledge of programming concepts.
Acknowledgements
We would like to thank Professor Jorma Sajaniemi for encouraging and stimulating the writing of this
paper. We wish also to acknowledge Professor Sajaniemi as the supervisor of the empirical study
presented here.
References
Clapham, C. (2000) Assessment for academic purposes: where next? System, 28: 511-521.

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen vi

Curtis, B. (1991) Techies as non-technological factors in software engineering? In the proceedings of

the International Conference on Software Engineering (Proceedings of the 13th international
conference on Software engineering 1991), Austin, Texas, United States, 147-148.
Evans, G.E. & Simkin, M.G. (1989) What best predicts computer proficiency? Communications of the
ACM, 32 (11): 1322-1327.
Huoman, J. (1986) Predicting programming aptitude. Master’s Thesis, Department of Computer
Science, University of Joensuu: Joensuu (in Finnish).
Mayer, D. B. & Stalnaker, A. W. (1968) Selection and evaluation of computer personnel – the
research history of SIG/CPR. In the proceedings of the 1968 ACM National Conference (23rd
ACM National Conference), 657-670.
Nakayama, M. & Sutcliffe, N.G. (2001) IT skills portfolio research in SIGCPR proceedings: analysis,
synthesis and proposals. In the proceedings of the Special Interest Group on Computer Personnel
Research Annual Conference (Proceedings of the 2001 ACM SIGCPR conference on Computer
personnel research 2001) , San Diego, California, United States, 100– 113.
Organ, D. W. (2001) Predicting success in school and at work. Business Horizons, May-June 2001: 1-
2.
Ransdell, S. (2001) Predicting college success: the importance of ability and non-cognitive variables.
International Journal of Educational Research, 35: 357-364.
Shaw, M. (2000) Software engineering education: a roadmap. In the proceedings of the International
Conference on Software Engineering (Proceedings of the conference on The future of Software
engineering 2000) , Limerick, Ireland, 371–380.
Weinberg, G. M. (1998) The psychology of computer programming (silver anniversary edition). New
York: Dorset House Publishing.
Whiley Jr, B.E. (1996) The relationship of psychological type to computer aptitude, attitudes, and
behavior. Computers in Human Behavior, 12 (3): 389-406.
Winrow, B. (1999) The Walden programmer analyst aptitude test. Dr. Dobb’s Journal, Fall 1999,
Software Careers. (http://www.ddj.com/documents/ s=894/ddj9914b/9914b.htm).
Wolfe, J. M. (1971) Perspectives on testing for programming aptitude. In the proceedings of the 1971
Annual Conference of the ACM , 268-277.
Wolfe, J. M. (1971) Wolfe programming aptitude test (school edition). In the proceedings of the Ninth
annual SIGCPR conference, Illinois, United States, 180 - 185.

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen vii

Appendix A. Huoman Programming aptitude test.

Task 1
The company has information of their employers in three different lists. You have acquired all three lists, which all have
little bit different information depending on the purpose of the list.
The lists have following information of the employees:
List 1: the number, the name, the occupation and the department
List 1 is ordered by the number of employee to ascending order
List 2: the name, the number, the address, the phone number and the social security number
List 2 is organized to alphabetic order by the name
List 3: the number, the social security number, salary and some secret information
List 3 is ordered by the number to ascending order
Your job is to make a report of those employees whose salary is greater than 2000 €. The repost has to display the name, the
address, the department and the salary of the employees.
Describe how you would solve the problem.

Task 2
Let i and j be integers.
List all values of i and j, that make the expression to be always true
a) (i>=1) and (i<=5)
The expression is true when i has values:
b) (j>=3) or (j<=7)
The expression is true when j has values:

Task 3
Write a letter sequence, that is next on the series. Try to determine the general form of the series:
a) bce , bbcde , bbbcdde , __________________________
The general form:
b) bcace , bcacacace , bcacacacacace , __________________________
The general form:
c) __________________ , abcccdd , abbccccdd , abbbcccccdd , …
The general form:

Task 4
Your job is to sum up 50 numbers and at the end reprt the sum and the count of numbers that were positive numbers (>0).
Describe how you would solve the problem or write in a pseudo-code.

Task 5
Your job is to deduce which of the given numbers is the smallest. How many comparisons you have to make if you have:
a) 3 numbers ___________________
b) 6 numbers ___________________
c) n numbers ___________________ (n>=1)

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen viii

Appendix B: The final exam questions.

Task 1.
The following statements/reserved words are from Ms Visual Basic. Describe shortly what they are:
a) Option Base 1
b) Cint
c) Boolean
d) String
e) InputBox

Task 2.
a) Write the following loop using For-statement
k=1
Do until k > 50
Number_table (k) = 0
print k
k=k+1
Loop

b) Write the following Case - statement using conditional statements (If – Then – Else):
Select Case intNumber
Case 1
Print ”Number is 1”
Case 6
Print ” Number is 6”
Case Else
Print ”All the other numbers”
End Select

Task 3.
What does the following program print?
Dim a,b,c,d,e As Integer
a= 2
if a = 1 then
a = a+1
end if
if a = 2 then
a = a+2
end if

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen ix

if a = 3 then
a = a+3
end if
if a = 4 then
a = a+4
end if
Print ”a=”+a
b=2
Select Case b
case 1
b=b+1
case 2
b=b+2
case 3
b=b+3
case 4
b=b+4
Print ”b=”+b
c = 11 \ Int(8/3);
Print ”c=”+c
d = 20
for e = 1 to 4
d=d–e
Next e
Print ”d=”+d
a=1
e=1
while a < 8
a=a+1
if a < 4 then
e=e+a
else
e=e+2
end if
wend
Print ”e=”+e

Task 4.
Write a program that first sorts into ascending order and then prints out the numbers in the Credits –table. The Credits –table
has been declared:
Option Base 1
Dim Credits (15) As Integer

PPIG 2002, Brunel University www.ppig.org

Tukiainen & Eero Mönkkönen x

Task 5.
Write a program that reads numbers from the user, until the number 0 is encountered or 20 numbers have been inputted. The
numbers must be saved into the Information –table. After this the program should go through Information –table, and figure
out:
a) Is the number 50 in the table?
b) How many numbers are < 50?
c) What is the average of all the numbers?

PPIG 2002, Brunel University www.ppig.org