Using Kintsch's Text Comprehension Model to Identify CS Students’

Conceptions and Misconceptions

Christina Kyriakou

, Agoritsa Gogoulou

and Maria Grigoriadou

Department of Informatics and Telecommunications, National and Kapodistrian University of Athens, Athens, Greece

Keywords: Introductory Computer Science Courses, Data Storing and Manipulation, Main Memory, Program Execution,

Kintsch's Construction-integration Text-comprehension Model, Conceptions, Misconceptions.

Abstract: Recent research attempts to figure out the hard points of Computer Science (CS) discipline curricula that seem

to trouble students. The presented work focuses on assessing first-year CS students’ knowledge on issues

related to fundamental computer architecture concerning main memory organization, operation, and role on

program execution. Formulating questions based on Kintsch's text comprehension model theory is a

promising diagnostic tool to identify students’ conceptions and misconceptions. The paper discusses the

Kintsch model as the basis for formulating meaningful questions, analyzes students’ answers, and attempts to

categorize and explain the revealed misconceptions. The emerging misconceptions may be helpful for

effective learning design and appropriate educational material.

1 INTRODUCTION

The in-depth learning of the internal computer

operations is of great significance not only for their

value but additionally because when students

internalize successfully such lower-level information,

they gain coherent knowledge of computer operation

(Clements, 2000). The non-observable nature of

internal computer operations renders learning of

computer architecture and organization a problematic

issue.

Efficient instruction should consider students'

difficulties in comprehending the concepts being

taught. This notion is supported by the results of

extensive research of cognitivists and educators. For

more than four decades, researchers have explored

how students learn and what affects their

understanding (Bransford, Brown, & Cocking, 2000;

Vosniadou, 2001). All findings converge on the great

significance of prior constructed students' mental

models (Johnson-Laird 1983; Ben-Ari, 2001) or prior

knowledge frameworks (Davis, Maher, and

Noddings, 1990) for the acquisition and assimilation

of new knowledge (Ausubel, 1968; Posner et al.

1982; Vosniadou, & Brewer, 1987). When students

are exposed to new knowledge or situations, they use

https://orcid.org/0000-0002-6835-4783

https://orcid.org/0000-0001-9025-0254

their pre-existing mental models to understand and

explain the new concepts. The scientific validity of

the previous mental models or knowledge

frameworks is a prerequisite in incorporating the new

knowledge without misunderstandings. Often

students' prior empirical representations of the “real

world” are not compatible with scientific

representations (Von Glaserfeld, 1995). Under these

conditions, misconceptions may be created in

students' minds. Such misconceptions or, at best

synthetic models are created when students try to

incorporate new incompatible knowledge into an

existing knowledge structure (Vosniadou 1994; 2003;

2007) or when familiar terms are used in unfamiliar

contexts (Clancy, 2004). Misconceptions may be

identified by the student’s explanations on answering

questions or engaging in activities of understanding

relevant to the specific concepts. Mistakes, faulty or

incomplete answers may reflect what exists in a

student's mind and should be considered precious

feedback for the instruction and learning process.

Following the theories mentioned above and the

cοgnitive approaches to learning, during the last

decades, research has focused on the hard points of

Computer Science discipline curricula that seem to

trouble students either by their inherent complexity or

384

Kyriakou, C., Gogoulou, A. and Grigoriadou, M.

Using Kintsch’s Text Comprehension Model to Identify CS Students’ Conceptions and Misconceptions.

DOI: 10.5220/0011084000003182

In Proceedings of the 14th International Conference on Computer Supported Education (CSEDU 2022) - Volume 2, pages 384-394

ISBN: 978-989-758-562-3; ISSN: 2184-5026

by their interrelatedness with other complicated

concepts. The focus has been on introductory

programming topics and less on computer

architecture and organization topics (Herman, Loui,

& Zilles, 2009; 2010; 2011; Herman, Kaczmarczyk,

Loui, & Zilles, 2011).

Porter et al. (2013) created a preliminary test of

conceptual understanding for an upper-division

computer architecture course. They designed nine

high-level concept questions on topics including the

performance implications of deeper pipelines, the

roles of various cache components, and performance

analysis of single-cycle, multi-cycle, and pipelined

processors. The questions were administered to four

separate computer architecture courses at two

different institutions and were intended to be

correctly answered by any student passing the class.

Disappointingly, the results showed a large

discrepancy between what instructors thought

students were learning and what they had learned.

The scholars suggest a further inquiry into student

understanding of core course concepts in Computer

Architecture.

Grigoriadou and Kanidis (2001) investigated CS

students’ conceptions on computer cache-memory-

related topics with the aid of questionnaires and

interviews. The scholars identified the difficulties

several students had in understanding the

organization and operation of cache-memory. This

fact raised concerns to scholars about the depth of

students’ memory-related knowledge. Another work

of the same researchers (2002) on the conceptions of

secondary education students indicated a lack of

knowledge on computer memory hierarchy,

organization, and operation.

The findings and suggestions of the studies

mentioned above made the need for efficient

instruction of memory-related topics apparent to the

authors. The notion that the CS students should be

aware of how a computer operates and that

encompasses a deep knowledge about memory,

motivated authors to conduct a sequence of studies.

The ultimate aim of these studies was to support

students in their learning of memory-related topics.

