Teaching computer programming: A connectionist view of pedagogical change.

One of the major issues of teaching computer programming is concerned with determining pedagogical factors that contribute to students' learning and instructional improvement. Classroom teachers are the keys to making learning happen as well as improving teaching in schools. Teachers' perspectives on teaching affect how teachers perceive the nature of the subject, the problems of instruction and how they adopt solutions or ways to improve instruction. Teachers' views of knowledge have great impact on teachers' perspectives and their pedagogical practices. Over the years, two predominant computational theories of mind have emerged from the study of artificial intelligence, namely, the symbolic and the connectionist models. In this article, I argue, based on a qualitative study of 12 computer studies teachers in Hong Kong, that the connectionist view provides an alternative way of thinking about understanding and knowledge which leads to better insight on teaching and facilitates pedagogical change.

Introduction

Computer programming has been a subject widely taught in schools since the 1980s (Mayer, 1988; Nickerson, Perkins, & Smith, 1985; Papert, 1980). Papert encourages a new kind of learning environment which demands `free contact' between children and computers. Children, in getting the computer to proceed on developing computer programs, also engage themselves in thinking development. He envisioned that powerful intellectual skills are developed in the programming process. His advocacy for computer programming as a worthwhile educational domain seemed to have implications for the instructional methods as well as the expected outcomes of teaching and learning computer programming in schools (Mayer, 1988). Research in the past two decades or more has supported the view that the teaching issue does not simply refer to the delivery of programming skills but paves the way for the cultivation of a higher level of intellectual development. Jonassen (1996, p. 233) puts forward the idea that `computer programming is more likely to be used as a mindtool if it is taught more as a problem-solving tool than as an academic subject to be mastered'.

Based on the investigation of the social characteristics of effective learning of BASIC and LOGO, Webb and Lewis (1988) suggest that detailed analysis of the social context can bring important insights into the nature of effective classroom learning environments. Sloan and Linn (1988) also suggest that a careful analysis of classroom conditions can give useful information about the nature of effective instruction for programming. Perkins, Schwartz, and Simons (1988) argue that the pedagogical environment is crucial in learning computer programming. Nevertheless it is widely agreed that learning to program is difficult (Smith & Webb, 2000). A number of recent studies attempt to develop and examine useful tools in helping students learn programming, such as algorithm animation (Guimaraes & de Lucena, 1995; Kann, Lindeman, & Heller, 1997), program simulation (Thomas & Upah, 1996; Yuen, 1999) and visualisation tool (Smith & Webb, 2000). However one of the major issues of teaching computer programming is concerned with determining pedagogical factors that contribute to students' learning and instructional improvement.

Classroom teachers are the keys to making learning happen as well as improving teaching in schools. The improvement of instruction in education is a complex professional challenge to teachers. This paper will start by developing an initial theory of teachers' perspectives on the teaching of computer programming and will explore the underpinning pedagogical assumptions reflected in their perspectives. Teachers' perspectives are likely to impact on their instruction and the change of teachers' perspectives is the first major breakthrough for instructional improvement. I argue that the connectionist view of understanding provides an alternative way of thinking that breaks the taken-for-granted perspectives of teachers, leads to better insight on teaching and facilitates pedagogical change. Following the implication of the connectionist view, I will finally discuss four issues regarding pedagogical change.

Teacher thinking and pedagogical practice

Teaching is a complex activity and usually associated with practical problems as well as teacher thinking. In the past two decades, an increasing amount of educational research has been conducted to study issues on the practice of teaching in relation to teacher thinking. Young (1981) notes: `Epistemologies, curriculum codes, assessment approach and approach to pupil control constitute elements of a coherent pedagogical ideology' (p. 199). This view has been supported by succeeding studies.

Ben-Peretz (1984) makes the point that George Kelly's theory of personal constructs

   offers one valid and interesting approach to the study of teacher thinking
   which may yield some true statements in this important research area ...
   Teacher education and professional development programs could benefit from
   the recovery of personal constructs in relation to various components of
   the educational environment, such as pupils, peers, parents, physical
   settings and curricula. Teachers' awareness of their own implicit theories
   would enable them to be more reflective in their professional decision
   making. Teachers would be able to analyze their own thinking about
   curriculum materials, old and new, and their modes of curriculum
   interpretation would be enriched and more complex, yielding more of the
   educational potential of these materials. (p. 110)

Following a common theoretical position, Pope and Scott (1984) drew on the theoretical perspective of George Kelly in examining the importance of teachers' perspectives on the nature of knowledge. Based on a case study of students attending four different courses (two courses were for initial teacher training and two were for in-service teachers) in primary education throughout one academic year, teachers' views of knowledge, learning and implications for their practice of teaching were investigated through semi-structured interviews and observation. The result suggests that teachers' views on knowledge and theories of learning would affect practice in the classroom. It also indicates that student teachers could, if encouraged, become aware of their initial epistemologies. Elbaz (1983) uses five categories as the basis for the analysis of teacher thinking: knowledge of self, knowledge of the milieu of teaching, knowledge of subject matter, knowledge of curriculum development and knowledge of instruction.

