Concept Maps and the Processes o f Compnhension: Explicating Cognition and

Metacognition, Structural Knowledge and Procedural Knowledge

Graham James Passmore

A thesis submitted in conformity with the requirements

for the degree of Ph.D.
Graduate department of Cumculum, Teaching and Lcaming
University of Toronto

O Copyright by Graham James Passmore (1999)

IM National Library
0fCa"ada
Biblioth6que nationale
du Canada
Aq@tions and Acquisitions et
Bibhographic Secvices seMces bibliographiques
395 Wellington Street 395,rue Wellington
Ottawa ON K1AûN4 OttewaON K 1 A W
Canada canada

The author has granted a non- L'auteur a accordé une licence non
exclusive Licence allowing the exclusive permettant à la
National Library of Canada to Bibliothèque nationale du Canada de
reproduce, loan, distri'bute or sell reproduire, prêter, disûî'buer ou
copies of this thesis in microform, vendre des copies de cette thèse sous
paper or electronic formats. la forme de microfiche/fb, de
reproduction sur papier ou sur format
électronique.

The author retains ownership of the L'auteur conserve la propriété du

copyright in this thesis. Neither the droit d'auteur qui protège cette thèse.
thesis nor substantial extracts eom it Ni la thèse ni des extraits substantiels
may be printed or otherwise de celle-ci ne doivent être imprimés
reproduced without the author's ou autrement reproduits sans son
permission. autorisation.
 A bstract

Concept Maps and the Processes of Comprehension: Explicating Cognition and

Metacognition Structural Knowledge and Procedural Knowledge

Graham James Passmore Ph. D. ( 1999)

Department of Curriculum, Teaching and Learning. OISUUniversity of Toronto

Concept maps are said to represent mental models of problems that form in working

memory. Integration of theories of cognition indicates that the structural and procedural

knowledge of concept map links are comparably complex. Integration funher suggeas that

comparably complex cognitive and metacognitive processes govem concept mapping. This

thesis develops methods to mess the structural and procedural knowledge of concept maps,

as well as their cognitive and metacognitive processes.

The SOLO (Structure Of the Learning Outcorne) taxonorny is adapted to assess

structurai knowledge in concept maps. The adaptation implies that five levels of inference

m a t e structural knowledge. Descriptions of the inference levels are used to develop an

afsessment scale for procedural knowledge. Applying the methods to students' concept maps,

shows that their structural and procedural knowledge is comparably complex. It is argued

that concept map procedural knowledge and cognitive processes are the same phenornenon.

Therefore, the procedural knowledge assessrnent scaie is used to rneasure cognition. A set of

interview questions is developed to assess metacognition. The methods indicate that snidents'

concept maps exhibit cornparably complex cognitive and metacognitive processes.

Confonnity between concept map properties and the predictions of theory integration is said

to support the argument that concept maps represent mental modets.

 Table of Contents

Chapter 1 The Educational Uses o f Concept Maps

1 .O OveMew of the chapter

1.1 Reading the thesis

Chapter 2 A Sumrnary o f the Thesis

2.0 How concept maps are used and the impetus for this thesis

2.1 h brief introduction to the goals of the thesis

2.2 Mental models and concept maps

2.3 The cognitive and metacognitive processes of concept mapping

2.4 The structure of the concept map link

2.5 Three methods for concept map analysis and an interview

2.6 The ernpincal component of the study

Chapter 3 Concept Maps and Mental Models

3 -0Integrating theories of cognition and concept map theory

3.1 The structure of rnemory and concept retneval

3-2 Structurai knowledge

3.3 A problem with the description of structural knowledge

3.4 Schema theory

3.5 Frames 31

3.6 Scripts 32

3.7 Marshall's schema theory: A summary 34

3.8 lntegrating schema theory and stmctural knowledge theory 34

3.9 The scaffolding of procedural knowledge 36

3.10 Schema theory and concept maps 38

3.1 1 A concept rnap to illustrate intesration of structural knowledge and schema theory 40

3.12 A different interpretation of the nature of knowledge 32

Chapter 4 The Metacognitive and Cognitive Processes of Concept Map Creation

4.0Concept mapping and reading comprehension

4. i A description of metacognition

4.2 Comprehension and reading

Chapter 5 An Introduction to Concept iMap Analysis

5 .O An example concept map

5.0 1 Knowledge to be expressed in the rnap

5.02 The map

5.1 Novak and Gowin ( 1984) and the structure of concept maps

5.2 The Novak and Gowin scoring technique

5.3 An example Novak and Gowin score

Chapter 6 The Next Ceneration of Assessrnent Methods

6.0 Concept map assessment methods and the goals of the thesis

6.1 The SOLO knowledge assessment method

6.2 The prestructural SOLO response

6.2.1 Capacity, closure and consistency

6.3 The unistructural SOLO response

6.4 The multistructural SOLO response

6.5 The relational SOLO response

6.6 The extended abstract SOLO response

6.7 General trends in SOLO stmctures

Chapter 7 The Analytical Methods and an Interview

7.0 Method 1: An adaptation of SOLO

7.1 Prestructurd links

7.2 Lfnistmctural links

7.3 Multistnicturd links

7.4 Relational links

7.5 Extended abstract links

7.6 Assessing the structural knowledge of entire maps and individual concepts
7.7Weaknesses of Method 1

7.8 Method 2: Progressive diferentiation and individual concepts

7.9 Limitations of Methods 1 and 2

7.10 Method 3 : Link-procedural knowledge

7.1 1 The interview

7.1 1.1 The interview questions

7.12 A summary of Methods 1.2 and 3

Chapter 8 A Description of the Empirical Components of the Experiment

8.0 The experiment

8.1 Methodolow and data collection

8.2 Written instructions presented to the teacher

8.3 Introducing students to concept maps

8.4 A map of the tarset text

Chapter 9 Assessing Structural and Link-Procedural Knowledge

9.0 An oveMew o f the chapter

9.1 Interrater reliability

9.7 The nature of link-procedural knowledge

9.1.1 Subject 1

9.3.2 Subject 2

9.2.3 Subject 6
9.3 A summary of the chapter

Chapter 10 Assessing the Psychologicai Validity o f Concept Maps

10.0 An o v e ~ e w
of the chapter

1 0.1 The relationship between structural and link-procedural knowledge

10.2 Progressive differentiation and integrative reconciliation

1O. 3 Cognitive and metacognitive activity

Chapter I l Conclusions

1 I .O Conclusions drawn From the data analysis

1 1 . 1 implications for pedagog and snident leaming

1 1.2 Future research

Bibliography

Appendices

Appendix A

Appendix B

Appendix C

Appendix D

Appendi~E

vii
Tables

Table 1 : Descriptions of the links of Figure 3 56

Table 2: A concept map s c o h g system (afler Novak and Gowin ( 1984))for Fisure 4 60
Table 3: A prestructural SOLO response expressed in link format 73
Table 4: A unistructural SOLO response expressed in Iink format 74
Table 5: Example unistructural links in Figure 5 75
Table 6: A multistructural SOLO response expressed in link format 75
Table 7: A relational SOLO response expressed in link format 76
Table 8: Exarnple relational links in Figure 5 77
Table 9: An extended abstract SOLO response expressed in link format 78
Table 10:The structural knowledge links of figure 3 81
Table 1 I : Nurnber-Link (Method 2) ratings of Figure 5 85
Table 12 : Method 1 and Number-Link ratings for concepts in Figure 5 87
Table 1 3: Method I and Method 3 ratings for the links of Figure 5 93
Table 14:Descriptions of the links of Figure 1 1 104
Table 15: The links of Figure 12 111
Table 16: Methods 1.2 and 3 ratings for the key concepts 112
Table 17: The links of subject 1 119
Table 18: The relationship between learnins modes and methods 1.2 and 3 123
Table 19: The links of subject 2 129
Table 20 :The links of subject 6 136
Table 3 1 : The stnictural-link-procedurd knowledge relationship of Figure 1 3 139
Table 22 : The stmctural-link-procedural knowledge relationship 140
Table 33 : The key concepts of subject 1 141
Table 21: The oil spills concept of subject 2 142
Table 25 : The oil spills concept of subject 6 142
Table 26: The cognitive activity of the six subjects 147
Table 27: Levels of inference in maps of high, medium and low level cognitive activity 147
Table 28: Co_&tive activity and metacogiitive i n t e ~ e wscores 149

Figures

Figure 1 : The goals of the thesis

Figure 2: The typical link structure that is required for this thesis
Figure 3: The integration of schema theory and structural knowled~etheory
Figure 4: The fermentation pathway
Figure 5: Concept map of fermentation
Figure 6: The prestructural SOLO structure
Figure 7: The unistructural SOLO structure
Figure 8:The multistructural SOLO structure
Figure 9: The relational SOLO structure
Fiy r e 10: The extended abstract SOLO structure
Figure 1 I : .An example concept map.
Figure 12: A concept map of the tarset text
Figure 13 : Subject 1. an example of high level cognitive activity
Figure 1 4: Subject 2' medium IeveI cognitive activity
Figure 15: Subject 6, low levei cosnitive activity
 Chapter 1

The Educational Uses of Concept Maps.