The first step towards this aim was to investigate

students’ understanding of memory.

The preliminary study investigated the level of

understanding that first-year undergraduates of the

Department of Informatics and Telecommunications

of the National and Kapodistrian University of

Athens have on memory operations after attending an

introductory CS course. The authors collected and

analyzed students’ answers to open-ended questions

administered in the written final exams at the end of

the semester in four different academic years. More

specifically, the focus was on the concepts related to

the communication between main memory and CPU,

the special-purpose registers, and the machine

language program’s execution. The results

demonstrated that although most students can recall

basic knowledge on these topics, they have great

difficulty applying and combining this knowledge to

make inferences and give correct answers (Kyriakou

& Grigoriadou, 2016).

The present study is the sequel to the preliminary

research. Learning about the role of the main memory

and its architecture in implementing the von

Neumann model is also fundamental to understanding

computer operation. This idea motivated the authors

to continue the investigation of the students’

understanding focusing on these memory-related

topics.

This study was also conducted at the same

department on first-year students. The investigation is

based on Kintsch's Construction-Integration Model of

Text Comprehension. The presented work attempts to

contribute to the field of Computer Science Education

as it proposes an approach of formulating questions

in order to assess students’ conceptions and provides

a categorization of the identified conceptions/

misconceptions.

The rest of the paper is structured as follows.

Section Τwo outlines the underlying text

comprehension theory. In Section Three, the

empirical study is described. Afterward, the study’s

limitations and the implications of the research results

are discussed, and the paper concludes with further

research directions.

2 THEORY BACKGROUND

To investigate students' conceptions on the

combination of issues relevant to main memory and

program execution, a text comprehension model was

sought that relies on questions to assess the degree of

comprehension accomplishment. The theory of the

Construction-Integration model of text

comprehension seemed to fill our requirements.

(Kintsch, 1988). Kintsch's theory of Construction-

Integration model of text comprehension is Kintsch's

extension of his and van Dijk's pre-existed

comprehension models (Kintsch and van Dijk, 1978;

van Dijk and Kintsch, 1983) and is based on the

notion that the comprehension process is moderated

by individual differences, such as prior knowledge,

abilities, preferences, and strategies, mainly stressing

the role of previous knowledge.

Using Kintsch’s Text Comprehension Model to Identify CS Students’ Conceptions and Misconceptions

385

According to this cognitive model of discourse

understanding, a reader develops two distinct levels

of text representation during a text comprehension

process which the text base and situation models can

describe. The text base model corresponds to the

propositional representation of a text, both at the level

of the micro- and macrostructure. The situation model

corresponds to a representation of the text integrated

with other knowledge like events, actions, persons,

and generally the situation a text is about. “A situation

model may incorporate previous experiences, and

hence also previous text bases, regarding the same or

similar situations. At the same time, the model may

incorporate instantiations from more general

knowledge from semantic memory about such

situations.” (van Dijk and Kintsch, 1983).

Several types of measures can be used to assess

the level of a reader’s text understanding, i.e., to

evaluate the extent to which the text base and

situation models have been developed (McNamara,

Kintsch, Songer, and Kintsch, 1996).

These include text base measures which are free-

recall and text-based questions. Free-recall is when

the reader is asked to recall the text without

explanations. While text-based questions are based

entirely on the text content, and the necessary

information for their answer is stated in the original

text and requires only a single sentence from it.

Additionally, the other being the situation model

measures which include problem-solving questions,

elaborative-inference questions, bridging-inference

questions, and the sorting task. In problem-solving

questions, a reader is asked to apply information from

the text to a novel situation, which requires a well-

formed situation model. Inference questions need

inferencing of some kind or analytic reasoning. More

specifically, when a reader is asked elaborative-

inference questions, he should combine information

from the text with outside knowledge, which can be

achieved even with a surface situational

understanding. On the other hand, in bridging-

inference questions, the necessary information is

stated in two or more sentences in the text. The reader

should combine them and infer their unstated

relations to answer the question, which requires a

deeper situational understanding and a solid text base.

In a sorting task, a reader is asked to relate the

concepts presented in the text, which reflects, at least

in part, the situation model.

3 THE EMPIRICAL STUDY

The present study aimed to apply Kintsch's theory of

the Construction-Integration model of text

comprehension and analyze first-year CS students'

answers to assess students’ knowledge on main

memory organization, operation, and role during

program execution and derive, record, and identify

possible students' misconceptions. The main research

question of the empirical study is:

Are the open-ended bridging-inference questions

formulated based on Kintsch's theory of the

Construction-Integration model of text

comprehension, effective in investigating students’

conceptions?

The study was conducted at the Department of

Informatics and Telecommunications of the National

and Kapodistrian University of Athens in the

undergraduate course “Introduction to Informatics

and Telecommunications.” The course is delivered

through two-hour lectures weekly during the first

semester, and students take a written examination at

the end. The topics covered are i) Data Storage, ii)

Data Manipulation, iii) Operating Systems, iv)

Networking and the Internet, and v) Security. The

topics involved in this study are Data Storage and

Data Manipulation. Data Storage covers issues

associated with data representation and storage within

a computer like bits storage, main memory

organization and capacity, mass storage,

representation of information as bit patterns, and

binary system. Data Manipulation covers themes

related to the way a computer manipulates data and

communicates with peripheral devices, the basics of

computer architecture, and includes the way

computers are programmed in machine language

instructions (Brookshear, 2009).