Barnes (1992) writes: `Of central importance is the teacher's conception of what constitutes learning, which normally includes an implicit epistemology' (p.20). He suggests that it is useful to consider the ways in which teachers perceive and execute their work in terms of frames, which refer to the default settings of our daily lives. A frame is an outline scheme which, running ahead of experience, defines and guides it. Thus frames are value-laden and dynamic. There are five significant frames:

   (1) preconception, often implicit, about the nature of what they are
   teaching, and--for secondary specialists--about the subject they teach and
   how to interpret it; (2) preconceptions about learning and how it takes
   place, though modified by a view of what can be achieved in the classroom;
   (3) preconceptions about students (in general, and about the particular
   group being taught) that place limits upon what is thought to be useful or
   possible; (4) benefits about priorities and constraints inherent in the
   professional and institutional context; and (5) the nature of his or her
   overall commitment to teaching--vocational, professional,
   career-continuance. (p. 19)

If changes in teaching are to take place, the teachers' frames must change.

Research on teacher thinking has grown rapidly since the 1980s and now constitutes a substantial area of inquiry in explorations of the nature of teaching. All the aforementioned studies reveal the assumption that implies a view of teachers as active agents who attach subjective meanings to what they know and think of classroom practice (Tabulawa, 1998). Teachers' perspectives on teaching affect how teachers perceive the nature of the subject, the problems of instruction and how they adopt solutions or ways to improve instruction. Furthermore teachers' views of knowledge have great impact on teachers' perspectives and their pedagogical practices.

Studying teachers' perspectives

If you ask teachers to tell you about their work, they usually describe very carefully what they have done or how they are planning to proceed in their classroom instruction. Teachers seldom speak about their perspectives underlying the action level. The aim of this study is to explore the perspectives held by computer studies teachers about teaching computer programming. In-depth interviews of 12 computer studies teachers (presented as T1, T2, ... T12 in the following sections) have been conducted in Hong Kong. Of the 12 teachers, all were trained computer studies teachers with 4 to 5 years of experience. On average each interview lasted for 45 to 60 minutes. The interviews were carried out using a semi-structured schedule. Questions were directed towards the areas of teaching objectives, contents, instructional methods, teaching focuses, students' learning difficulties and possible solutions for improvement. All interviews were tape-recorded and transcribed later.

Interview data were coded and analysed according to the constant comparative method of grounded theory. The label `grounded theory' means the discovery of theory from data (Glaser & Strauss, 1967). Constant comparative method allows the researcher simultaneously to code/decode and analyse data in order to develop theoretical concepts. By continually comparing specific incidents in the data set, the researcher refines these concepts, identifies their properties, explores their relationships to one another, and integrates them into coherent theory (Strauss, 1987).

The iterative coding of data led to the generation of specific themes and theoretical categories. The results of data analysis are discussed in the following sections, which are organised in five themes: (a) teaching objectives and content, (b) teaching foci, (c) teaching methods, (d) perceived factors affecting students' learning, and (e) instructional improvement.

Teaching objectives and content

The teaching objectives of the 12 teachers tended to fall into 2 categories: exam-oriented objectives and developmental objectives. The teachers with exam-oriented objectives were much concerned about the Hong Kong Certificate Examination of Education (HKCEE). These teachers were keen to help their students prepare for the public examination, which takes place by the end of their Secondary 5.

   For the objective, I would like my students to catch up with the HKCEE
   level. For the content, I follow the HKCEE syllabus. (T6)

   [My objective is] to help my students to achieve the HKCEE standard. (T7)

   The objective is quite exam-oriented. I follow the syllabus of HKCEE in my
   teaching. (T9)

Compared with exam-oriented objectives, developmental objectives are more profound and beneficial to students in the long-run. This is because they are about fostering a desirable further development of students' ability or a positive change of students' behaviour. These boost the personal qualities of students. For 0instance, among the 12 interviewees, some teachers with developmental objectives attached their teaching agendas to cultivating a computer-using habit, helping students to apply computers in daily life problems or developing students' logical thinking skills.