1.0 Overview of the thesis

This thesis is primarily concemed with the argument that concept maps represent

mental models that form in working memory in response to problems or situations. Theones

of cognition (structural knowledge (Jonassen. Beissner and Yacci. 1993), schema theory

(Marshall, 1999, semantic network theory (Ashcrafl, 1989), and spreading activation theory

(Collins and Loftus. 1975)are integrated to identify properties of the structural and

procedural knowledge associated with mental models. lntegration of the theories is also used

to develop three rnethods for the assessrnent of structural and procedural knowledge in

concept maps. When audents' maps were analyzed with these methods they were found to

exhibit properties that parallel the predicted properties of mental models. This conformity

between the properties of concept maps and mental models is said to be supportive of the

notion that concept maps represent mental models.

The thesis begins by considering the integration of four theories of cognition (schema

theory, stmcnirai knowledge theory, semantic networking theory and spreading activation

theoiy) and concept map theory. Integration is based on the argument that components of

different theories that perform the sarne function are actuaiiy the same phenomenon. The

process indicates that the structural and procedural knowledge of schemata develop

simultaneously, and with isomorphic' quality. It also indicates that these knowledge forms are

'By isomorphic 1 mean that structurai and procedural knowledge develop simultaneously
in response to the sarne stimuli. Increases in the quality of structural knowledge are associated
with concomitant increases in the quality of procedural knowledge.
operative during the creation of mental models. The second goal of the thesis is to develop

methods for the assessrnent of structurai and procedural knowledge in concept maps. When

these methods are applied to students' concept maps they indicate t hat their structural and

procedural knowledge are isomorphic in quality. Therefore it is argued that concept maps

exhibit properties that compare to the properties of mental models as predicted by theory

integration. This finding suggests that the rationale for theory inteyration is correct, it is also

said to lend weight to the notion that concept maps represent mental models.

Three methods are developed for concept map analysis. Method I assesses the quality

of structural knowledge. Method 3 assesses the quality of procedural knowledge. Method 2

is concerned with a subplot of the thesis. It is used to assess progressive differentiation. In

assessing the quality of aructural knowledge, Method 1 also measures integrative

reconciliation. Methods 1 and 2 are used in conjunction to assess the integrative

reconciliation and progressive differentiation associated with individual concepts. This use of

methods 1 and 2 complements the earlier work of Novak and Gowin ( 1984). These workers

enumerated links to provide total scores for concept maps. In counting links and awarding

them a score. Novak and Gowin bundled the effects of progressive differentiation and

integrative reconciliation. Methods 1 and 2 disentangle progressive differentiation fiom

inteprative reconciliation. They provide a separate score for each process.

The third goal of the thesis seeks to strengthen the argument that concept maps

represent mental modeis. A strengthening effect is achieved by triangulating the analyses of

nnicnual and procedural knowledge. Theones of reading comprehension (Jacobs and Paris,

1987 and Kirby, 1991) contend that the cognitive and metacognitive processes that govem an
 3

instance of reading are comparable in their complexity. These theories are integrated with

concept map theory by way of a Functional equivalence argument. Integration indicates that

the same cognitive and metacognitive processes control reading comprehension and concept

map creation. lt also susgens that the procedural knowledge of concept mapping is

functionally equivalent to the cognitive activity of reading comprehension. Therefore. in

measuring the quality of procedural knowledge. Method 3 is said to assess the cognitive

processes of concept mapping. Metacognition is assessed wit h an i n t e ~ e wscnpt developed

especially for this thesis. When Method 3 and the interview scnpt were applied to students'

concept maps, they indicated that cognition and metacognition are comparably complex

processes. This finding indicates that concept maps exhibit properties that conform to the

predictions of theory integation. It therefore supports the rationale for theory integation and

by extension it lends weight to the notion that concept maps represent mental models.

I. 1 Reading the thesis

Two descriptions of the thesis are provided in the following pages. Chapter 2

provides an introductory description of the goals and arguments of the thesis. Section 2.0

provides some background information about concept maps. It also descnbes the impetus for

this research. Section 2.1 introduces the thesis and its ~oals.In section 2.2 the argument that

concept maps represent mental models is discussed. The cognitive and metacognitive

processes of concept mapping are described in section 2.3. Methods 1,2 and 3 are described

in section 1.4 and section 3.5 descnbes the structure of concept map links. Section 2.6 closes

the chapter with a description of the empiricai component of the thesis.

 A second highly detailed run at the materiai begins in section 3.0. Chapters 3 and 4

present the psychological validity argument. Chapters 5,6 and 7 descnbe methods of concept

map andysis and the empincal component of the thesis is descnbed in chapter 8. Chapters 9

and 10 descriibe the assessrnent of the concept maps of a gifleci grade 9 biology class. These

students were select& for the snidy as it was thought that their academic ability would

increase the likelihood that high quality links would be produced. The students were also

chosen as they were snidying environmental biology. Al1 the maps created in this study were

based upon a target text that descnbes an aspect of environmental biology. The text was

taken from the World Wide Web at: http://www.soton.ac.uk/engenvir/enviro~.

It

concems water pollution and oii slicks. Conclusions derived from chapters 9 and 10 are

presented in chapter 1 1.
 Chapter 2

A Sumrnary of the Thesis

2.0 Bon concept maps are used and the impetus for this thesis

Concept maps have been put toward a multitude of educational uses. In a discussion

of their utility, Edmondson and Smith (1996) make note of applications for teaching and

learning. Two of the most important applications are, increasing the potential for recalî and

maximizing the usefulness of knowledge. To explain, concept map creation requires the

reprocessing of extant conceptions. Reprocessing strengthens the links between concepts and

thereby increases the potential for h r e recall. The usefûiness of knowledge can be

increased if students reaect on the content of a map. Such reflection provides an awareness

of the "big picturen as it allows for link creation across multiple themes. Multiple thematic

connections d o w for the transfer of knowledge to multiple situations (Spiro, Feltovich,

Jacobson and Coulson, 199 1a). Knowiedge transfer can provide unique solutions to problems

as the thematic connections that make it possible are unique to their creator. Teachers can

actively promote the creation of multiple thematic connections by using several concept maps

to present the same knowledge. Each of these maps should present the topic from a dinerent

perspective or theme. Aithough this process makes thematic connections expiicit, dadents

can still create their own connections if they are provided with oppominities for reflection.

m e r uses that teachers have for concept maps include: planning curricula, evaluating

leaming, identdyurg erroneous thinking and making qualitative judgements about the

understanding arpressed in maps (Barenholtt and Tamir, 1992, Edmondson., 1993,

Edmondson and Smith, 1996, Fisher, 1996, Liu, 1994, Morhe-Dershimer, 1993).
 The usefulness of concept maps is often based on the assumption that they render

mental models available for analysis. That is, concept maps are assurned to represent

schernata in some way.Schemata are components of long term memory that develop in

response to repeat experiences. They conskt of interco~ectednodes. Verifjmg that concept

maps represent schemata has proved to be a difficult undertaking. Part of the problem is that

educators are still debating the very existence of schemata. This leaves advocates of concept

maps in the doubly precarious position of having to extol the virtues of a tool that

theoretically explicates (on some unidentifieci level) a phenomenon that is itself only

theoreticai in nature.

The difficulty of determining if concept maps are reflective of schernata provided the

impenis for this thesis. An investigation was launched into the nature of mental models and

concept maps. The author felt that the use of concept rnaps in education is inhibited by a lack

of evidence to indicate thiit they represent mental models. It was hoped that a greater

understanding of concept maps' relationship to mental models would increase their

acceptance within the educational community.

2.1 A brief introduction to the goais of the thesis

An investigation into the representation of mental models by concept maps is the

primary goal of this thesis. The second goal is to develop methds for concept map analysis

and the third is to assess the cognitive and metacopnitive processes of concept mapping. The

latter goal is designed to provide further information about model representation by concept

maps. The second goal led to the production of three concept map assessrnent methods
(Method 1, Method 2 and Method 3) and an intewiew script. Two factors enabled the

development of these methods. The first was the argument that enabled the integration of

concept map theory and the four theories of cognition. When M y developed it suggested

that the structural and procedural knowledge of mental models and concept maps are

isomorphic in quality. It also revealed a iittle of the nature of these knowledge fonns.

Descriptions of the knowledge forms made it possible to develop methods to investigate

them. The second factor was a description of the cognitive and metacognitive processes of

concept rnapping. Here again, descriptions of the processes revealed their nature and

methods were developed to assess these properties.

Figure 1 sumrnarizes the goals of the thesis. in regard to the fint goal it illustrates

that structural and procedural knowledge are said to have isomorphic quality in concept

maps. It also suggests that this isomorphism can be checked by way of Methods 1 and 3.

Finally Figure 1 illustrates that the comprehension processes (cognitive and metacognitive)

that govem concept map creation can be elucidated through an i n t e ~ e wand the third

anaiyticai method (Method 3).