3.1 Educational Material

The educational material used in this introductory

course is in written (text) and electronic form. Two

course books support students’ learning (Brookshear,

2009; Forouzan, 2003). Electronic material is based

on the course books. It includes lecture notes,

delivered through an LMS (called “eclass”) and

additional educational electronic material, in activity

form, offered through SCALE, a web-based activity-

oriented learning environment. SCALE supports the

knowledge construction process by engaging students

in activities that address specific learning goals in the

context of fundamental concepts and by providing

tutoring and informative feedback (Gogoulou et al.,

2007).

CSEDU 2022 - 14th International Conference on Computer Supported Education

386

In the course framework, students' engagement

with the activities in the SCALE is optional and

motivated by a pre-defined offered reward of up to

one grade out of ten towards the students’ final course

performance.

3.2 Method

To investigate students’ conceptions on the role and

organization of the main memory in program

execution (von Neumann model implementation),

written data were collected from two different groups

of students using two diagnostic tools (open-ended

questions and small-group discussions). A set of

open-ended questions were formulated based on

Kintsch’s text-comprehension theory. In particular,

the bringing-inference type of question was used to

assess students’ text understanding. Then a

categorization scheme was defined using a

methodology based on Thematic Analysis, as Braun

and Clarke (2006) outlined. Inter-coder reliability of

the categorization was assessed, and the

categorization was then analyzed and interpreted.

3.3 Participants and Data Collection

In the study, a total of 360 students participated.

Group A of 270 students participated in the

compulsory final course exams in the academic year

2011-2012. 84% were first-year, 16% were second-

year, and seniors. Their responses to a set of questions

related to computer memory are the first dataset

(Dataset1) that was collected and analyzed. Group B

of 90 students took part in an optional mid-term

course project in 2016-2017. 92% were first-year, 8%

were second-year, and seniors. Their participation

was motivated by a one-grade bonus reward (out of

ten). The project reviewed the topics of Data Storage

and Manipulation after the relevant lectures. The

course project provided two datasets to the study

(Dataset 2 & 3). Dataset2 consisted of students’

written responses to two open-ended questions. As a

follow-up method, the small-group discussions

provided detailed students’ opinions on one of the

questions. Students’ ideas constitute the Dataset3 of

the study.

3.4 Bridging-Inference Questions

Both experimental groups answered bridging-

inference questions based on Kintsch's text

comprehension model. To answer the questions

requires combining knowledge from Data Storage

and Data Manipulation and then concluding with

inferences. To answer the questions, students should

have a good text-based understanding of the topics:

1) Data Storage (the organization and operation of the

main memory and hard disk); 2) Data Manipulation

(CPU’s architecture and operations, machine-

language program’s execution), and 3) the von

Neumann architecture.

Indicative questions are:

• Many things have changed since the first PCs

were made, such as size, electronic circuitry,

software, applications. What do you think has

remained the same?

• A student with a Simple Computer wrote the

program, executed it, and then stored it on the

hard disk before shutting down the computer.

a) Describe the format of the program stored

on the hard disk.

b) Another day, when the student wanted to

open and edit the program, what actions

would the Simple Computer take?

• Please comment on the correctness of the

following suggestion and justify your answer:

“A computer may run using the hard disk when

the main memory is missing.”

Here follows the analysis of the latter question, which

was articulated to encompass the following research

questions (RQ):

RQ1 What do the students believe about the role of

main memory on a program execution?

RQ2 How do the students relate the main memory

organization (ordered directly-accessed

addressable cells/bytes) with the

implementation of the sequential execution of a

program’s instructions (as defined in the von

Neumann model)?

RQ3 What is the students' knowledge of main and

secondary memory data storage?

Moreover, to answer this question, students should be

able to infer the text-unstated relations from the

mentioned above topics and to describe: 1) the

reasons why the main memory and the hard disk have

the particular structure and their role on a program’s

execution; 2) the communication between the main

memory and CPU; 3) the direct access of the main

memory contents on cells/byte-level through their

addresses and the significance of this capability for

the sequential instructions’ execution as is specified

in the von Neumann’s model, 4) the inability to

access data/instructions on byte-level when stored in

a hard disc. These inferences depend on situational

understanding.

Using Kintsch’s Text Comprehension Model to Identify CS Students’ Conceptions and Misconceptions

387

By linking all this information, students should

conclude that, as the von Neumann architecture

defines, a computer cannot run when the main

memory is missing. The correct justification is that,

contrary to the hard disk structure, only the main

memory’s organization in ordered addressable

cells/bytes enables the distinct storage and fetching of

a program’s instructions which is essential for the

machine cycle (fetching, decoding, and execution).

3.5 Data Analysis

3.5.1 Dataset1 Analysis

Group A students’ answers (Dataset1) were filed,

manually content-analyzed, and classified according

to the correctness and justification.

Each response was initially classified as correct,

incorrect, and partially correct by one of the authors.

75% of the students failed to make the proper

inferences. These students answered the question

either: 1) incorrectly, arguing that a computer may

run using only the hard disk when the main memory

is missing, or 2) partially correct/incompletely,

arguing that a computer cannot run without the main

memory, but giving incomplete or wrong reasoning.