   There is another objective, that is, I hope my students will continue using
   and learning computer after graduation. (T6) [I would like] to enable my
   students to solve their problems in daily life by computers. Indeed, the
   main objective is `problem solving', that is, it is hoped that my students
   are able to write programs in order to solve their problems in daily life.
   (T3)

   I would like [to help] my students to develop their logical thinking skill.
(T10)

When mentioning content, most teachers (both with exam-oriented objectives and developmental objectives) instantly referred to the HKCEE (the public examination) syllabus or the textbook.

   I follow the textbook and teach my students computer language. I also teach
   them how to use computers to solve the problems in daily life. (T2)

   Concerning the content, my teaching covers the syllabus of HKCEE. (T3)

   I follow the syllabus of HKCEE in my teaching. (T5)

   I teach what the HKCEE pass papers have included. I have found the cases in
   the pass papers are quite simple. If I find a topic which is commonly being
   asked in the pass papers, for example, electronic data processing, I will
   put my teaching focus on that topic. (T11)

Teaching foci

The teaching focus is defined as the primary concern of teachers in their teaching plans. The focus of T1 and T2 was that their students can understand how computers work and the focus of T10 was the attainment of students' logical thinking ability.

   My primary concern is to enable the students to understand how computers
   function. This is because computers function in a way which is very
   different from human's. So, I have to teach them the differences first
   before teaching them how to use computer programming to complete tasks.
   (T1)

   I have to teach my students how computers work, that is, input, process and
   output. (T2)

   Therefore, I think the focus is on the development of my students' logical
   thinking ability. I think it is useless if students' logical thinking skill
   cannot be trained when they are learning Pascal, as they can apply this
   skill on studying other programming language or other subjects. This skill
   can be used on learning other subjects. (T10)

Many teachers showed hesitation when asked about their teaching foci. They either did not have a clear teaching focus or they did not think they could achieve their focus. For example, T10 said, `I would like to help them to develop their logical thinking ability, but I think it is hard to achieve that within the two lessons per cycle and under the current curriculum'. T6 said, `I have no specific focus' just before he told the interviewer that his focus was on HKCEE.

Teachers tended to complain that time constraint, HKCEE and the school curriculum did not allow room for them to achieve their teaching foci. T5 even put his teaching focus on an area which he did not think fit. In fact, he had put his focus on syntax, but he desired not to do so.

   I do not want to put my focus on the syntax. However, in reality, there are
   many computer languages. Students may think the syntax of one computer
   language can also be applied on another computer language. There is not
   enough time to do what I want to focus on, that is, telling the students
   the best way of writing a program. This is because students have to spend
   much time on learning syntax ... They have to spend a month to understand
   the use of semi-colon. (T5)

Teaching methods

Two teaching methods emerged from the data analysis, namely teacher-dominated and subject-centred methods. Teachers who followed the teacher-dominated method `are serving the immediate needs of the dependent, authority centred, linear thinking students. The teacher would direct students' learning through textbook and lecture mode' (McBeath, 1995, p. 425). T3 used the teacher-dominated method in his teaching. He stated that his teaching method tended to be a primitive one. Most of his lesson was spent on teacher's talk. In other words, teacher dominated the classroom activities.

   For my teaching strategy, it tends to be a traditional one ... With the
   time constraint, I cannot put some interesting elements, such as games and
   activities into my teaching. In the lessons, most time is spent on my
   presentation. In other words, my teaching tends to be a traditional one ...
   I think under such a lot of constraints, this is a way to teach ... If
   there is more time, I will use another teaching approach. (T3)

Teachers who taught with the subject-centred method `are providing more information and use a greater variety of presentation method. The responsibility for learning is placed upon the students, while the teacher primarily provides opportunities for learning to take place' (McBeath, 1995, p. 425). Some teachers adopted this subject-centred method in their teaching. They illustrated with examples, metaphors, pictures or games in their instruction. Apart from textbooks, they also used notes, laboratory sheets and supplementary exercises as teaching materials. T6 and T7 used mathematical examples to promote students to correlate thinking in computer programming.

   It is too hard for them. These programs are very long. Most F.4/5 students
   are not able to write these programs. So I use very simple examples, like
   examples taken from mathematics to teach them. (T6)

   There are mathematical examples. It is because my students' standard cannot
   write programs to solve daily life problems. However, they can do some
   programs related to mathematics, for example, solving two equations,
   finding the prime numbers or solving quadratic equations. I can find many
   good mathematical examples. (T7)

T8 and T1 enriched the lessons with metaphors, pictures or games to stimulate students' understanding of concepts.