An argument is laid out in this thesis wherein concept maps are said to represent

mental models of problems. These mental rnodels are said to consist of components of

schemata and the problern that stimulateci formation of the model. Supportive evidence for

this argument is sought by assessing whether the structural and procedural knowledge of

concept maps has comparable properties to that of mental models. Fwther evidence for the

argument is sought fiom an investigation hto the cognitive and metacognitive processes of

concept mapping. Here again, the purpose of the investigation is to determine whether these
 processes exhibit comparable properties in concept maps and mental models.

Before a more detailed sumrnary of the thesis' argument and investigations can be

provided, some background knowledge of schema theory and concept rnap theory is

required.

1- Asseashg knowleâge in concept maps

Methoâ 1 assesses structural knowldge l

Method 2 enurnerates links

1

-II3 Elucidatethe comprehsnsion pracesses ofconcept mapping

, Metacognition 1 1s isomorphic to Cognitive pcesses I
I _ _ I J

; lntenriew scores Are isomorphic to

I A
Method 3 Wngs

me t The goals of the thesis

r

Schemata develop fiom repeat experiences (travelling to work, making a cup of tea etceteni)

with an action or problem (Marshall 1995). They are components of long tena memory that

consist of Uitercomected nodes. Explanations for the co~ectionsbetween nodes are known
as schema knowledge. The quality of schema knowledge inmeases as experience with the

action or problern is acquired. During initial exposure to an action or problem type, the

knowledge extracted is specificand concrete. During repeat exposures, increasingly abmact

and generalized schema knowledge is extracted. In a highly developed schema, schema

knowledge may be so abstract as to be stereotypical of a problem type or action.

Concept map links are created by a form of procedural knowledge we will cal1 link-

procedural knowledge. The prefix "link" distinguishes this form of procedural knowledge

fiom the procedwal knowledge that enables the performance of tactile operations (sawing

wood, skiing a slalom etcetera). Link-procedural knowledge identifies relationships (links)

between concepts. Structural knowledge is the expression of understanding that connects two

concepts in a concept map link. It results from the actions of link-procedural knowledge.

That is, link-procedural knowledge foms a basis of understanding that is used to generate a

linking siatement (structural knowledge).

Descriptions of the components of schema theory and concept maps suggest that

schema knowledge is fÙnctionalIy equivalem to the structural knowledge of concept maps.

These descriptions also suggest that the procedural and schema (structural) knowledge of

schemata have isomorphic qualities and are active during concept map creation. An

investigation into the two knowledge forms that finds they exhibit comparable quaiity in

concept maps lends would lend weight to the rationale for theory Unegration. By extension it

would also be supportive of the notion that concept maps represent mental models. This is

the primary argument of the thesis.

Besides the notion that structurai and link-procedurd knowledge are isomorphic,
 10

theory integration provides two additional revelations that are wonhy of note. The 6rst is the

process by which schema components are retrieved and subsequently expressed in concept

maps. The second is the manner in which structural knowledge guides the actions of

procedural knowledge.

Semantic network theory and spreading activation theory indicate how schema

components are retrieved from long term memory and subsequently expressed. These

theories maintain that constraints of problems (or any environmental stimuli) are carried to

working memory where they provoke a search of long term memory. During this search,

nodes that are contex~allyrelated to the constraints may become activated. Nodes that attain

a threshold level of activation can be retrieved to working memory. Once in working memory

the nodes' relationship to the problem's constraints may be reflected upon and manipulated.

The results of this refleçtive process are used to form a mental mode1 of the problem. The

memal mode1 may be expressed as a concept map.

Jonassen, Beissner and Yacci (1993) refmed to the process by which structural

knowledge guides the actions of procedurai knowledge as the scafluildig of proçecfrrrai

knawfedge. The knowledge that concept maps represent mentai models of problems was used

to acquire an understanding of the sdolding process. The foUowing problem is used to

describe the sdolding process. Suppose a machinist has to cut a circular hole in a sheet of

aluminum. Suppose also that the hole has to be sized so that a cyündrical bar will pass

through it and create an annulus between the bar and the edge of the hole that is 1 m h e t r e

thick (i.e. the diarneter of the hole has to be 2mrn p a t e r than the diarneter of the bar).

F i y suppose that the ody information available to the machinist is the circumference of
 11

the bar. According to Marshall, when a problem is encountered its constraints are carried

from the environment to working memory by way ofa sensory register. Marshall also noted

that components of schemata are recalled to working memory in response to the problem's

constraints. In the hypothdcal problem, the sensory register would carry the circderence

of the bar (20mm)and other problem constraints (create an annulus with a diameter that is

2mm greater than that of the bar) to working memory. Recailed schema components would

likeiy include the formula for the circumference of a circle (21r).

Marshall's theory then holds that probiem resolution continues by rnanipulating

r d e d components and the problem's constraints in working memory. This manipulation

connects components and constraints through expressions of association (structural

knowledge). The rewlt is a mental model that is used by procedurai knowledge to develop

plans for problem resolution. The mental model of the hypothetical problem might be that

"2xr = 20mrn." Procedural knowledge might devise the following steps for resolution: divide

2 h by n, divide 20mm/n by 2 to determine the radius of the cyhdncal bar, add 1 mm to

accommodate the annuius and to determine the radius of the circular hole. Procedural

knowledge would then carry out the steps of this plan and find that the circle requires a

radius of 4.183mm.That i s procedural knowledge uses the mental model to develop a plan

of action for problem remlution. Given that mental models are constniaed from s t m d

knowledge, t can be said that structural knowledge guides or d o l d s the actions of

procedurai knowledge.

ùi the following sections of this chapter the goals of the thesis are discussed in greater

detail. Section 2.2 considers the integration of theories of cognition as means for
 12

investigating the nature of the stmctural and procedural knowledge of concept maps. The

isomorphic nature of cognitive and metacognitive activity is discussed in section 2.3. Section

2.4 presents an aside, it describes the structure of concept map links.The anaiytical methods

and the i n t e ~ e wscript that are used to assess structural knowledge, procedural knowledge

and metacognition are presented in section 2.5. Finally, a description of the empirical

component ofthe thesis is provided in section 2.6.

2.2 Mental models and concept mips

This thesis seeks to investigate the nature of the aructural and procedural knowledge

associated with concept maps. The nature of these knowledge forms is revealed through an

integration of concept map theory and four theories of cognition (xhema theory, suuctural

knowledge theory, semantic networking theory and spreading activation theory).

Descriptions of the theones reveal that cenain components in different t heories perforrn the

same functions. These fiinctionally equivalent cornponents are thought to represent different

views of the same phenornenon. Components that occur in multiple theories represent points

at which the theones overlap. They make integration possible.

Several assertions about the nature of knowledge are derived fkom the integration

process. For example, structurai and procedural knowledge are said to develop

isomorphicdy in schemata. In addition it is suggested that the structurai and procedurai

knowledge of schemata are f'unctionaily equivaient to the structural and link-procedural

knowledge of concept mapping.

MarshaU (1995) provides evidence that scherna knowledge is a psychologically valid

 13

c o n a m . For example, she reports that researchers (Anderson, 1983 and SteRenson, Joag-

dev and Anderson, 1979) have demonstrateci that performance is influenced by prior

expenence. Marshall aiso cites Bransford and Johnson's (1972) work in which stories are

used to trigger inappropriate schemata and invoke rnisunderstandings. These examples

indicate that schemata are psychologically vaiid.

Evidence for the psychological validity of schemata lends weight to the argument that

concept maps are useful educational tools. In this thesis it is assened that schema knowledge

is functionally equivalent to the structural knowledge of concept maps. Concept map theory

holds that mental models contain elements of a problem that have entered working memory.

It also holds that mental models contain components of schemata that are retneved from long

term memory in response to the problem's components. tn this thesis, it is argued that the

structural and procedural knowledge of schemata and mental models are isomorphic in

quality. It is aiso argued that mental models are expressed in concept maps. When the

structural and procedural knowledge of concept maps are assessed they are found to exhibit

isomorphism. The hdîng is taken to indicate that the knowledge associated with concept

maps has properties that compare to the structural and procedural knowledge of mental

models. By extension the finding suggests that concept maps have properties that compare to

schema theory's description of schemata Evidence for the psychologicai validity of schemata

is therefore supportive of arguments for increashg the use of concept maps in education.

2.3 The cognitive and metacognitive processes of concept mapping

The argument that concept maps emulate theories of cognition would be greatly
assisted if the results of the structural-link-proceduraiknowledge analysis could be

triangulated in some way. A triangulation argument is developed here by integrating theories

of reading comprehension (Jacobs and Paris, 1987 and Kirby, 1991) with concept map

theory. Again the basis for integration is that tiinctionaily equivalent components (in difiierent

theones) are actually the same phenomenon. This second integration reveals that artifacts of

the cognitive processes of reading comprehension are retrieved fiom long tem rnemory and

expressed in concept map links. More specifically, integration suggests that the cognitive

processes that govem reading comprehension are aiso active during the creation of concept

maps. Reading t heory stipulates t hat cognitive and metacognitive processes are hierarchical

and proportionate. If the rationale for integration is correct then the cognitive and

metacognitive processes of concept mapping should exhibit comparable qualities. Such a

hding would indicate that concept map components behave in the same manner as their

wunterparts in theuries of reading comprehension. By extension, this finding would also

Sugged that concept maps have a psychological validity that is comparable to that of the

reading theories.

