The Application of Language Acquisition Theory
to Programming Concept Instruction:

Chunks versus Programs from Scratch


Anne Applin

Center for Science and Mathematics Education

University of Southern Mississippi

Hattiesburg, MS 39406

Topic Overview

This study is designed to ascertain the relative effectiveness of two forms of programming assignments in an introductory computer science course. The premise is that to order to write well, one must be well read. A further goal of the study is to ascertain the effect of advance organizers (well-written, well-documented programs) on the student's understanding of the concepts. The template programs that the experimental group adds to or modifies as assignments are meant to act as advance organizers for concepts not yet covered and examples of how a 'good' program is written.

Theoretical Framework and Previous Research

Two of the dominant theories of learning in practice today are the interactionist theories of the cognitivists and the stimulus-response conditioning theories of the neo-behaviorists.

Cognitive psychology is concerned with what we know and how we know it. Students of education are familiar the work of Dewey, Piaget, Bruner, and Vygotsky The following summarizes the major points that are germane to this study.

Dewey's work suggests that knowledge is a series of continuous reconstructions. Piaget's studies explained this reconstuction. His model of how children learn suggests a cycle during which new information is incorporated into the current cognitive structure using whatever logic allows the information to fit within it. When presented with new facts that create problems with that framework, the child makes adjustments to the organization of the knowledge. This constant refinement of the cognitive structure is how children construct a knowledge base that makes sense and how critical thinking is developed. Vygotsky read Piaget's work but maintains that language and interaction play a larger role in children's understanding of their world.

George Miller proposed information processing theory in an article published in 1956. His meta-research indicated that humans process information in "chunks" or meaningful units of information. According to his study, the optimum number of chunks that can be processed at any given time is seven plus or minus two. This proposition has been verified at all levels of cognitive processing by independent research.

Levels of processing theory proposes that deeper processing leads to more permanent memory. Several studies have been conducted on methods to force deeper processing of new information and in the area of reading, it has been shown that teaching words in the context of paragraphs rather than in isolation leads to greater retention and increased fluency in the language.

David Ausubel's subsumption theory contends that meaningful learning and permanent retention of material is a function of the stability of existing anchoring ideas. The main teaching tool identified from his work is the advance organizer which is used to introduce a new concept in a familiar framework.

In the last twenty years there have been literally thousands of informal studies conducted in computer science classrooms. Empirical studies are less prevalent but some very important conclusions have been drawn from both. Among these is the connection between mathematical experience and success in computer science and the importance of critical thinking skills to the process. (Almstrum, 1991, Denning, 1989, Konvalina, 1983, Campbell, 1984).

In the midst of the constant battle over which language and which paradigm should be taught, it is important to note that in a study published in 1996, Brilliant and Wiseman concluded that neither the paradigm nor the language used in the first programming course had any impact on a student's future course work.

A recent study published in 1998 used two assignment based teaching methodologies to ascertain which method trained programmers to use better programming style and enhanced concept understanding. The methodologies were essentially programming in chunks within a framework versus programming from scratch. Frazer's study was conducted on junior level computer science majors and suggested that the impact of the two methods might be seen if they were used in a beginning course. She also suggested that learning style might play a part in the relative success of one method over the other.

Overview of the Study

Can advance organizers, in the form of template programs that the students add to or modify, facilitate a deeper understanding of the first course topics?

The nonequivalent control group design was chosen for this study. This quasi-experimental design allows educators to use the "naturally assembled groups" that we have as classes or sections of a course. Pretest measures are the student scores on the Watson-Glaser Critical Thinking Appraisal (WGCTA). This was chosen for several reasons, among them was the need to control for pretest-treatment interaction. By having a pretest instrument that does not reflect course content we control for that interaction. Student learning styles are being measured with the Productive Environment Preferences Survey (PEPS). The post-test is a final exam that has been used in previous semesters so that bias would not be introduced into the post-test measure.

Statistical equivalence of the two groups will be established using the following controlling variables: ACT mathematics score, college GPA, ACT English score, the PEPS scores and the WGCTA scores.

The study contained a control group and an experimental group. The experimental group received the independent treatment involving the use of pre-written, well-documented, programs that the students added to or modified as assignments for the course. The control group wrote their assignments from scratch. The learning objectives for corresponding programs were identical for the two groups and the amount of actual coding required for the corresponding programs was roughly equivalent. The weekly lecture material was presented using overhead transparencies and was therefore identical for the two groups as were the handouts. Great care was taken to keep the in-class discussion as close as possible.

Current Status

I will be presenting my proposal to my committee on February 17th, 1999. The study itself began last semester. In early October the experimental group size dwindled to ten students and became statistically insignigicant. My statistician informed me that it was possible to use the single section from this semester as an addition to the original control group and treatment will begin in March if my committee approves this extension.

Open Issues

It would have been nice if I had contacts at larger institutions to have this study replicated at a university where the teaching could have been done by someone other than myself and the chances of the students in the different groups discovering each other was minimized. One committee member strongly opposes my involvement with the subjects and I have had to be very careful to ensure that my behavior in no way influences the outcomes.


Almstrum, V. L. (1991). The relationship between pre-college mathematics and the undergraduate computer science curricula. Proceedings of the twenty-second SIGCSE Technical Symposium on Computer Science Education, USA, 23 (1), 124-129.

Brilliant, S. S & Wiseman, T. R. (1996). The first programming paradigm and language dilemma. Proceedings of the twenty-seventh SIGCSE Technical Symposium on Computer Science Education, USA, 28 (1), 338-342.

Campbell, P. F. & McCabe G. P.(1984) Predicting the success of freshmen in a computer science major. Communications of the ACM, 27 (11), p.1108-1113.

Frazer, K. (1998) Comparing the impact of two assignment-based teaching methodologies on student programming. Journal of Computer Science Education, 12 (1&2), p.21-26.

Konvalina, J, Wileman, S. A. & Stephens, L. J. (1983) Math proficiency: a key to success for computer science students. Communications of the ACM, 26 (5), pp 377-382.