FIRST PRINCIPLES OF CS INSTRUCTION
Katrin Becker
Graduate Division of Educational Research,
University of Calgary
Calgary, Alberta, Canada
becker@minkhollow.ca
0TU

U0T

Abstract
Much attention has been paid in recent years to finding more flexible and less
prescriptive approaches to the design of instruction than those put forward during the
latter part of the twentieth century. A view of instruction as causal and largely
behaviouristic has given way to one that is guided and primarily constructivist. In his
current research, David Merrill outlines five fundamental principles of instruction which
have broad implications for teaching computer science (CS). These five principles are: 1)
solving real-world problems, 2) activating existing knowledge to build new knowledge,
3) demonstrating new knowledge to the learner, 4) allowing learners to apply new
knowledge, and 5) integrating knowledge into the learner’s world. The following paper
describes these principles and discusses how they related to instruction in CS.
1.0 Introduction
M. David Merrill’s career in instructional technology, with a career has spanned
40 years, and include numerous significant contributions to the field. He is probably best
known for his Component Display Theory [1]. In the 1990’s he was one of the foremost
proponents of “Second Generation Instructional Design (ID2)” [2] which acknowledged a
more open-ended and less prescriptive approach to ID, and included Instructional
Transaction Theory[3] and ID based on knowledge objects.
The First Principles of Instruction are the result of a systematic review of
instructional design theories, models and research. Each of the principles included
satisfies the following properties: 1) promotes more effective, efficient or engaging
learning, 2) is supported by research, 3) is general enough to apply to any delivery system
or instructional methodology, and 4) is design oriented with direct relevance to
promoting learning activities. Merrill defines a ‘principle’ as a basic method, and
describes it as a “relationship that is always true under appropriate conditions regardless
of program or practice (variable methods).” [4, p43] These principles constitute a set of
fundamental elements common to all effective instructional design. Merrill hypothesizes
that 1) “Learning from a given instructional program will be facilitated in direct
proportion to the implementation of first principles of instruction”, and 2) the “learning
from a given instructional program will be facilitated in direct proportion to the degree
that first principles of instruction are explicitly implemented rather than haphazardly
implemented.” [4, p43] If true, then we stand to gain from examining how Merrill’s First
Principles apply to computer science instruction.
2.0 Computer Science Education
Computer Science is fortunate to have two international organizations (ACM &
IEEE) who together have produced widely accepted curricula describing undergraduate

programs in computing [5]. Although it is and will probably remain a subject of debate,
they nonetheless stand as standards against which programs and courses can be
compared. These extensive documents go a long way towards describing what computer
science is, but the purview of this document does not really extend to theories or models
of how to teach computer science. Instructional Design methods for teaching science and
math receive considerable attention [6, 7], but less is paid specifically to computer
science education, and a search at GoogleScholar for “instructional design” and
“computer science education” turns up fewer than 140 hits if we look at the last ten years.
The vast majority of these references concern themselves with web-based and distance
education, and the use of knowledge objects. Specific ID theories and models are rarely
mentioned in the context of designing instruction for computer science education.
Several areas in computer science have had significant influence over learning
and instructional theories, notably human-computer interfaces, and artificial intelligence,
but it does not seem to be a reciprocal relationship. On the whole, learning and ID
theories have had relatively little impact on how computer science is taught, and a great
many courses are still taught much the same way they were a generation ago. Admittedly,
we have always made extensive use of technology, which has evolved over time, but in
many cases the technology is used less as a system for the support of learning and more
as an apparatus on which to run programs and develop documentation.
3.0 Merrill’s First Principles of Instruction
According to Merrill, the success of a given instructional program will be directly
proportional to how well and how deliberately the first principles are implemented. [4]
First principles are intended to be domain NON-specific and so should be applicable to
almost any discipline. The following paragraphs provide a general description of these
first principles, and the subsequent section relates each to computer science education.
3.1 Problem
“Learning is facilitated when learners are engaged in solving real-world
problems.” Problems should be authentic, real world, and personal. Not mentioned by
Merrill, fantasy can also have authentic and personal elements that result in highly
motivating scenarios. Learning to solve problems involves several levels of instruction,
from understanding the problem through identifying the tasks and implementing the
operations that make up the tasks. Often instruction concentrates on the implementation
level and fails to involve learners in the problem level. Complex problems are mastered
only with adequate coaching, or after sufficient time has been spent building up from
simpler problems.
3.2 Activation
“Learning is facilitated when existing knowledge is activated as a foundation for
new knowledge.” Also described as scaffolding – this involves making explicit
connections between new knowledge and past experience. This must be more than
merely testing prerequisite knowledge. It involves activating those mental models that are
relevant to the current context – with an emphasis on active participation.
3.3 Demonstration
“Learning is facilitated when new knowledge is demonstrated to the learner.” This
principle tells us we must show people what we want them to learn, not simply tell them.
Included here are examples as well as counter examples, demonstrations, visualizations,