An additional revelation can be drawn from the triangulation argument. Tnat is that

the link-procedural knowledge of concept maps is fhctionally equivalent to the cognitive

processes of reading comprehension. &en the contention that fiinctionaily equivalent

components are t h e same phenomenon, this revelation is said to indicate that iink-procedural

knowledge represents the cognitive activity of concept mapping. Thus the tnangulation

argument has a common component in hic-procedural knowledge. The argument is that it is

isomorphic to structural knowledge and that its quality is comparable to the metacognitive
processes of concept mapping.

Metacognition and cognition are interactive processes that adapt dynamicaily to the

needs of comprehension while learning occus. To Jacobs and Paris (1987), metacognition is

a general concept that concerns the lemer's knowledge of cognitive tasks such as reading,

memory and leaniing. They also say that it comprises executive arategies that plan and

monitor thinking. Kirby (1 99 1) cornplements this reasoning by maintaining that

metacognition operates on a higher level than cognition. He posits that metacognition and

cognition operate hierarchically. Metacognition guides the selection and operation of

cognitive processes. Thus, the application of a metacognitive strategy results in the adoption

of a proportionately complex cognitive strategy.

A 13 question i n t e ~ e w(section 7.1 1) was developed to masure the metacognitive

activity of concept mapping. Method 3 was used to assess cognitive processes. Given the

integrated nature of cognitive and metacognitive processes, it was anticipated that students

with high Method 3 scores would also score well on the UrteMew. Such a finding would

support the rationale for theory integration. That is, it wouid imply that concept map

components exhibit the sarne properties as theù counterparts in theories of reading

comprehension. if equivalent components exhibit sirniiar properties in merem theories then

it is reasonable to argue that the psychological validity of the theories is comparable.

2.4 The structure of the concept map ünlr

The three analytical methods require that concept map Links be constnicted in a

particular way. This Link structure is now discussed dong with an introduction to the nature
m r e 2, The typicai link structure that is requireâ for this thesis

and link-procedural kwwledge. The top half of Figure 2 is an illustration of the ideal link

structure. ffere, concepts are c o ~ e c t e dby a propositional Linking statement ( ~ t n i c t ~ d

knowledge). This linking statement summarizes the mapper's understanding of the tink. Items

of data are subsumeci beneath it. Each d a m describes a component of the relationship that is

expressed in the linking statement. Linking statements are formed by generating an inference

that connects data. The connedon should reveal the nature of the relationship that associates
the concepts of the link.

A hypothetical link comecting the concepts Aneries and Veim provides a concrete

example of the requisite Iùik structure. It is also used to illustrate the nature of link-

procedurai knowledge.
Linking Statement: Arteries and veins are co~ectedby capillaries.

Data:

a. Blood passes through a continuous system of arteries, veins and capiilaries.

b. Arteries cany oxygen rich blood f'rom the heart to the tissues.

c. Veins cany oxygen depleted blood from tissues to the heart.

d. Capillaries are the site of oxygen exchange.

Suppose that this link was created after reading a text that provided the data presented

above. Suppose dso that this text fded to explain the structural arrangement of arteries,

veins and capillaries. Finally suppose that the mapper's objective when creating the link was

to determine how blood flows from arteries to capillaries to veins. The following paragraph

hypothesizes about how the scorer of a concept map might associate the data and linking

staternent of this link to deduce the fùnctioning of the mapper's link-procedural knowledge.

Link-procedural knowledge can be identified by reasoning as to how data were

co~ectedto fom an inference about the structural arrangement of artenes, veins and

capillaries. For example, in seeking a co~ectionbetween Arrrries and Veim, Link-procedural

knowledge might have first considered the oxygen status of blood in artenes and veins. Next

the second and third data items may have been considered as they reveal that arterial blood

bas an abundance of oxygen and that vascuiar blood has a dearth. Next, the knowledge that

capillaries are the site of oxygen exchange rnight have been added to the equation . When the

knowledge that arteries, veios and capillaries fom a continuous system is added to the

picture, an inference about the structurai arrangement can be formed. This iaference rnight be
 i8

"if artenes have lots of oxygen and veins have ody a little, then oxygen must be lost or

gained as blood passes between them. If capillaries are the site of oxygen exchange and blood

passes through a continuous systern, then capillaries mua comect artenes and veins."

Aithough this example is a simple one, it illustrates how link-procedural knowledge seeks

connections between items of data. tt also illustrates that al1 connections across data are

considered when linking statements are fashioned.

Many of the maps created for this thesis contained over 25 links. Placing 25 linking

aatements and their data on a map can lead to clutter so that the logic of the mapper is

difficult to follow. To reduce this tendency toward clutter, the hidents of this study were

asked to number the links of their maps and place the linking statements and data on the

reverse side of the page. This is the Link structure illustrateci in the bottom half of Figure 2.

2.5 T h m methods for concept map a n a s i s and an interview script

The second goal of this thesis is to develop methods for concept rnap analysis. Three

methods are developed in total. Method 1 assesses the qudity of structural knowledge. It

also assesses integrative reconciliation. Integrative reconciliation refers to increases in the

undemanding of schema connections, which occurs as a result of the process of learning. The

renilts of integrative reconciliation are expresseci in the propositional links (stnictural

knowledge) of concept maps.

In assessing integrative reconciliation, Method 1 quantifies the effect of the cognitive

nrain that is placed on working memory during link creation This quantification is achieved

by comparing Illik descriptions to a £ive stage hierarchy of cognition. The hierarchy in

 19

question was devised by Biggs and Coltis (1 982) and it is referred to as The Structure of t k

Obserwd Leming Outcorne or SOLO.From lowest to highest the SOLO stages are:

presmictural, unistructural, rnultistnictural, relational and extended abstract. Successive

stages involve the juggling of evermore data in working memory. They also require that data

be integrated with increasing creativity. The higher the SOLO stage, the higher the quality of

the link. The higher the quaiity of the link the greater the quality of the integrative

reconciliation.

Method 2 assesses progressive differentiation. Progressive differentiation refen to the

thinking that concepts are never fully leamed. Evermore links can be added to a node as

leamhg progresses. Progressive dserentiation manifests in concept maps as the number of

links associated with a concept. Method 2's scoring system refen to concepts with just one

link as I-Link concepts. A concept with two links is called 2-Link concept. The more links

the higher the Number-Link score and the greater the progressive dserentiation. Method 2 is

not concerned with the argument for the psychological validity of concept rnaps. It is used

alongside Method 1 to complement the evaluation technique of Novak and Gowin ( 1984).

Novak and Gowin's assesment strategy is often quoted in current research efforts. It requires

the enurneration of links in a rnap. The resulting score combines the effects of integrative

reconciliation and progressive düferentiation. Methods 1 and 2 untangle Novak and Gowin

scores to provide assessments of the integrative reconciliation and progressive difrentiation

associated with individual concepts.

Method 3 assesses the quaiity of Link-procedural knowledge. Link-procedural

knowledge is the knowledge that generates the inferences that comect the data items of links.
Linking statements summarize the understanding that is acquired fiom the actions of link-

procedural knowledge. Assessments of link-procedurd knowledge are achieved by

comparing link descriptions to a d e that recognUes five levels of inference quality. This

scde was developed by integrating the work of Biggs and Collis (1982) and Rumelhart and

N o m (1978). Descriptions of the inference levels and illustrative example links are

provided below. Each example link describes the interaction of factors that govem the

morphology of lakes (lake morphology refers to the shape of the undenvater basin). Although

the example links contain inferences of different quality, they were al1 created fiom the

following statement "Lake morphology is a function of undenvater contour lines, the shape

of the lake, and geologic origin" (Home and Goidman, 1994, p. 14). The concepts Lake and

Mwphougy were used to create examples of level 1, level2, leveI3, and level 5 inferences.

An example level4 inference was created with the concepts Lake and Water Sbucture.

Level 1. At this level inferences involve incorrect reasoning. Two scenarios are
possible. In the k s t scenario, the inference may connect two items of data that are
not tmly associated with the concepts of that iink. Although the inference may be
correct, the nature of the data involved renden it irrelevant to the concepts of the
link. In the second situation data may be appropriate but the inference is incorrect.

Example link:
Linking statement: Lake morphology is determined by the shape of the lake, its
geologic origin and prevailing winds.
a) Morphology is determined by the lake's shape.
b) Morphology is determined by geologic origin.
c) Morphology is determined by prevailing winds.

The third datum, datum "c" is incorrect. Prevailing winds affect the structure of lake water

but they do not impact on morphology. The inference of this link would be rated as incorrect

as it involves a single moneous datum.

 Level2. Simple listings of facts that fail to connect &îa through inferences.

Example link:
Linking statement: Lake morphology is affecteci by the shape of underwater contour
lines. It is also affecteci by the shape of the lake, and it is also a f f e d by the M e ' s
geologic origin.
a) Morphology is determined by the laketsshape.
b) Morphology is determined by geologic origin.
C) Morphology is determined by contour lines.