Then the correct answers were removed from

Dataset1. Two of the authors collaborated and

processed the Dataset1 answers to derive students’

conceptions. They went through the answers and

coded them using Thematic Analysis methodology.

The method is data-driven. Many students’ responses

included several justification cases. Each different

justification case that was observed defined a new

category. Categories were modified and refined after

repeated reading of the data. Each case was recorded

and coded with a small phrase, i.e., “ROM,” “hard

disk synchronization issue,” “RAM volatility,” etc.

Then, for category validation purposes, the third

author and another experienced instructor (Inst) from

the Department specialized in relevant topics like

computer architecture, digital logic design, and digital

systems, worked independently. They coded 100

randomly chosen answers from Dataset1, and the inter-

coder reliability was calculated (see Results section).

Lastly, only when a justification case occurred in

more than five student answers was it kept as

significant. The relative frequency of each case was

calculated, i.e., the percentage of students’ responses

that mentioned each justification case.

3.5.2 Dataset2 Analysis

In order to investigate the possible re-occurrence and

persistence of the students' conceptions about main

memory over time and under different conditions, the

study was extended to Group B students. Moreover,

the motivation was to ensure that students’

incomplete answers were due to their lack of

knowledge or misunderstandings and not due to a

lack of test time or just because they didn’t consider

it necessary to add more information to their

justification.

In particular, Group B students participated in a

mid-term optional review project on Data Storage and

Manipulation topics after the relevant lectures had

been delivered. The first part of the project consisted

of several activities in the WbLE supporting the

course. The activities included questions relevant to

the topics of the lectures. The students worked

individually on the activities. Two open-ended

bridging-inference questions were posed to Group B

students in this context. The previously described

question is one of the two.

Again Group B students’ answers (Dataset2) were

filed, manually content-analyzed, and classified

according to the correctness and justification.

Classification according to their correctness showed

that 82% of the students' answers were either wrong

or partially correct. After removing the right answers

from Dataset2, two authors followed the same

category-development procedure described in 3.5.1.

Afterward, the third author and the Inst working

in collaboration used a sample of 40 randomly-chosen

answers of Dataset2 and confirmed the categories.

Once again, the justification cases occurring in

less than five student answers were considered

infrequent and omitted.

3.5.3 Data Validation

For data validation, another input source of students’

conceptions was used. In the second part of the

optional course project, Group B students were asked

to discuss the open-ended bridging-inference

question described above in the “eclass” LMS. The

students were separated into small groups of 5-6

members (each member holding a different opinion).

The guidelines asked students: a) to express their

opinions about the open-ended question and agree or

disagree with their peers justifying their views, and b)

to elicit a team response to the question under

consideration.

This process provided insight into students’

thinking and conceptions because it required students

to explicitly articulate content in their own words.

Consequently, it served as an indicator of their

understanding or misunderstanding.

CSEDU 2022 - 14th International Conference on Computer Supported Education

388

Table 1: Outline of the most frequent justification cases of the incorrect and partially correct students answers.

Cases Type of answer / Description of justification case

Dataset1

(

)

Dataset2

(

)

Incorrect Answer: When the main memory is missing, a computer may operate slower

using only the hard disk because both are similar storage systems that work at different

speeds.

57% 23%

Partially Correct Answer: According to von Neumann's architecture, the main memory

is one of the fundamental computer subsystems. The hard disk communicates only with

the main memory, not the CPU.

17% 12%

Partially Correct Answer: A program should be in the main memory for its execution,

so the main memor

is necessar

for a com

uter's o

eration.

14% 23%

Partially Correct Answer: ROM is part of the main memory, so the computer cannot

start up if it is missing.

11% 41%

Partially Correct Answer: The main memory is volatile and used for data storage

during program execution. A hard disk is a non-volatile storage system, so it would

soon be out of s

ace if it re

laced the main memor

ardless of its ca

acit

19% 22%

Partially Correct Answer: The electronic circuitry of the main memory renders its

speed fast enough for synchronization with CPU circuits. A hard disk has mechanical

arts and re

uires a

sical motion for its o

eration, makin

its s

eed extremel

slow.

15% 7%

In the end, the screenshots of the online discussions

were collected and filed (Dataset3). The authors

studied and compared the opinions of each student

with their previous answers in the first part of the

project (Dataset2). The Dataset2 answers were

confirmed and enriched by the extra information

provided through the discussion session.

The discussion session served as a follow-up

method that provided clear information about

students’ knowledge since many of them elaborated

on their justifications to persuade their peers.

3.6 Results

The most frequent justification cases of the incorrect

and partially correct student answers are outlined in

Table 1. In both datasets, the same justification cases

prevail, regardless of the variation in the percentages

of students.

The interest of the study is focused on these most

frequent justification cases since it is likely that future

instructional interventions will concentrate on the

most common students’ conceptions. For this reason,

we calculated the inter-coder agreement for these

cases, and it was 93%.

As shown in Table 1, in both datasets, the

distribution of answers with cases 2, 3, 5, and 6 are

similar. On the contrary, there is a significant

difference in the percentages of cases 1 and 4. Great

interest arises from the fact that the percentage of

students who answered wrongly in 2011-2012 is

double than that of 2016-2017 (57% vs. 23%). A

possible explanation is that, before answering the

question, the Group B students had been engaged in

the relevant activities in SCALE. This probably

boosted the refinement of their knowledge. This

conjecture aligns with the study results about the

positive effects on students' learning after engaging in

SCALE (Verginis, et al, 2009).