   I teach them the basic concepts first. It is not suitable to teach them a
   program at the beginning. I teach them the concepts with metaphors and
   pictures and then lead them to understand the programs. Otherwise, the
   students cannot understand how the programs run. (T8)

   So when I teach computer programming, I have to find a way to relate the
   topic to daily life first. Sometimes we play games first, for example,
   playing `Matchbox Computer'. It is hoped that after playing a game which is
   related to the topic, the students will understand why the computers
   function in such a way. (T1)

T4 asked his students to solve real-life problems.

   I give them some real-life problems to solve. For example, the use of
   computer programs in supermarkets, that is, the use of bar-code readers or
   other devices. (T4)

None of the 12 teachers maintained the textbook-only method. They put effort into preparing `home-made' materials such as notes, laboratory sheets and exercises:

   I take the examples from the textbooks and simplify them. This is because
   the examples in the textbooks are quite difficult for my students and they
   do not like reading the textbooks. We use notes as the teaching material.
   For the 2-year syllabus, we have 80-90 sets of notes. (T4)

   Indeed, I have designed some lab sheets which indicate the steps of
   designing the programs. The students work with these lab sheets in the
   computer lab. (T5)

   I teach through examples. I explain the program and then give an example to
   the students. Later on, I ask them to do some `fill in the blanks'
   exercises. (T7)

Perceived factors affecting students' learning

Several factors were found to be affecting students' learning progress. Generally speaking, teachers perceived factors ranging from students' learning motivation, students' background to exposure of computers outside schools. Whether students inherently had a curiosity in learning the subject affected their learning heavily. Some students worked harder because they had intrinsic interests in computers. `If they (students) are interested in computer, they work hard on this subject' (T6). On the other hand, motivating students who had no interest in studying computers was an issue, too.

   Some students have no learning motivation. They think that they are being
   forced to study this subject. They think what they have to learn in
   computer, for example, syntax, is very hard. For this kind of students, I
   find no way to motivate them. (T11)

Most teachers thought that the lack of relevance of the subject content for daily life impeded students' motivation.

   They (the students) think they do not need to use Pascal programming when
   they are using computers at home. They do not need to use Pascal
   programming to help them to solve problems. For example, if they want to do
   word processing, there is application software to help them to do so. If
   they want to do calculation, they can use spreadsheets. If they want to
   draw patterns, there is graphic software to help them. They think that they
   cannot apply what they have learnt in the subjects in daily life. They
   think they learn this subject for examination only. It is useless to their
   daily life. It cannot meet their expectation. They expect the content of
   computer would be something like the areas cover in AS level computer. They
   expect to learn Microsoft word, PowerPoint and Excel. Therefore, after they
   have realized that their expectation is wrong, they have strong negative
   feeling (painful). (T3)

Examination-oriented and result-driven students were more hardworking in studying the subject. `For the students who are exam-oriented, they are more willing to spend time on studying Computer and their results in this subject are better' (T10).

Apart from students' interest, teachers considered students' ability as another major factor that was an impetus for student learning.

   It is their ability that affects their learning. If they are weak in both
   computer and mathematics, they are unable to understand the subject
   content, though they are interested in it. (T2)

From their experience, teachers found that the science students were more capable of learning computer programming than arts students.

   Yes. At least the science students are better in mathematics. They have
   more experience in understanding equations and formulae which can be used
   as the examples in computer. On the other hand, arts subjects, for example,
   history, are hard to relate to computer. (T6)

Some teachers related students' mathematics ability positively with computer programming.

   If the students are better in Mathematics, they tend to have higher ability
   in designing algorithms. (T7)

   This is because this subject is closely related to mathematics, especially
   the computer programming. There are many calculations and formulae
   involved. (T11)

Teachers stated that students with previous knowledge in LOGO were generally quicker learners of the subject.

   Personally, I think it is easier for the students who have learnt Logo
   before to understand computer. For example, they know how computers work
   (sentence by sentence or statement by statement). They understand that each
   statement carries a command. For the students who do not have this idea, it
   is very difficult for them to understand a program which has 5-6
   statements. They think the statements are related to each other. (T1)

   Computer is a very special (difficult) subject for the students, especially
   for the students who have not learnt Logo before, as they know nothing
   related to the subject content. (T10)

Instructional improvement

Patterns of 12 teachers, as summarised in the previous sections, indicate that difficulties were prevalent in programming instruction. Given time constraints, the examination syllabus, absence of teaching focus, students' diminishing interests in the subject and relatively low academic ability, there are no short-cuts to teaching computer programming. However difficult, teachers were still working out solutions in order to improve their teaching practices. Just like one of the steps, the problem-solving step, in computer programming, `debugging' is never easy. Teachers had to resort to balancing time, resources, syllabus and provisional strategies to remedy difficulties.