The data involved in this link are correct. The level2 rating derives from the structure of the

linking statement. It lists component data but rnakes no attempt to integrate them. The

integration of data constitutes the essence of an iderence. The absence of integration implies

that no inference was fonned. In a sense a level2 rating constitutes fded cornprehension as it

indicates that the mapper lacked the understanding to mate an iderence.

Level3. Inferences based on fact (inferences that replicate descriptions found in the
instruction)

Example tink: Lake morphology is determined by the interaction of undenvater

contour lines, lake shape and geologic ongin
a) Morphology is determined by the lake's shape.
b) Morphology is determined by geologic origin.
c) Morphology is deterrnined by contour lines.

This link replicates the inference that was presented in the instruction. It integrates data in

such a w a y that the instruction is summarized.

Level4. Merences based on facts that were not ovenly linked to the concepts in the
instruction (correct reasoning that builds upon information presented in the
instruction).
 Example tink:
Linking statement. Once the lake basin is formed, physical, chemical and biological
factors interact to produce discemible structure within the water.
a) Water structure develops d e r the basin is fomed.
b) Water structure is determined by physical factors.
C) Water structure is determined by chemical factors.
d) Water stmcture is determineci by biological factors.

LeveI4 inferences associate concepts in the instruction to related concepts that the mapper

was aware of before the instruction was made available. In this case the concept Water

Stmcl~rewas retrieved fiom the mapper's long term memoty and associated with Water. A

level4 rating is applied whenever extant knowledge is associated with information presented

in the instruction.

Level 5. Inferences that improve upon the instruction's descriptions of relationships.

They contain reasoning that could not have been Uiferred fiom information presented
in the instruction.

Example link:
Linking staternent: Lake morphology is determined by the interaction of underwater
contour lines, lake shape, geologic origin and sedimentation.
a) Morphology is determined by the lake's shape.
b) Morphology is determined by geologic origin.
C) Morphology is determined by contour lines.
d) Morphology is determined by the sedimentation of biologicai matter.

This link incorporates extant knowledge of the deposition of biological matter into the

instruction's description of the Lake-Mophoiogy link.

The interview script was developed fiom the work of Paris and Jacobs ( 1984). These

workers developed a series of questions to assess the metacognitive activity of reading. I"neir

questions were adapted to assess the metacognitive activity of concept mapping. Paris and

Jacobs also provided example answers for the thirteen questions. The answers were also
 23

adapted to the interview. Three example answers were developed for each question. Student

answers were rated against the example answers. Students were interviewed individuafly and

each intemiew was tape recorded.

2.6 The ernpirical component o f the study.

The ernpirical component of the t hesis involved 27 rnembers of a grade 9 Biology

class. The study began with the author providing a short lecture to introduce concept

mapping techniques. Students were then asked to read a target text and create a map of it.

The process took place over two consecutive class periods (the Thursday and Friday of a

school week). The target text concemed water pollution and oil spills. When the maps were

complete, the author rated each of them by methods 1 and 3. A positive correlation was

obsened between Method 1 and Method 3 scores. This finding was taken to be supportive

evidence for the notion that the structural and procedural knowledge of concept mapping are

isomorphicdly developed phenornena. Methods 1 and 2 were used to evaluate the

progressive differentiation and integrated reconciliation associated with individual concepts.

Once the maps were rated, six subjects were selected for metacognitive inte~ew.

The interviews were run over three consecutive lunch houn (the Tuesday, Wednesday and

Thursday) in the week foiiowing the mapping exercise. One interviewee was selected as he

created a map with low Method 3 scores.Two inte~eweeswere chosen as they produced

maps with medium scores. The remaining inte~eweeswere used as they produced maps

with high Method 3 scores. It was a s m e d that high Method 3 scores were achieved by

applyiag superior cognitive processes toward the creation of concept maps. It was aîso
 24

assumeci that the higher the interview scores the greater the quality of the metacognitive

processes that were put toward concept map creation. When the Method 3 and i n t e ~ e w

scores of individual students were coneiated they were found to be comparable. The 6nding

was taken to be supportive evidence for the thinking that cognitive and metacognitive

processes have comparable quality. It also indicates that concept maps emdate behavioun

predicted from the integration of concept map theory and theories of cognition. Finaily it

supports the contention that concept maps represent mental models.

 Chapter 3

Concept Maps and Mental Models

3.0 Integrating theories of cognition and concept map theory

The strategy for integrating concept map theory and the four theories of cognition

(stnictural knowledge theory, schema theory, semantic network theory and spreading

activation theory) can now be descnbed in detail. This description assumes that mental

processes operate as per cognitive information processing theory (CP).CIP (described by

the Education Department of the University of Indiana at:

http://www .educ.indiana.edu/-p54O/webcoumhiphd)recognizes three components of

memory, a sensory register, long term memory, and working memory. Information passes (by

way of the sensory register) from the environment to working memory in response to a

problem. The problem may also stimulate information retneval from long term memory. In

working memory relationships are sought between the information cornponents. Once

identified these relationships permit the construction of new knowledge structures that can be

saved to long term memory.

Theory integration begins by s u d n g semantic network theory and spreading

activation theory. These summaries explain how concepts are retrieved tiom long term

memory and expressed in concept maps. Integration then identifies relationships between

structurai knowledge and schema knowledge. The relationships identifieci are used to develop

a new interpretation of stmctural knowledge wherein it f o m mental models ofproblems. It

is argued that these mental models guide the actions of procedural knowledge. Imegration

also reveais paraiiels between schemata and concept maps. These p d e l s are used to
 26

propose a method to investigate the contention that concept maps represent mental models.

3.1 The structure of memory and concept retrieval

According to Ashcrafi (1989), Collins and Quiiiian (1972) provide a description of

memory structure. They envision a network of sernantically arranged and interconnected

concepts. Semantic relatedness govems the positioning of concepts so that categories of

related concepts are formed. Each category has a superordinate descriptor concept and a

number of subordinates. The more a subordinate concept is accessed, the shoner its link to

the superordinate descriptor. A networked nructural arrangement is also favoured by

woikers such as Marshall (1995), as it allows for the human capacity for rapid information

retrieval.

Leaming results in changes in the arrangement of concepts in memory. Regardless of

the structure of knowledge the processes of retrievai are the same. A description of concept

retrieval is provided by the Collins and Loftus' ( 1975) interpretation of Quillian's (1966)

spreading activation theory. According to Coilins and Loftus, Quillian said that activation of

semantic memory begins with the recognition of input words. Starting nodes become

activated when input words are rewgnized. This activation is foliowed by a search of

semantic memory. At first nodes that are directly attached to the starting nodes become

activated. From there the activation spreads outward to nodes that are attached to the nodes

that are attached to the starting nodes. Activation may continue in this rnanner unal several

schemata are activated-

When a node is activateci, a "tag" is attached to it that hdicates which node was
activated previously. The tag also rnakes note of the node where activation began (the

starting node). Eventually a node is located that is on the path of the spread of activation

from two or more starting nodes. This intersecting node is evaluated by tracing tags back to

the starting nodes. Once the starting nodes are located the evaluation determines if they are

semantically matched. For exarnple, the input words Full and l e m s may stimulate activation

at the nodes Fall and lemes. This activation may spread to the intersection node Fall leaves.

Tags are traced fkom Fall leaves back to the starting nodes and contexts of al1 three are

compareci. If the concept Fall and the intersection Full lemes both referred to the third

season of the year a point of intersection would be confirmeci. If the input word fdlreferred

to the effect of gravity the intersection Falf ieaves would be rejezted. Intersections are

evaiuated to accommodate multiple meanings of the same terms. Several intersections may be

rejected before the correct one is located.

Collins and Loftus consider the process of intersection evaluation a Little further.

They say that every intersection located in a memory search provides either positive or

negative evidence of semantic matching. For example, locating the concept c u n q is strong

negative evidence for the suitement ail bir& me robinç. Collins and Loftus propose several

processes for evaluating intersections. Each involves the nimmation of evidence until a

positive or negative threshold of activation is reached. Once the threshold is exceeded, a

decision about a semantic match is made. Only concepts and links that reach a positive

activation threshold can be recaiied to w o r h g memory.

Spreading activation theory reveals the means of concept map creation. Concepts

retneved fiom long term memory are ananged in working memory to form a mental mode1 of
 28

a problem. These models are expressed as concept maps. The following sections interpret

structural knowledge theory and schema theory. These interpretations reveal the processes

behind the creation of mental models. They also indicate the role that mental models play in

problem resolution.

3.2 Structural Knowledge

Jonassen. Beissner and Yacci ( 1993) provide a description of structural knowledge

theory that is adaptable to our needs. They say that dalarative knowledge is knowing what,

procedural knowledge is knowing h m , and structural knowledge is knowing why. Mer

reviewing several studies, Jonassen et al. concluded that anictuml knowledge mediates the

acquisition of procedural knowledge and that it is important to problem solving success.

ûther workers add weight to this conclusion. For example, Shavelson (1972) and Thro

(1 975) determined that a student's knowledge structure cornes to resemble that of the

instnictor during leaming. Gordon and Gill(1989) found that problem solving was facilitated

when knowledge structures resembled the insmictor's. Chi and Glaser (1985) deduced that

pertinent knowledge structures promote effective problem solvhg. Robertson (1990)even

concluded that havhg the right structure is more important than ability when it cornes to

solving physics problems.