and modeling. Media should play an instructional role, multiple representations should be
used, and learners should receive appropriate guidance for such things as: finding
relevant information and explicit comparison of demonstrations.
3.4 Application
“Learning is facilitated when new knowledge is applied by the learner.” Learners
must be guided in their problem solving. This involves a coaching process where more
guidance is offered to start, including error detection and correction, gradually
withdrawing assistance until they are able to manage on their own. Merrill
says, "Appropriate practice is the single most neglected aspect of effective instruction."
And as we all know, “You can't rush Mother Nature.” Unfortunately, education cannot be
made into an exercise in efficiency. Learning is a natural process and takes time.
3.5 Integration
“Learning is facilitated when new knowledge is integrated into the learner’s
world.” In and of themselves, animation, multimedia, even games have only a temporary
effect on motivation. Ultimately, learning is its own reward. A case in point is the video
game industry, where meeting challenges and demonstrating improvement are the
motivators that have made this industry into a multi-billion dollar business. In games,
once a new skill is learned, players seek ways to apply that skill and improve upon it. The
popularity of games is less about winning than it is about meeting challenges and
showing improvement. This is also true in real life. Reflection, discussion, and finding
new ways to use new knowledge and skills all facilitate learning. Students quickly learn
to recall and recite on cue, but the information they so readily regurgitate is not available
to them for application elsewhere. Knowledge that is not used remains inert.
4.0 First Principles Applied to CS
This section applies the principles defined above to computer science instruction,
using examples. Note that each principle interconnects with others, so that a single, wellchosen example can meet the criteria for all of them. The more of these carefully-chosen
problems we have in our courses, the more effective the instruction will be.
4.1 Problem (engagement in solving real-world problems)
Some would say that problem solving is the very core of all computer science,
and one would be hard-pressed to find a programming class of any description that did
not include assignments posed as problems to be solved. However the kind of problembased learning proposed as Merrill’s first principle is more deliberate and structured [810] than most typical CS assignments. This kind of problem-based learning is distinct
from showing examples, and also from assigning a problem and leaving them to it. It
involves a structured, guided approach that begins with a delineation of the problem and
its domain, and progresses through the research and discovery of relevant knowledge and
data, and finally through a presentation of a solution and a reflection on what was learned
and deliberate and conscious connection with existing knowledge.
Problems should be relevant – realistic, but also ones that consider learner
characteristics. Problems should be interesting for our students while still forcing them
into contact with the necessary content. If building on a student’s expertise results in a
more engaging problem, we should look seriously at how students spend their time.
Problems drawn from accounting or management are common in CS, but seriously, how
many students do you know that do accounting as a hobby or pastime? They do play