From the interview, `survival' teaching was found to be very common among the 12 teachers. Teachers faced various constraints and limitations. As a result, their resolutions turned to provisional remedies. Their main target was to get the whole syllabus covered. Slight and passive improvement was settled on. Teachers claimed they had no choice and such survival teaching seemed to be the only way. Some of them even showed disappointment and uneasy feeling on achieving a balance between survival teaching and quality teaching. As T1 said, `I try hard to teach my students through games, but the students with low ability sometimes make me disappointed. And until now, I still cannot find an effective way to help them.' Table 1 summarises the perceived problems and respective solutions of the 12 teachers.

Table 1 Teachers' perceived problems and solutions

Perceived problems         Solutions

Difficult subject          Relate the subject content to daily life (T1)
content                    Begin with something easier (T2)
                           Slow the teaching pace (T8)
                           Go through all steps of problem solving with
                             students (T11)
Students' low ability      Stimulate students' thinking by games (T1)
                           Go through all steps of problem solving with
                             students (T6)
Time constraints,          Adopt teacher talk, the `traditional teaching
limited lesson time          method', as the major classroom
                             activities (T3)
                           Offer supplementary lessons after school
                             (T10)
Students' low logical      Encourage students to think verbally (T3)
thinking ability

Students' low ability      Go through all the steps of problem solving
Student cannot write a       with students (T6)
program                    Ask students to memorise some standard
                             structure (T7)
Low applicability of the   Talk about something extra if there is
subject content in daily     time (T9)
life

Students' low learning     Provide more daily life examples (T12)
motivation

Connectionist view of understanding

Over the years, two predominant computational theories of mind have emerged from the study of artificial intelligence (AI), namely, the symbolic model fashioned after the digital computer, and the connectionist model modelled after the human brain (Korf, 1991). In this section, I will discuss neither the debate between these two models nor their epistemological foundations though they have different implications for how the brain relates to mind and knowledge. However I will provide an introduction to the two models and discuss their relations to pedagogical practice.

The symbolic models, building on the widely accepted conception of the Physical Symbol Systems Hypothesis (Newell & Simon, 1976), view symbol systems as machines, possibly human, that process symbolic structures through time. In this model, the theories take the form of computer programs executed on a standard symbolic machine. Early AI efforts in this model focused primarily on problem-solving tasks (Newell & Simon, 1972) such as playing games and proving theorems in symbolic logic. As more complex tasks began to be tackled, however, the emphasis shifted from general algorithms to rule-based expert systems that performed at expert levels in particular domains by virtue of having encoded very large domain-specific knowledge bases. In symbolic model, concepts are represented by symbols, knowledge is conceived of as objects in individual minds (Bereiter & Scardamalia, 1996). Learning in such mind-as-container metaphor is a linear problem-solving process, in which symbols are manipulated in a sequence of steps in two directions, forward from a set of givens or backward from a goal. This metaphor suggests that when we learn something, there is a sentence in our head representing that knowledge (Strauss & Quinn, 1997).

The connectionist models, often referred to as artificial neural networks, have a long history, dating back to the early 1950s when they were called parallel-processing systems or self-organising systems (Boden, 1988). In these models, knowledge is not represented by symbols linked together in sentences, but by simple processing units arranged in layers. An artificial neural network consists of a very large number of individual elements (processing units), modelled after neurons, each connected to a large number of neighbouring elements, and each computing a very simple function, such as a weighted sum of its inputs. Connectionist models presume that cognitive functioning can be explained by collections of processing units that operate in this way. Concepts to be learned in connectionist systems are represented as patterns of activity over a set of processing units. Propositions are represented by patterns of activation over many units. `All kinds of experiences, including those provided by texts and other symbolic materials, can be represented in a connectionist model' (Strauss & Quinn, 1997, p. 53). The knowledge in connectionist systems is in the connections (Rumelhart, 1989). Thus the connectionist mind can be knowledgeable without containing knowledge (Bereiter, 1991). This is in line with the advocacy of Minsky (1990), `We will have powerful, modular learning machines equipped with no knowledge at all' (p. 240).