Nevertheless, it is noteworthy that regardless of

the distribution variations, both groups of students

seem to hold the same conceptions even if they

expressed them under different conditions.

3.7 Students' Conceptions and

Emerging Misconceptions

The analysis of students’ answers revealed their

conceptions about main memory, hard disc, program

execution, and their interrelations. Here follows the

detailed description of these conceptions that answer

the research questions of this study and the related

misconceptions that appeared.

57% and 23% of the students in the two academic

years 2011-2012 and 2016-2017 respectively

answered incorrectly that a hard disk could replace

the main memory, with the drawback being the

slower operation of the computer, considering them

both storage systems. Case 1 answers revealed the

major misconception of the similarity between the

main memory and hard disk. Both are considered

storage systems with the same functionality. The fact

Using Kintsch’s Text Comprehension Model to Identify CS Students’ Conceptions and Misconceptions

389

that each unit has a different role, organization, and

method of data storage was ignored or underestimated

(RQ1, RQ2, and RQ3).

The majority of the students gave a wide range of

reasoning regarding the capability of the computer to

run when the main memory is missing. The most

common answers were: “hard disk is slower than the

main memory so the computer will run slower”; “a

hard disk doesn’t have an addressing system like the

main memory does, so the storing and reading

processes would be slower”; “hard disk doesn’t

communicate with the CPU through the bus as the

main memory does so the data transfer rate would be

lower”; “the implementation of the virtual memory

already uses a part of the hard disk and it may use the

whole capacity of the hard disk, when the main

memory is missing, with the flow of the slower

retrieve/store processes in a fully occupied disk”;

“cache memory could replace main memory during a

program execution, but as it has less capacity the data

retrieval from the hard disk would need much more

time.”

Even though there are some hints of knowledge

about the different structures and organization of the

two units (main memory and hard disk) in these

justifications, it seems that the students have

constructed these fragments of knowledge in the

context of their alternative frameworks.

A misconception about the performance of the

main memory derives from these justifications.

Students attribute the high performance of the main

memory either to its cell addressing system, which

enables the reading and writing processes to be quick,

or to the fast bit transfer rate of the bus connecting the

main memory with the CPU. Still, students overlook

the fact that this performance results from the

electronic circuitries of the main memory.

Either way, all the above justifications ignore the

role of the main memory and its organization for

program execution (RQ1 and RQ2).

Besides, some students hold a misconception

about the virtual memory mechanism, as they

consider that it can be implemented with a standalone

hard disk when the main memory is missing.

Furthermore, other students include the cache

memory in their justifications, stating that it may

substitute the main memory. These students consider

it feasible to run a program from the cache memory

when the main memory is missing, a statement that

reveals a misconception about the operation of cache

memory.

Students expressing the answers of case 2

superficially recall the von Neumann architecture

without providing any further information concerning

the research questions.

Case 3 students seem to recognize the role of main

memory (RQ1), but no information is provided for

RQ2 & RQ3.

Case 4 students claim that the hard disk cannot

displace the main memory as the start-up information

stored in the ROM is crucial. As evidence to their

claims, most of the students mention their practical

experience of a non-running computer that causes

“beep” sounds in case of RAM failure. The

misconception here is that the significance of the

main memory seems to be constrained only to a

storage place for the start-up information, i.e., the

ROM is the only important part.

Moreover, many of the answers mentioning this

justification case proposed overcoming the start-up

problem by storing the ROM data on the hard disk. It

seems that the underlying misconception here is about

the similarity of the two units (described above).

Case 5 students state that the main memory is

irreplaceable for a computer’s operation because of

its volatility. These students seem to acknowledge the

significance of the role of the main memory on

computer operation (RQ1), but not for the

scientifically correct reason (RQ2). Besides, many

students mention that hard disks are non-volatile, and

they would soon run out of space if they replaced the

main memory on program execution. This opinion

adds another aspect to the previous misconception,

suggesting that only volatility is the crucial difference

between the main memory and the hard disk. Students

ignore the data storage distinction between the two

units (RQ3). A few of them propose to solve the out-

of-space disk problem with proper programming,

which may erase the unnecessary data from the hard

disk whenever necessary. The background of such

thoughts is the misconception of similarity between

the two units.

Students mentioning the justification case 6 seem

to acknowledge the significant role of the main

memory (RQ1). They express the scientifically valid

opinion that the electronic circuitry of the main

memory renders it fast enough for synchronization

with the CPU circuits. These students state that the

main memory’s high performance is the reason for

storing a program’s instructions and data during its

execution. They seem to believe that only the main

memory's high performance makes it indispensable

for computer operation. So it seems that they ignore

the relation of the main memory’s unique structure in

ordered directly-accessed addressable cells for the

execution of the sequential instructions of a program

(RQ2). In addition, students think that poor

CSEDU 2022 - 14th International Conference on Computer Supported Education

390

Table 2: Brief description of students’ misconceptions and reasoning. Outcomes concerning the RQs.

Descri

tion of Misconce

tion Ex

lanation/Reasonin

/RQs results

The similarity between main memory

and hard disc.

Both are storage systems with the

same functionality.