Mandlef s (1983) study (as cited by Jonassen et al.) claimed that the understanding

expressed in concept relationships gives meaning and structure to knowledge. Presrmiably

then, Mander wouid hold that relationships (stnictural knowledge) contain concept meaning

or more accurately the leamer's understanding of concepts. Jonassen et al. go so fàr as to say
that an understanding of the relationships between concepts are essential for the

comprehension of ideas. If the above is true it implies that structural knowledge is the

platform on which successful problem solving is built. Successful problem solving requires

the application of procedural knowledge. Structural knowledge provides a guiding

h e w o r k for the actions of procedural knowledge (Winograd, 1977). Jonassen et ai. say

that stmctural knowledge manifests in concept maps as descriptions of the relationships

between concepts. That is, it manifests in concept rnaps as the links between concepts. An

examination of concept map links should therefore reveal the quaiity of structural knowledge

and the likelihood that procedural knowledge is or can be acquired.

3.3 A problern with the description of structural knowledge

The Jonassen et al description of structural, declarative and procedural knowledge

implies a continuum of knowledge development. In this continuum, knowledge is acquired in

a declarative format. It passes then through a structural phase before it reaches maturity as

procedural knowledge. Kooeman and Jonassen (1 994) add weight to this thinking. They

clah that besides knowing whar, students have to know why before they can know how. This

argument is flawed. Someone can be told to change the oil in a car and they can be told how

to do it. Someone can be told how to wash their hands before preparing a meal. in both

cases, knowing why is not an essentid prerequisite to the performance of the task.

It is possible to infa how not telling sorneone why a task should be performed will

impact on their concept map of a topic. Suppose a person firom a mie with no knowledge of

machines was told how to change a car's oil. Their concept map would likely be a series of
 30

boxed concepts (car components involved in the oil change)joined in a linear fashion by

arrowed links. Link explanations in this case would be non-existent. If a person with

experience of machines other than cars were asked to create the same map then their

knowledge of the functions of oil would likely be incorporateci into the links. The greater

their understanding of oil's role in the functioning of machines, the more detailed their link

explanations. The mapper automatically associates the car's oil change procedure to other

machine systems and expresses this association in the links of the oil change map.

Three things are apparent kom the above. First, explanations of concept relationships

are found in concept map links. That is, the knowledge of why (stmctural knowledge) is

found in concept map links. This supports Mander's ( 1983) conception of structural

knowledge. Second, stmcniml knowledge is not intermediate to procedural knowledge. It is

a dEerent component of knowledge that is concemed with conception rather than

performance. Third, if the why is missing from an explanation it rnay be inferred from existing

knowledge. Inferences can be made as soon as a procedure is understood. That is structural

and procedurai knowledge are, or can be, acquired simultaneously.

3.4 Schema theory

Marshaii ( 1995)traces the meaning of the word scherna back to the ancient Greeks.

Plato and Aristotle thought that a schema was an aspect of memory that coaains a

phenornenon's characteristics. These characteristics were organized in memory to form a

hnework so that similar phenornena could be recognked in the £Ùture. The modem view of

schemata is that they assist in the resolution as weil as the recognition of problems.
 31

Marshall goes on to describe the contribution that Bartlett and Piaget have made to

Our current understanding of schema According to Marshall, both of these psychologists

thought that schema knowledge consists of generalized data denved fiom multiple, similar

interactions with the environment. Marshall provided examples of repetitious interactions that

lead to schema formation: the joumey to work, cooking a favourite meal, and solving

mathematical problems. During each performance of an action information is extracted and

added to the schema. ui the early stages of schema formation the information that is extracted

is specific and concrete. Increasingly abaract information is extracted and added to the

schema over repeat exposure to the action. Eventually the schema contains a generalized or

stereotypical representation of an action or problem. Marshall proposeci that people develop a

separate schema fot different repetitious actions. Although people have similar SChemata for

sirnilar actions, dserences arise in response to different prior knowledge as well as different

experiences with the action.

3.5 Frames

According to Manhail (1995), three bodies of work, (Minsky, 1975.;Schank, 1975;

and Rurneihart and Onony, 1977) have been the primary infiuences on schema theory

development. This section concerns Minsky's work with h e thmry. The work of Schank

and Rurnelhart and Ortony is considered in section 3.6. These works are the foundation

blocks of Marshall's schema theory. In tum Marshall's theory serves as a building block for

the descriptions of structural and link-procedural knowledge that are developed in this work.

Like schemata, h e s are theoretical data structures that hold large chunks of reiated
 32

information. According to Marshall, Minsky held that vimially any repeat expenence qualifies

as the subject of a frame. Section 3.4 makes note of Marshall's claim that schemata form ffom

repetitious acts such as the joumey to work, cooking a favourite meal etcetera. In Minsky's

theory these acts lead to the formation of frames. Like schema theory, h e theory hoids

that the typical characteristics of an action (or problem type) are extracted from repeat

experiences and nored as the default values of the fiame. Each action or problem type

generates a separate hme.

Schemata and frames are similar in that both can evolve from almost any repeat

expenence. Both hold generalized data that is extnicted d u ~ repeat

g exposure to a problem

type. Before a frame can be applied, the fixed attributes of the problem are matched against

the generalized characteristics of various m e s . Only the fiame that matches the problem's

fixed attributes is applied. During problem resolution the unique aspects of the problem, its

variables, replace the h e ' s default values or dots. If information about problem variables is

incomplet+ the fhne may still be applied by replacing information gaps with default values.

The mechanism that lads to schema application is similar. Various schemata are matched

againa a problem to determine its type. During schema application, default values may be

used where problem information is lacking.

3.6 Scripts

Another variant on the schema theme is Roger Schank's scripts. A script is "a special

data structure containing a specific sequence of events in a wel understood context"

(Marshall, 1995, p. 19). Scripts are concemed with repetitious social events whose
components unfold in a specific order. Acwrding to Marshall, Schank's ( 1975) study

provided a restaurant script as an example. The fixed data of the restaurant script data

unfolds sequentially as the meal progresses. For example, the menu arrives before the food

and the food precedes the tip. Variable data such as the type of food and the particular waiter

are integral to the script just they are in a frame. Marshd viewed scripts as subsets of

schernata that lack the flexibility to resolve new situations and problems.

Marshall embraced M e theory and script theory for their ability to partially explain

her conception of schemata. Rumelhart and Ortony (1 977) provided the following points for

her schema theory:

1.Schemata are data structures that represent generic concepts stored in memory.

2. Schemata exist for generalized concepts that underlie objects, situations, events,

actions, and sequences of actions.

3. Schemata are not atoMc. Each contains, as part of its specification, the network of

interrelations that is believed to generally hold the constituents of the concept in

question.

4. Schemata in some sense represent concept stereotypes. Accordhg to this thinking,

a schema is a structure that contains a generalwd interpretation of a situation, action

or event. The interpretation is made up of regularities that are extracteci and

abstracted during repeat exposûe to an action type. These regularities CapNe the

essence of the abject matter and as such they are thought to be stereotypical of the

action-
3.7 Marshall's schema theory: A sumrnary

Marshall's work culminated in a comprehensive definition of schema. This definition

takes from the work of Minsky (1 979, Schank (1 973, Rumelhart and ûrtony (1 977) and

Winograd ( 1977). It is presented below.

A schema is a vehicle of memory, allowing organization of an individual's similar

experiences in such a way that the individual:

* c m easily recognize additional experiences that are also similar, discnminating

between these and ones that are dissirnilar;

*cm access a genenc framework that contains the essential elements of al1 these

similar experiences, including verbal and nonverbal componems;

* c m draw inferences, make estimates, create goals, and develop plans using the

fiarnework;

*cm utilize skills, procedures, or rules as needed when faced with a problem for

which this particuiar framework is relevant. (Marshall, 1995, p. 39)

3.8 Iategrating scherna theory and structurai knowledge theory

Marshail's schema theory is readily applicable to the present study. Like Collins and

Quiuian (1972), Marshall says that schemata and their contents are interrelated through a

nested structural arrangement. The idea of nesteâ schemata aiiows concepts (nodes) to be

f'unctional components in more than one schema The understanding (of an action) containeci

in schemata is represented by concept knowledge and schema knowledge. Marshall holds that
 35

concept knowledge is the knowledge extracted during early exposure to an action. It evolves

into schema knowledge as experience is acquired. Marshall compares concept knowledge to

a dictionary definition that can be looked up quickly to differentiate one concept from

another. She describes schema knowledge as a higher fonn of knowledge that allows one to

reason about a concept and solve problems. Concept knowledge declares what a concept is

and schema knowledge elaborates its meaning.

Marshall's interpretations of concept and schema knowledge are obviously similar to

the Jonassen, Beissner and Yacci (1 993) interpretations of declarative and stmctural

knowledge. Declarative knowledge was referred to as the knowledge of whaf. It declares

what a concept is in dictionary definition format. Structural knowledge was calied the

knowledge of why as it described interrelationships between concepts. Marshall's concept and

schema knowledge are fùnctionaliy equivalent to Jonassen et al's declarative knowledge and

structural knowledge. This functional equivalence is interpreted to mean that declarative

knowledge is concept knowledge and that structurai knowledge is schema knowledge.