The results of studying teachers' perspectives discussed in the previous sections reveal that most teachers take knowledge as objects and treat understanding as a characteristic of something in students' minds. These pedagogical assumptions are fully consistent with the mind-as-container metaphor. `Learning with understanding' is one of the catch phrases of educational reform nowadays. However a particularly complex problem is how to generate changes aiming at a pedagogy of understanding, that is, teaching so that all students can come to understand ideas deeply enough to apply them to novel situations and to perform proficiently in their own right (Darling-Hammond, 1998).

There are indeed important kinds of learning that do not readily fit the container metaphor and the connectionist model provides an alternative way of thinking about knowledge and understanding. Strauss and Quinn (1997) argue that connectionist models seem better than symbolic models:

   at simulating how people build up cultural schemas and use them to extract
   meaning from events ... It does give us a way of starting to think about
   the issue of learning, one that could be built upon and modified as
   necessary to explain various forms of culture acquisition. (pp.56-57)

Bereiter and Scardamalia (1996) argue that the connectionist view is a more enlightened theory than the account of understanding provided by the symbolic model. The connectionist view of understanding is based on commonsense criteria. First of all, there is no privileged way that ought to be expected of student. In judging a student's understanding of x, the following criteria are applied: (a) how intelligently the person acts with respect to x, (b) his or her ability to explain whatever is judged to be in need of explaining in x, and (c) awareness of and interest in doing something about shortcomings of the first two criteria. An immediate benefit of this approach to understanding is that `it permits liberality without total collapse into relativism' (Bereiter & Scardamalia, 1996, p. 499).

Pedagogical implications of connectionist approach

The change of teachers' perspectives is the first major breakthrough for instructional improvement. It is evident that the connectionist view of understanding provides an alternative way of thinking that breaks the taken-for-granted perspectives of teachers. Following the discussion of the previous section, I describe four implications for pedagogical change in teaching computer programming.

Given time, resource constraints, students' low ability and motivation, as well as complicated subject content, teachers fell into the dichotomies of whether to stick to the syllabus or to shift to the interesting and lively presentation materials; whether to focus on the public examination or to cultivate students' logical thinking ability; whether to adopt short-term survival teaching or to provide long-term quality instruction. All problems are deep-rooted. `In the connectionist view of mind, there is no mental content to talk about. There are only abilities and dispositions' (Bereiter & Scardamalia, 1996, p. 499). The first pedagogical change, supported by the research in the 1980s, is to adopt a specific teaching focus, a focus geared towards the cognitive development of students, for computer programming. The content should go beyond the public examination syllabus, and emphasise thinking processes and transfer of learning. This focus is also described as one very crucial objective in the HKCEE syllabus. This change is obviously linked to the second change.

Knowledge cannot simply be transferred by means of words, and learning is the product of self-organisation (von Glasersfeld, 1998). A connectionist mind is a self-organising mind. Thus the second pedagogical change is to move the transmission-oriented pedagogy (teacher-dominated or subject-centred) to student-centred and self-organising learning. Traditionally, teacher-dominated or subject-centred methods refer to the use of learning activities which typically involve direct teacher involvement, participation and interaction with pupils, in which, through the use of informing, describing, explaining, questioning, modelling, demonstrating, and coaching, the teacher transmits the knowledge, understanding, skills and attitudes that he or she wishes to develop (Kyriacou, 1995). This teaching-as-telling method is never, on its own, sufficient to ensure deep understanding. However, the student-centred approach builds on the idea that students learn best when engrossed in the topic and motivated to seek out new knowledge and skills because they need them in order to solve the problem at hand (Norman & Spohrer, 1996).

Third, collaborative project work should be encouraged. Project work is a complex cognitive and metacognitive process that requires both hands-on and minds-on learning, that is, concrete subject-based knowledge and abstract high-order thinking skills. Project-based learning is action-oriented and focuses on doing something rather than learning about something (Moursund, 1999). In projects, students engage in a complex process of learning, inquiry and knowledge construction. The result is an artifact, a product of student knowledge that can be shared and critiqued. The resultant artifact becomes a product for review and reflection (Dede, 1998). Projects are contextualised in issues and topics related to the real world and, properly chosen, the context can situate learning to promote authentic learning and improve transfer. In project learning, students construct knowledge by manipulating and extending ideas and information. Moreover, in project work, students share ideas and exchange information as a group and improve knowledge collaboratively. This change is particularly important in learning and teaching computer programming.