Regardless of their differences, both systems can store data, so the main

memory’s replacement by a hard disk will only slow the system’s

performance.

RQ1 & RQ2: The main memory’s role and organization on a computer’s

operation are ignored or underestimated.

RQ3: This conception shows ignorance in data storage on both systems.

The significant role of the main

memory is the storing of start-up

information.

The only important part of the main memory is ROM.

RQ1: This conception disregards the role of the main memory during

ram execution.

a. Volatility is the reason for using

the main memory for data storage

during program execution.

b. A hard disk is non-volatile, so it

cannot replace the main memory

during program execution.

Volatility is considered the most significant difference between main

memory and hard disk and the reason for storing data in the main memory

throughout program execution.

RQ1: This conception recognizes the role of main memory

RQ2: This conception underestimates the actual reason for this role. The

main memory’s unique organization in ordered addressable cells allows

the direct access of a program’s data and instructions for sequential

execution.

RQ3: This conception shows ignorance in data storage on both systems.

a. High-speed performance justifies

the use of main memory during

program execution.

b. Hard disk has such a poor

performance that it cannot replace

the main memory during program

execution.

High-speed performance is the most significant difference between main

memory and hard disk and explains the use of the main memory to store

data during program execution.

RQ1: This conception recognizes the role of main memory

RQ2: This conception underestimates the actual reason for this role. The

main memory’s unique organization in ordered addressable cells allows

the direct access of a program’s data and instructions for sequential

execution.

RQ3: This conce

tion shows i

norance in data stora

e on both s

stems.

5 Virtual memory mechanism

A standalone hard disk can implement the virtual memory mechanism.

RQ1: This conception demonstrates ignorance about the role of main

memor

durin

ram execution.

6 Cache memory operation

A program can run from the cache memory when the main memory is

missing.

RQ1: The conception implies ignorance about the role of main memory

durin

ram execution.

Main memory’s performance is

dependent on the addressing system

or the data transfer rate of the bus.

This conception ignores the positive effects of the main memory’s

electronic nature/circuitry

performance is the reason for the inability of the hard

disk to replace the main memory, which is another

misunderstanding about what differentiates the two

units. Again the difference in data storage is ignored

(RQ3). Furthermore, some of the students propose the

SSD type of hard disks as a possible substitute for the

main memory, considering this technology

advancement a step towards replacing the main

memory in program execution. The latter

misconception is the source of such thoughts. The

reasoning is that since the discrepancy in the

performance of the two units is eliminated, the hard

disk could serve as the main memory. A synopsis of

the results of the study (Table 2) is that students: 1)

ignore or underestimate the role of main memory in

program execution (RQ1); 2) cannot link the main

memory’s organization in ordered directly accessed

addressable bytes/words with the sequential

execution of the program’s instructions (RQ2), even

when they recognize the role of the main memory in

program execution, and 3) in many cases overlook the

difference of data storage between the main memory

and the secondary memory (RQ3).

Using Kintsch’s Text Comprehension Model to Identify CS Students’ Conceptions and Misconceptions

391

3.8 Limitations of the Study

Bouvens stated that identifying students’

misconceptions from open-ended tests becomes

difficult since language problems make students

generally less eager to write their answers in complete

sentences (as cited in Kaltakci Gurel, Eryilmaz, and

McDermott, 2015). Thus, students’ potential

language problems may be a limitation since all the

input data were based on written responses.

Nevertheless, the group-discussion sessions allowed

students to re-think their opinions, study peers’ views,

reflect on different ideas and formulate their

responses. Thus, group-discussion sessions served as

a diagnostic tool that strengthened the results of the

open-ended test.

Group A and Group B students had different

course instructors during the two academic years.

However, the educational setting of the introductory

course remained the same, i.e., the educational

material (course-books, supplementary electronic

material) and the instructional approach (lectures).

Even though the particular course books are broadly

used in higher education CS introductory courses

worldwide, this may limit the extent to which these

results may be replicable in studies using different

educational material and possibly different teaching

methods.

The fact that the data were collected only from the

students at one institution imposes another possible

limitation to the study. Nevertheless, the students of

this department are considered top-qualified students

on a national level, as they achieved very high grades

on the national entrance exams. Moreover, the input

data of the study were collected with a time interval

of five years that strengthens the robustness and time

persistence of the results.

4 DISCUSSION AND

CONCLUSIONS

The analysis of students’ answers to understanding

questions reflects what exists in their minds. The

bridging-inference type of questions based on

Kintsch's text comprehension model provides a

suitable context to explore students’ ideas on a range

of concepts. To answer such questions, the reader

needs to combine knowledge from several text parts

and infer their unstated relations.

Using a bridging-inference question, the study

revealed that a considerable percentage of students

have such superficial knowledge on the concepts of

main memory organization, operation and role, hard

disc structure, data storage, program execution, and

their interrelations, that in some cases it is

scientifically invalid. These naive or intuitive

students’ conceptions are possibly based on everyday

experience, their social network, and knowledge on

Informatics which was acquired during the school

years (e.g., “beep” sounds in case of RAM failure,

volatility of main memory)

In line with Grigoriadou and Kanidis’s (2002)

research, a significant percentage of students attribute

some of the characteristics of secondary memory to

the main memory. What is more, those scholars have

shown that although most of the school students

acknowledged that data are stored on main memory

temporarily (volatility) or that the main memory is a

high-speed memory, nearly none recognized the role

of main memory addressing or the necessity of

storing the program on main memory for its

execution. These results resemble the misconceptions

of volatility and high-speed performance revealed in

the present research. The similarity between the

school students' and the CS students' conceptions may

be explained by the notion that in discourse

comprehension, prior knowledge “provides part of

the context within which a discourse is interpreted.