Marshail was also influenced by Wînograd (1977). Winograd intimated that schemata

are primarily composed of declarative knowledge. More important for this work, Marshall

concurred with Winograd's suggestion that a schema's declarative knowledge "serves as a

guide for procedural knowledge" (Marshall, 1995, p.24). Jonassen et al. suggested a similar

fiinction for stmctural knowledge. They said that it mediates the acquisition of procedurd

knowledge and that it is essential for the comprehension of ideas. These descriptions hint at

the fbnctional equivalence of declarathe knowledge and structurai knowledge. Both

knowledge forms scafFold the actions of procedural knowledge.

 36

The proposed functional equivalence of structural, declarative and scherna knowledge

lends itself to a deeper interpretation of the nature of structural knowledge. In this

interpretation, structural (schema) knowledge is said to evolve fiom declarative (concept)

knowledge during repeat performances of an action. During its evolution, structural

knowledge increases in quality and in the degree of its abstraction.

3.9 The scaffolding o f proceduril knowledge

We are interested in plan development and ski11 utilization once the generic

framework of a schema has been accessed. The reason for this interest is that the processes of

plan development and ski11 utilkation can be used to speculate about the scafTolding of

procedural knowledge. Marshall proposes that schemata contain several kinds of knowledge:

identification, elaboration, planning and execution knowledge. When the constraints of a

problem enter working rnemory, they are compared to identification knowledge in a process

known as pattern matching. Elaboration knowledge is primarily declarative knowledge that

derives fiom repeat exposure to situations that stimulate schema development. Once a

problem/situation type has been recognized by pattem matching, elaboration knowledge is

used to forrn a mental model of the current exemplar. Elaboration knowledge is aiso used to

compare the mental model to the schema's template of this problem type.

The application of identification knowledge generates a hypothesis about the nature

ofa problem. Elaboration lmowledge tests the hypothesis to see ifit is correct.A schema can

be adopted with somejustification once the problem's type has been identifid correctly. The

adopted schema is applied with planning knowledge that breaks down the resolution of a
 37
problem into goals and sub-goals. Planning knowledge becomes more efficient with repeated

use of the schema. The goals and sub-goals of planning knowledge are carried out by

execution knowledge.

An example serves to illustrate how Manhall's different knowledge types fùnction

together. Solving the addition problem, 9 plus 7 plus 5, begins with the application of

identification knowledge. The problem is identified as one of addition. Next, elaboration

knowledge is used to fom a mental mode1 of the problem and venfL that the problem is

indeed one of addition. Once it has been venfied that the problem's type has been correctly

identified the addition schema is adopted. The addition schema's planning knowledge breaks

resolution of the problem up into sub-goals. For example, "If there is only one coiumn of

numbers, add the second digit to the first to generate a subtotal. Add subsequent numbers

one by one generating subtotals with each addition untii al1 the numbers have been added".

Execution knowledge resolves the problem by carrying out the sub-goal steps that planning

knowledge generates.

The above interpretation of Marshall's theory indicates how stmctural knowledge

U o l d s or guides the actions of procedural knowledge. We have said that structural

knowledge is a sophisticated form of declarative knowledge that is essentiai for the

comprehension of ideas. Marshall says that elaboration knowledge is a form of declarative

knowledge. Identification knowledge must be declarative in nature as it is concemed with

knowing whu? type of problem has been encountered. Thus identification and elaboration

knowledge are forms of declarative or structural knowledge. Their actions culminate in the

creation of mental models in workuig memory.

 38

Jonassen, Beissner and Yacci ( 1993) called procedural knowledge the knowledge of

h m . Planning and execution knowledge are fiinctionally equivalent to procedural knowledge,

they identie a plan for solving a problem and then they carry out the steps of those plans.

The plans fomed by procedural (planning or execution knowledge) knowledge are based on

the mental model that forms in working memory. The implication taken from this is that

anictural knowledge scaffolds procedural knowledge by way of mental models of problems.

The case of an oil change for a car can be used to illustrate the scaffiolding of

procedural knowledge. The problem of changing a car's oil begins with the identification of

concepts that are involved in the oil change. The concepts are retrieved to working memory

and relationships between them are used to assemble a mental model of the oil change

process. An oil change concept map is an expression of this mental model. Tactile procedural

knowledge uses the mental model to plan and cany out the oil change procedure.

The scaffolding process requires a stmctural knowledge whose development is

related to that of procedural knowledge. If srnictural knowledge were less sophisticated than

procedural knowledge then the mode1 may lack the complexity that problem resolution

requires. If the structural knowledge were more complex than the correspondhg procedural

knowledge, the mental model could not be acted upon. Thus the contention that stnimrai

and procedural knowledge have comparable qualiàes is a conclusion that arises fiom their

funaional relationship.

3.10 Schema theory and concept maps

The relationship between schema theory and concept maps bas been touched upon
 39

previously. Boxed concepts in concept rnaps have been equated to declarative (concept)

knowledge. Relationships between concepts are describeci in propositional links (stnictural or

schema knowledge). Marshall pennits funher consideration of this schema theory-concept

map cornparison. Her descriptions of the components and elements of schemata wggest that

they are fùnctionally equivalent to the concepts and links of concept maps. For exarnple,

Marshall describes elements as constituents of schemata that contnbute to schema

knowledge. Thus far in our integration of theories, schema knowledge has been equated to

structurai knowledge and structural knowledge has been equated to concept map links. If

elements contribute to schema knowledge they must also contribute to the understanding

expressed in concept map links.

The integration of schema theory and concept map theory would be made easier if

schema components could be equated to concepts in concept maps. Suitable evidence for this

is found in Marshall's contention that comectedness contributes to xhema strength and

cohesion. In addition, Marshall States that the C ~ ~a schema

O M ~ C ~ of ~ S S component refen

to the number of elements associated with it. Kconcept maps containing equd numbers of

concepts are compared, the one with the most links wodd be regardecf as the most comected

and cohesive. Increasing the co~ectednessof a scherna component is therefore comparable

to increasing the number of links to a concept in a concept map. This fùnctiond equivalence

is said to suggest that schema components are concept rnap concepts and that schema

elements are concept map links.

 40

3.11 A concept rnap to iiiustrate the integratioa o f stmctuttl knowledge and schema

theory

Severai revelations about the nature of different forms of knowledge were derived in

the above. These revelations are illustrateci by the blue links of Figure 3. The most important

are that: stnictural knowledge is schema knowledge and that the quality of structural and

procedural knowledge are isomorphic.

Links 1 to 6 illustrate the thinking that concept maps represent mental models. Link

number 1 indicates that concept map concepts have been equated to schema theory

components. Link number 2 serves a similar function by illustrating that concept map links

have been equated to schema elements. The supposition that declarative knowledge and

concept knowledge are fùnctionally quivalent is indicated by links 3 and 4. Links 3 and 4

also indicate that declarative knowledge and concept knowledge both serve as the source of

structural knowledge. Link 5 consolidates the rûwning of links 3 and 4. tt proposes that

dechrative and concept knowledge are one and the same.

Links 1 to 5 permit a cornparison of the declarative-structural knowledge continuum

on the left of Figure 3 to the concept-schema knowledge continuum on the right. Link 6

illustrates the outcome of this cornparison as it equates stmctural knowledge to schema

knowledge. Link 6 seems reasonable as both knowledge types elaborate concept meaning and

both of them develop from declarative knowledge in response to acquired experience.

 Structural KnowlodgF Schuna Tbally

me3*The integration of schema theory and structurai knowledge theory

r

Links 1 to 6 argue that concept maps contain expressions of schema (structural)

knowledge. Wth this thinking in place the scaffolding of procedural knowledge can be
 42

considered. Blue links 7 to 1 2 illustrate the scaffolding of procedural knowledge. Links 7 and

8 point to the supposition that identification and elaboration knowledge are two forrns of

declarative knowledge. Links 9 and 10 have it that procedural knowledge comprises planning

and execution knowledge. Marshall argued that elaboration knowledge is a form of

declarative knowledge used to fom mental models. Links 1 1 and 12 indicate that structural

and schema knowledge are sophisticated foms of declarative knowledge that are used to

form mental models. Planning (procedural) knowiedge uses mental models to determine the

steps for problem resolution. That is, structural knowledge scafTolds procedural knowledge

through mental models of problems.

3. 12 A different interpretation of the nature of knowledge

Section 3.3 (structural knowledge) begins with the Jonassen, Beissner and Yacci

( 1993) contention that knowledge changes fonn as it evolves. Their view is that knowledge is

acquired in declarative format. It then passes through an intemediate structural stage en

route to maturity as procedural knowledge. The integration of the theories of cognition

suggests a different interpretation of the nature of knowledge. In Figure 3 it is proposed that

declarative and structural knowledge are the same f o n of knowledge. They differ only in the

sophistication of their qualities. Figure 3 also proposes a CO-evolutionfor mcnirai and

procedurai knowledge. The intent is not to imply that structural and procedural knowledge

are independent of one another. Rather. knowledge is a singular entity that possesses both

structural and procedurai modalities. Marshall's schema theory (section 3.8) reveais that the

structural and procedural modalities cm be divided further. The structural modality has
 43

identification and elaboration functions. The procedural modality has d e s in planning and

execution. Link-procedural knowledge adds to the conception of procedural knowledge. It is

a subcomponent of the procedural mode that has both planning and execution capabilities..