Connectionist systems `always start with some initial constraints only, and gradually acquire the rest of their knowledge through exposure to a variety of specific examples and repeated correction of inferences about those instances' (Strauss & Quinn, 1997, p. 57). Thus, lastly, students need guided instruction and exploration to insure they acquire the concepts and strategies underlying the diverse activities of computer programming. This emphasis is in line with the research on teaching computer programming since the 1980s.

Conclusion

Research since the 1980s has presented quite varied opinion about how to teach computer programming. The development of simulation, visualisation or animation tools has not fully solved the problems. The tension between the nature of teachers' work and the time, energy and expertise necessary to improve pedagogical practice is a constant dilemma for teachers. What are the factors that contribute to effective teaching? How can teachers improve their teaching? These questions are significant not only for computer studies teachers but for teachers of all subjects. Understanding teachers' perspectives on pedagogical practice, I believe, is important and it shows the direction of instructional improvement.

In this article, interviews with 12 computer studies teachers in Hong Kong were presented to outline the scene of teachers' perspectives. Results of the data analysis can be summarised as follows: (a) examination is the focus of teaching in schools, (b) teaching materials and textbooks are essential in instruction, (c) teaching-as-telling is the popular teaching method, and (d) learning means objects of knowledge are represented and stored in students' minds. Thus the improvement of teaching is limited to the emphasis on method (Berry, 1999), such as providing more teaching materials or talking in different ways. These results reflect that teachers have a taken-for-granted perspective rooted in the mind-as-container metaphor. I have discussed the connectionist view as providing an alternative way of thinking about knowledge and understanding which leads to better insight on teaching and facilitates pedagogical change. Furthermore I have discussed four implications of the connectionist view in relation to pedagogical change in teaching computer programming. Nevertheless further advances in epistemology will continue to influence work in psychology and education (Case, 1996). Teachers and other people involved in the educational arena need to be aware of the epistemological impact on educational and pedagogical practice.

Keywords

cognitive processes            learning strategies   student attitudes
computer assisted teaching     programming           teaching process

References

Barnes, D. (1992). The significance of teachers' frames for teaching. In T. Russell & H. Munby (Eds.), Teachers and teaching: From classroom to reflection (pp.9-32). London: Falmer Press.

Ben-Peretz, M. (1984). Kelly's theory of personal constructs as a paradigm for investigating teacher thinking. In R. Halkes & J.K. Olson (Eds.), Teacher thinking: A new perspective on persisting problems in education (pp. 103-111). Lisse: Swets & Zeitlinger.

Bereiter, C. & Scardamalia, M. (1996). Rethinking learning. In D.R. Olson & N. Torrance (Eds.), The handbook of education and human development: New models of learning, teaching and schooling (pp. 485-513). Cambridge, MA: Basil Blackwell.

Bereiter, C. (1991). Implications of connectionism for thinking about rules. Educational Researcher, 20, 10-16.

Berry, K. (1999). Destabilizing educational thought and practice: Post-formal pedagogy. In S.R. Steinberg, J.L. Kincheloe, & P.H. Hinchey (Eds.), The post-formal reader: Cognition and education (pp. 331-348). New York: Falmer Press.

Boden, M.A. (1988). Computer models of mind. Cambridge: Cambridge University Press.

Case, R. (1996). Changing views of knowledge and their impact on educational research and practice. In D.R. Olson & N. Torrance (Eds.), The handbook of education and human development: New models of learning, teaching and schooling (pp. 75-99). Cambridge, MA: Blackwell.

Darling-Hammond, L. (1998). Policy and change: Getting beyond bureaucracy. In A. Hargreaves et al. (Eds.), International handbook of educational change (pp. 642-667). Dordrecht: Kluwer Academic Publishers.

Dede, C. (Ed.). (1998). Association for Supervision and Curriculum Development 1998 Yearbook: Learning with technology. Alexandria, VA: ASCD.

Elbaz, F. (1983). Teacher thinking: A study of practical knowledge. London: Croom Helm.

Glaser, B. & Strauss, A. (1967). The discovery of grounded theory. Chicago: Aldine.

Guimaraes, M.A.M. & de Lucena, C.J.P. (1995). A system environment for teaching introductory algorithms. SIGCUE OUTLOOK, 23(3), 2-11.

Jonassen, D. H. (1996). Computers in the classroom. Englewood Cliffs, NJ: Prentice-Hall.

Kann, C., Lindeman, R. W., & Heller, R. (1997). Integrating algorithm animation into a learning environment. Computers Education, 28(4), 223-228.

Korf, R.E. (1991). Artificial intelligence as information science. Information Sciences, No. 57-58, pp. 131-134.