The context is thought of as a kind of filter through

which people perceive the world” (Kintsch, 1988).

What is being argued here is that when the CS

students (both newcomers and older ones) were

exposed to the new knowledge (in text or oral form,

by reading the course books or attending the lectures),

they constructed discourse representations

constrained by their prior school knowledge, among

other factors. An implication of this is the need to

diagnose CS students’ prior conceptions early on

before delivering lectures, plan the sequence of topics

instructed, and refine and reorganize preconceptions

by using specific instructional interventions

Undoubtedly, another important influencing

factor of the learning process is the educational

material, as presented in the offered course books.

Although both are highly accepted and used

worldwide in higher education CS introductory

courses, a review of these course-books resulted in

the observation that the concepts under investigation

are not presented in a satisfactory manner. Some

knowledge components are described in detail, others

briefly, or even confusingly, and information about

their linkage is unclear or missing. For example, the

Forouzan course book firstly presents the memory

hierarchy – registers, cache, main memory accurately

- both in text and visual form (see Figure 5.4) so that

the readers can become aware of their underlying

CSEDU 2022 - 14th International Conference on Computer Supported Education

392

similarities. But afterward, while presenting the mass

storage systems it includes the main memory in a

misleading comparison that possibly confuses readers

and enhances the misconception of similarity. It is

noteworthy that relevant books (e.g the well-known

Patterson-Hennessy book (2005) contain visual

representations of the memory hierarchy (see Figures

7.1 and 7.3) that compare computer memory with the

mass storage systems, referring to them as memory

too. This kind of presentation may mislead readers by

creating the impression of the similarity of the

systems. One immediate implication of this is the

need for teachers to review textbooks or other

educational material to detect points that may cause

misunderstandings. The teachers should emphasize

the misleading references during the lectures to

prevent the potential consolidation of

misconceptions.

Another aspect regarding the educational material

is that most of the questions provided at the end of

each section in the two course books are text-based,

focusing on memorization or comprehension.

Consequently, there is a lack of bridging-inference

questions that might help or challenge students

thinking about the unstated interrelations of the

concepts. The bridging-inference question used in the

study proved challenging enough for students to

attempt to link the topics. This fact implies that the

enrichment of the educational material with this type

of question seems a promising practice to enhance

students’ linking-inference skills and consequently

their learning.

Furthermore, the effect of the course-book text

coherence on the students’ understanding of the

concepts under study is not negligible. Studies in the

domain of text comprehension (McNamara et al.,

1996; McNamara & Kintsch, 1996) and specifically

on learning from texts on computer science

(Gasparinatou and Grigoriadou, 2013) have

concluded that high-knowledge readers benefit from

a low cohesion text. Conversely, low-knowledge

readers benefit from a high cohesion text. The

students of the empirical study were mainly first-year

students who are low-knowledge readers and do not

seem to benefit from the low cohesion text of the

course books. This indication is in agreement with the

results of the text comprehension studies. Α resulting

implication of this issue is the need to revise the

educational material under the prism of text

coherence. That is, to present content with high

cohesion texts to facilitate first-year students’

understanding of the concepts under study.

In conclusion, formulating bridging-inference

questions based on Kintsch’s text comprehension

model seems an effective tool to assess and

investigate students’ understanding of CS curricula.

Whether the use of this model and especially of well-

designed bridging-inference questions, is a promising

tool in the direction of supporting knowledge

restructuring and refinement, needs to be further

investigated.

REFERENCES

Ausubel, D. (1968). Educational Psychology. N.Y, Holt,

Rinechart & Winston.

Ben-Ari, M. (2001). Constructivism in computer science

education. J. Comput. Math. Sci. Teach. 20(1), 45–73.

Bransford, D. J., Brown, L. A., & Cocking, R. R. (editors)

(2000). How People Learn Brain, Mind, Experience,

and School, Committee on Developments in the

Science of Learning. National Research Council.

National Academy Press.

Braun, V., & Clarke, V. (2006). Using Thematic Analysis

in Psychology. J. Qualitative Research in Psychology,

3(2), 77-101.

Brookshear, J. G. (2009). Computer Science: An Overview,

10th Edition. Pearson Education, Inc, publ as Addison

Wesley Higher Education.

Clancy, M. (2004). Misconceptions and attitudes that

interfere with learning to program. In S. Fincher and M.

Petre, (eds), Computer Science Education Research.

Taylor and Francis Group, London.

Clements A., (2000). The Undergraduate Curriculum in

Computer Architecture, Computer Architecture

Education. Micro May/June 2000, pp 10-22

Davis, R., Maher, C. and Noddings, N. (1990).

Constructivist views of the teaching and learning of

mathematics. Journal for Research in Mathematics

Education, Monograph No.4.

Forouzan, B. A. (2002). Foundations of Computer Science

– From Data Manipulation to Theory of Computation.

Publ. by Brooks/Cole, Thomson Learning.