games. For many students, digital games provide a rich source of problems that they find
both challenging and engaging. [11] In senior courses it is often possible to allow the
students to propose problems to solve, but in introductory classes, they will typically
require more direction. Still, it is possible to offer a choice among several alternatives,
where each problem to be solved addresses the same conceptual issues. For example, the
classic arcade game Frogger, a hospital emergency room simulation, and a dynamic
restaurant menu system are all problems that can involve the use of inheritance and
polymorphism. When pedagogical objectives are clearly defined, then the subject matter
can become flexible, thereby creating relevance, and the potential for a personal
connection with the problem.
4.2 Activation (activate existing knowledge as foundation for new knowledge)
To activate existing knowledge, we must know what that existing knowledge is.
These days, most probably have cell phones, know how to do email, surf the web, find
music and videos, and play games. They will feel themselves quite proficient, yet in
reality will typically lack sophistication in such things as searching and the critical
assessment of resources. It is important to acknowledge their skills, without either
assuming too high a level of sophistication, nor too much nescience. One will discourage
students, while the other will bore them. Either way, they will become disengaged and
motivation will suffer.
Introducing the notion of greedy algorithms by examining or developing an
algorithm for making change is one example of activating existing knowledge. These
days, few people seem to calculate change themselves or count it out from the amount
charged, but the problem of how to allot a minimal number of bills and coins still has a
real-life connection that can be used to build an effective bridge. Another example could
use the typical human approach to searching for a name in a phonebook as a means of
introducing a binary search. As an aside, this particular example will probably lose its
relevance soon, as searching becomes more and more automated. It will then become
necessary to locate other examples – perhaps finding a specific music CD in a collection,
or one’s final grade in a large class list. This highlights a key point: relevance changes
over time. If we wish to start from where the students are, then we must be prepared to
assess the knowledge they bring to the situation regularly.
4.3 Demonstration
Learning objects can be useful here, although as with any other resource, they
should be carefully assessed before they are used. [12] Students report that live, in class
demonstrations of programs are more effective than simply reading through and
explaining code. Another effective mechanism is modeling behavior such as answering a
question by performing a quick search on the internet, which of course would include a
demonstration of rapid assessment of potential resources. Instructor competence with the
technologies available is crucial if students are to be expected to demonstrate competence
themselves. Credibility is lost if the instructor cannot give the impression that he or she is
able to solve the problem using the tools available to the students.
Another approach often reported as welcome by students is to develop a solution
to a problem on the fly – including errors and blind alleys. One dilemma for many
novices is that the process used by the instructor in class when developing a solution to a
problem is one that has typically been well-rehearsed and delivered many times, yet when

the novice attempts to emulate this approach later, the result is not nearly so wrinkle-free.
We often learn a great deal from our mistakes and to some extent we can also learn from
the mistakes of others. We should allow our students to observe error recovery.
4.4 Application (by the learner)
The computer science curriculum is overflowing with content. The body of
knowledge associated with the discipline has grown and evolved over the last 40 years
and in our desire to provide students with as much information and knowledge as we can,
we sometimes forget that they still need time to absorb the information - and this includes
time for practice. Chess masters, musicians, swimmers, and others have been shown to
require on average ten years to achieve expert status! [13] What makes us think we can
create expert programmers and computer scientists in just four?
We can no longer cover the same ground in an undergraduate program as we did
when we were students, yet we cling to a desire to do so, and to add all that we have
learned since graduating as well. Modern theories of education, including Merrill’s imply
that students will be better prepared through the acquisition of deep knowledge in fewer
areas than through a shallow or cursory acquaintance with many. We do our students a
disservice by attempting to move on to the next topic too quickly, and often in our
eagerness we end up teaching content that is “a mile wide but only an inch deep”. One
the other hand some topics can be taught using a spiral approach. The topic can be
introduced but treated superficially in one course, and then addressed in greater depth in
one or more subsequent courses. Recursion, algorithm analysis, and program testing are
all topics that lend themselves to such an approach. Others, like ethics, professional
practice, and communication skills can also be broken up and spread across multiple
courses, but lend themselves more to sectioned, in-depth study than to layering from
superficial overviews to deep learning.
4.5 Integration (new knowledge is integrated into the learner’s world)
It is like coming full circle – we start from where the student is, and end by
helping them convert new knowledge into a new starting point. If the curriculum is well
integrated, subsequent courses can quite literally pick up where the others left off. If it is
not, time will be required in each course to re-assess what knowledge the students bring.
The assignment component of the freshman class taught by the author on and off
for over twenty years consisted of no less than six nor more than ten programming
assignments – each one focusing on a different concept or set of concepts. The time
pressure is high, and students barely get one handed in than they are headed full speed
into the next one. There is no time in between to reflect, and in an effort to maintain
interest, each problem bears little relationship to the last. This kind of approach does not
promote integration of learning. While it is sometimes impractical to create assignments
that logically follow one another, attempts should be made to do so. When this is not
possible, both practical and conceptual connections can be made explicit.
At the course level, content can and should be tied to the students’ lives, current
events, research, and professional practice whenever possible. For example, reading an
article like Walpole’s “Designing Games for the Wage Slave” [14] connects well with
student’s lives, can be used to discuss software design on many levels, and can be used as
the basis for assessment criteria on a programming assignment.