Kyriacou, C. (1995). Direct teaching. In C. Desforges (Ed.), An introduction to teaching (pp. 118-131). Oxford: Basil Blackwell.

Mayer, R.E. (1988). Introduction to research on teaching and learning computer programming. In R.E. Mayer (Ed.), Teaching and learning computing programming (pp. 1-12). Hillsdale, NJ: Lawrence Erlbaum Associates.

McBeath, R. (1995). Facilitating transitions in learning and teaching. In Improving university teaching: Twentieth international conference, 10-13 July 1995, Hong Kong, pp. 420-428.

Minsky, M. (1990). Logical vs. analogical or symoblic vs. connectionist or neat vs. scruffy. In P.H. Winston & S.A. Shellard (Eds.), Artificial intelligence at MIT: Expanding frontiers (pp. 219-243). Cambridge, MA: MIT Press.

Moursund, D. (1999). Project-based learning using information technology. Eugene, OR: International Society for Technology in Education.

Newell, A. & Simon, H. (1976). Computer science as empirical inquiry: Symbols and search. Communication of the ACM, 19(3), 113-126.

Newell, A., & Simon, H.A. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-Hall.

Nickerson, R.S., Perkins, D.N., & Smith, E.E. (1985). The teaching of thinking. Hillsdale, NJ: Lawrence Erlbaum Associates.

Norman, D.A. & Spohrer, J.C. (1996). Learner-centered education. Communication of ACM, 39(4), 24-27.

Papert, S. (1980). Mindstorms: Children, computers and powerful ideas. New York: Basic Books.

Perkins, D.N., Schwartz, S., & Simmons, R. (1988). Instructional strategies for the problems of novice programmers. In R.E. Mayer (Ed.), Teaching and learning computing programming (pp. 153-178). Hillsdale, NJ: Lawrence Erlbaum.

Pope, M.L. & Scott, E.M. (1984). Teachers' epistemology and practice. In R. Halkes & J.K. Olson (Eds.), Teacher thinking: A new perspective on persisting problems in education (pp. 112-122). Lisse: Swets & Zeitlinger.

Rumelhart, D.E. (1989). The architecture of mind: A connectionist approach. In M.I. Posner (Ed.), Foundation of cognitive science (pp. 133-159). Cambridge, MA: MIT Press.

Sloane, K.D. & Linn, M.C., (1988). Instructional conditions in Pascal programming classes. In R.E. Mayer (Ed.), Teaching and learning computing

programming (pp. 207-236). Hillsdale, NJ: Lawrence Erlbaum Associates.

Smith, P.A. & Webb, G.I. (2000). The efficacy of a low-level program visualization tool for teaching programming concepts to novice C programmers. Journal of Educational Computing Research, 22(2), 187-215.

Strauss, A. (1987). Qualitative analysis for social scientists. Cambridge: Cambridge University Press.

Strauss, C. & Quinn, N. (1997). A cognitive theory of cultural meaning. Cambridge, MA: Cambridge University Press.

Tabulawa, R. (1998). Teachers' perspectives on classroom practice in Botswana: Implications for pedagogical change. Qualitative Studies in Education, 11(2), 249-268.

Thomas, R.A. & Upah, S. C. Jr. (1996). Give programming instruction a chance. Journal of Research on Computing in Education, 29(1), 96-107.

von Glasersfeld, E. (1998). Cognition, construction of knowledge, and teaching. In M.R. Matthews (Ed.), Constructivism in science education: A philosophical examination (pp.11-30). Dordrecht: Kluwer Academic Publishers.

Webb, N.M. & Lewis, S. (1988). The social context of learning computer programming. In R.E. Mayer (Ed.), Teaching and learning computing programming (pp. 179-206). Hillsdale, NJ: Lawrence Erlbaum.

Young, R.E.A. (1981). Study of teacher epistemologies. Australian Journal of Education, 25(2), 194-208.

Yuen, H.K. (1999). Traffic light simulation system: A learner-centered multimedia tool for learning algorithm design. In W.S. Chow (Ed.), Multimedia information systems in practice (pp. 244-250). Singapore: Springer-Verlag.

Acknowledgements

The research reported in this paper was funded by the Committee on Research and Conference Grants of the University of Hong Kong.

Allan H.K. Yuen is an Assistant Professor in the Department of Curriculum Studies and a Deputy Director of the Centre for Information Technology in School and Teacher Education, the University of Hong Kong, Pokfulam Road, Hong Kong. Email: [e-mail]