Gasparinatou, A., & Grigoriadou, M. (2013) Exploring the

effect of background knowledge and text cohesion on

learning from texts in computer science. Educational

Psychology, 33:6, 645-670.

Gogoulou, A., Gouli, E., Grigoriadou, M., Samarakou, M.,

and Chinou, D. (2007) “A Web-based educational

setting supporting individualized learning,

collaborative learning and assessment,” Educ. Tech.

Soc. J., vol. 10, no. 4, pp. 242–256.

Grigoriadou, M., & Kanidis, E. (2001). Students’

approaches to the Computer Cache Memory and their

Exploitation in the Development of a Web-based

Learning Environment. In Proc. of 8th PanHellenic

Conf. in Informatics, Cyprus, 472-481.

Grigoriadou M., & Kanidis E. (2002). Secondary education

students’ conceptions about computer main memory

organization and operation. In Proc. of 3rd Pan Hellenic

Conference with International Participation:

Using Kintsch’s Text Comprehension Model to Identify CS Students’ Conceptions and Misconceptions

393

Information and Communication Technologies in

Education. Rhodes, Sep. 2002, pp. 249-258

Herman, G. L., Zilles, C., Loui, M. C. (2009, October).

Work in progress: Students' misconceptions about state

in digital systems. Proceeding of 39th ASEE/IEEE

Frontiers Education Conference, pp. T4D1-T4D2.

Herman, G. L., Loui, M. C., Zilles, C. (2010, October).

Work in progress: How do engineering students

misunderstand number representations? Proc. 40th

ASEE/IEEE Frontiers Educ. Conf., pp. T3G1-T3G2.

Herman, G. L., Kaczmarczyk, L., Loui, M. C., Zilles, C.

(2011). Discovering students' misconceptions in

boolean logic. Trans. Comput. Educ.

Herman, G. L., Loui, M. C., Zilles, C. (2011). Students'

misconceptions about medium-scale integrated circuits.

IEEE Transactions in Education.

Johnson-Laird, P. N. (1983). Mental models: towards a

cognitive science of language, inference, and

consciousness. Harvard University Press.

Kaltakci Gurel. D., Eryilmaz, A., & McDermott, L. (2015).

A Review and Comparison of Diagnostic Instruments

to Identify Students’ Misconceptions in Science.

Eurasia Journal of Mathematics, Science and

Technology Education, 11(5), 989-1008.

Kintsch,W. and van Dijk,T.A. (1978). Toward a Model of

Text Comprehension and Production. Psychological

Review, 85(5), 363-394.

Kintsch, W. (1988). The Role of Knowledge in Discourse

Comprehension: A Construction-Integration Model.

Psychological Review, 95(2), 163-182.

Kyriakou C. and Grigoriadou M. (2016, October).

Students’ cognitive difficulties on basic informatics

concepts. Proc. of 8th CIE2016 – 8th Conference on

Informatics in Education, pp. 314-326.

McNamara, D. S., Kintsch, E., Songer, N. B., and Kintsch,

W. (1996). Are good texts always better? Text

coherence, background knowledge, and levels of

understanding in learning from text. J. Cognition and

Instruction, 14(1), 1-43.

McNamara, D. S., & Kintsch, W. (1996). Learning from

texts: Effects of prior knowledge and text coherence.

Discourse Processes, 22, 247–288.

Patterson, D. A., & Hennessy, J. L. (2005). Computer

Organization and Design, The Hardware/Software

Interface/3rd Edition. Publ. by Elsevier Inc.

Porter, L., Garcia, S., Tseng, H. W., & Zingaro, D. (2013).

Evaluating student understanding of core concepts in

computer architecture. In Proc 18th ItiCSE,

Canterbury, UK, 279–284.

Posner, G. J., Strike, K. A., Hewson, P., W. & Gertzog, W.

A. (1982). Accommodation of a Scientific Conception:

Towards a Theory of Conceptual Change J. Science

Education, 66, 211-227.

Van Dijk, T. A., and Kintsch, W. (1983). Strategies of

discourse comprehension. San Diego, CA:Academic

Press.

Verginis I., Gogoulou A., Gouli, E., Boubouka M., and

Grigoriadou M. (2009). Enhancing Learning in

Introductory Computer Science Courses through

SCALE: An empirical study. IEEE Transactions on

Education, 54(1), pp. 1-13.

Von Glaserfeld, E. (1995). Radical Constructivism: A way

of Knowing and Learning, London: The Falmer Press.

Vosniadou, S., & Brewer, W. F. (1987). Theories of

knowledge restructuring in development. Review of

Educational Research, 57, 51-67.

Vosniadou, S. (1994). Capturing and modeling the process

of conceptual change. J. Learning and Instruction, 4

(1), 45-69.

Vosniadou, S. (2001). How Children Learn (Educational

Practice Series). Brussels, Belgium: International

Academy of Education.

Vosniadou, S. (2003). Exploring the relationship between

conceptual change and intentional learning. In G.M

Sinatra and P.R. Pietrich (Eds.) Intentional Conceptual

Change. Mahwah, Lawrence Erlbaum Associates, 377-

406.

Vosniadou, S. (2007). Conceptual Change and Education,

in Human Development, vol. 50, 47–54.

CSEDU 2022 - 14th International Conference on Computer Supported Education

394