5.0 Conclusions
Many courses would from a more deliberate approach to the design of instruction,
and following the principles laid out here can help. Freshmen have a broader base of
experience than they had even five years ago, and integrating this experience into our
courses forms connections that engages students in ways not otherwise possible. For
example, today’s freshmen have never really known a world without PCs, the internet
and computer games. Most own cell phones and some sort of high-capacity music player
that doubles as mass storage. These are as much a part of their immediate surroundings as
telephone land lines and televisions were for the generation before. Contextualizing their
experience draws them in, and maintaining relevance while guiding their practice will
sustain them. In short, following these first principles will not only help students learn
about computer science, but will also teach them how to be computer scientists.
6. References
1. Merrill, M.D., Component Display Theory, in Instructional-design theories and models, C.M.
Reigeluth, Editor. 1999, Erlbaum: Hillsdale, N.J. p. 279-333.
2. Merrill, M.D., Z. Li, and M.K. Jones, Second generation instructional design. Educational
Technology, 1991. 30(2): p. 7 - 14.
3. Merrill, M.D., Z. Li, and M.K. Jones, Instructional transaction theory: an introduction.
Educational Technology, 1991. 31(6): p. 7 - 12.
4. Merrill, M.D., First Principles of Instruction. Educational technology research and
development: ETR & D, 2002. 50 Part 3: p. 43-60.
5. CC 2001: Final Report of the Joint ACM/IEEE-CS Task Force on Computer Science
Education, E. Roberts and G. Engel, Editors. 2001, IEEE Computer Press: Los Alamitos, CA.
6. Mintzes, J.J., J.H. Wandersee, and J.D. Novak, Teaching science for understanding: a human
constructivist view. 1998, San Diego: Academic Press. xx, 360.
7. Taylor, P., P.J. Gilmer, and K.G. Tobin, Transforming undergraduate science teaching: social
constructivist perspectives. 2002, New York: Peter Lang. xxiv, 481 p.
8. Jonassen, D.H., Learning to solve problems: an instructional design guide. Instructional
technology & training series. 2004, San Francisco, CA: Pfeiffer. xxiv, 218 p.
9. Schank, R.C., T.R. Berman, and K.A. Macpherson, Learning by Doing, in Instructional-design
theories and models: vol. 2, a new paradigm of instructional theory. 1999, Lawrence Erlbaum
Associates: Mahwah, N.J. p. 161-181.
10. Savery, J.R. and T.M. Duffy, Problem based learning: An instructional model and its
constructivist framework, in Constructivist Learning Environments: Case Studies in
Instructional Design, B.G. Wilson, Editor. 1996, Educational Technology Publications:
Englewood Cliffs, NJ. p. 135-148.
11. Becker, K. and J.R. Parker. All I Ever Needed to Know About Programming, I Learned From
Re-writing Classic Arcade Games. in Future Play, The International Conference on the Future
of Game Design and Technology. 2005. Michigan State University, East Lansing, Michigan.
12. Wiley, D.A., The instructional use of learning objects. 2002, Bloomington, Ind.: Agency for
Instructional Technology: Association for Educational Communications & Technology.
13. Ericsson, K.A., R.T. Krampe, and C. Tesch-Romer, The Role of Deliberate Practice in the
Acquisition of Expert Performance. Psychological review, 1993. 100(3): p. 363-406.
14. Walpole, S., Designing Games for the Wage Slave. 2004, gamedev.net.