Link-procedural knowledge was introduced in chapter 2. The link structure that is

needed to elucidate the nature of link-procedural knowledge was also described in that

chapter. Links should consia of a linking statement and one or more items of data. In CIP

tenninolom, items of data represent chunks of information that enter working memory in

response to a problem. The linking statement is an expression of understanding that is fonned

when relationships between items of data (chunks) have been identified. Link-procedural

knowledge is the knowledg that seeks relationships between items of data to form the

linking statement. It generates the understanding that is expressed in structural knowledse

(linking statements).

In combination the structural knowledge statements of a rnap represent a mental

model of a problem. This mental model scaffolds the actions of procedurai knowledge. In

elucidating the nature of the scaffolding process, theory integration indicates that link-

procedural, stnictural and procedural knowledge are isomorphic. This indication derives from

the realization that the scafTo1ding process can only be successful if the t h e knowledge

forms are complementary in quality. Knowing that these knowledge forms are isomorphic

suggests two rnethods for investigating the psychological validity of concept maps. The

relationship between stmcturd knowledge and procedural knowledge can be explored as can

the relationship between amctuml and link-procedural knowledge. The assessments are

equivalent as both involve a cornparison of the structural and procedurai modalities. In both
 44

investigations, finding that the structural and procedural modes are isomorphic indicates that

these knowledge forms emulate behaviours predicted by integrating theories of cognition.

Thus such a findiny supports the rationale for theory integration. By extension, it aiso

sugyeas that concept maps are representations of mental models.

 Chapter 4

The Metacognitive and Cognitive Processes of Concept Map Creation

4.0 Concept mapping and reading comprehension

The argument that concept maps represent mental models would be enhanced if

additional evidence could be found to illustrate that concept maps exhibit propenies predicted

by theories of coynition. This chapter seeks such evidence by integrating two theories of

reading cornprehension (Jacobs and Paris. 1987 and Kirby, 1991). The integration process

hints at properties of the cognitive and metacognitive processes of concept map creation. in

doing so it meets the third goai of this thesis.

Metacognition is seldom defined in the literature in a clear precise manner. Definitions

of the cognitive strategies of leaming are also rare. Scarcer still are studies that clearly

delineate the dynarnic interactive roles played out by metacognitive and cognitive processes

dunng leaming. A search of the literature Failed to locate a description of the roles played

out by metacognitive and cognitive processes dunng concept map creation. This problem can

be circumvented by assuming that concept mapping is controlled by the metacognitive and

coynitive processes of meaningfùl leaming. The assumption is reasonable as the heuristic of

concept mapping is a metacognitive tool for meanin@l learning (Edmonson and Smith

1996).

Concept mapping retrieves and expresses the products of meaningful leaming. If that

learning is acquired from reading then the metaco_oiitive and cognitive processes of reading

comprehension will influence its expression in a map. Jacobs and Paris ( 1987) and Kirby

(1 99 1) provide complementary descriptions of the metacognitive and cognitive processes of

 46

reading comprehension. Their work is integrated here and posited as a description of the

heuristic of concept mappiny.

4. 1 A description of metacognition

To Jacobs and Paris ( 1987), metacognition is an executive process that plans and

monitors thinking. Kirby ( 1991) also maintains that metacognition operates on a higher level

than cognitive processes. During reading (and concept map creation) metacognition guides

the selection and operation of cognitive arategies. According to Kirby. little is known about

cognitive strategies. He holds that interpretations of them depend upon the scale on which

they are viewed. Macroscopically they manifest as ~eneralIeming styles. On closer

examination they appear as the set of processes involved in an instance of reading

comprehension. Just as cognitive strategies are mentored by metacognitive strategies, they in

tum guide the activities of co_gitiveprocesses. Cognitive processes hold units of information

in working rnernory and seek relationships between them. They are fùnctionally equivalent to

link-procedural knowledp.

Jacobs and Park (1987) define metacognition as "knowledge about cogitive States

or processes that cm be shared between individuals" (p. 258). This definition limits

metacognition to coh@ive activities that can be reponed to others. Once an activity becomes

so familiar to the learner that it is applied without conscious thought it ceases to be a

metacognitive tool. For example, a novice leamer may consciously reread words or phrases

to better determine their meaning. When asked ifwords are reread the leamer would give a

positive reply. Rereading in this instance is a metacognitive process. Rereading undertaken by

 47
an expert may be an automatic process that would not be reported in response to a prompt.

Once the action bypasses the awareness of the learner Jacobs and Paris say that it ceases to

be metacognitive.

Jacobs and Pans divide metacognition into two components: self-appraisal of

cognition and selfhanagement of thinking. Each component is funher divided into three

elements. Declarative. procedural and conditional knowledge are elements of the self-

appraisal of cognition. Declarative knowledge is knowing thm. Leamers may know thut

underlininy a word increases retention of it. Procedural knowledge refers to knowing huw.

Leamers rnay understand huw underlining increases retention and recall. Conditional

knowledge is the knowledge of why strategies work and whrtl they can be used appropriately.

Paris?Cross and Lipson (1 984) include declarative. procedural and conditional knowledge

within the rubric "knowledge about cognition." Element s of the self-appraisal of cognition

wiil be referred to as metacognitive declarative knowledge, metacognitive conditional

knowledge, and metacognitive procedural knowledge. The term "metacognitive"

distinyuishes these knowledge foms fi-omthe cognitive equivalents (Jonassen Beissner and

Yacci, 1993) that were defined earlier.

The components of self-manasement that Jacobs and Paris describe are executive

processes. PlamÏrtg is the process whereby a cognitive strategy is selected to achieve a

leaniing goal. Electing to read quickly when an oveMew is required is the example of

planning that Jacobs and Paris provide. Evalrrarion is the process of examinin8

comprehension as leamhg progresses. The third component, regzdation. takes feedback fiom

and adjusts the manner in which strategies are applied. Adjustments are made to
rv~Iucftio~i
meet a desired level of comprehension. Paris, Cross and Lipson incorporate these self-

management components into the category "self-directed t hinking ."

Jacobs and Pans' work can be interpreted to expiain the role played by metacognition

in concept map link creation. At the nart of the linking process, platz~ii~ig
is used to select a

metacognitive strategy. Metacognitive cottditiotiai knowledge is important in strategy

selection as the learner has to know when and why strategies are effective before making a

suitable selection. Strategy selection also requires metacognitive declarutivr knowledse or

knowledge of what the strateg does. The strategy is applied by metacognitive procrcl11rd

knowledge. Durinç concept map creation rvalrcatiml ensures that metacognitive procedural

knowledge is applied correctly. If a problem is encountered the application of metacognitive

procedural knowledge is adjusted via regu1utk.m. Metacognition is primarily concerned with

global aspects of learning. Lrltimately however it exerts an influence on the minutiae of

meaningful leaming (link creation) as it influences the selection of comprehension strategies

and processes.

4.2 Comprehension and reading

Kirby's ( 199 1) reading comprehension theory cm be adapted to explain the

comprehension processes of concept mapping. Kirby ( 199 1) descnbes comprehension as

follows:

The essence of comprehension is the recopition of relationships betweem two or

more separate units of information held simultaneously in working memory. ïhose

 49

elements can be recent input or can have been recalled fiom long-term memory. When

relationships are recoynized, the elements may then be integrated, to form a new,

higher-level unit in memory, which may then be stored or used in subsequent relating

and inteyrating (p. 107).

Kirby argues that reading can occur on ei&t ievels: features. letters, syllables, words,

syntactic chunks (phrases) ideas, main ideas and themes. Beginning readers operate below the

level of words. Their cognitive resources are so absorbed in recognition tasks that they are

incapable of the inregration that comprehension requires. From the word level up recognition

is sufficiently automatic that surplus resources are available for comprehension.

Comprehension begins with the chunking of incoming data from the text. The product of

chunking may be a phrase. an idea or a main idea depending on the proficiency of the reader.

Integration of the chunks cornpletes the comprehension process. For the purposes of this

thesis. the thematic level of reading is thought of as a process that connects chunks to extant

knowledge in other domains. This processing facilitates the use of knowledge in domains

other than that in which it was acquired. Reading to the level of themes is a popular goal in

education as knowledge that is associated with multiple themes has a high degree of utility.

The level at which readers operate depends on interaction between their reading skills

and goals and the complexity of the text. Kirby maintains that adults generally read at the

Ievel of main ideas. At this level little effort is required to chunk incoming data and form main

ideas. Two or three main ideas may reside in working memory simultaneously. Newly input

information is compared to the main ideas and reasons for integration are sought. By
 50

operating at the level of main ideas the learner is able to comprehend the plot of a novel or

the context of an expository text.

Readers operating at the level of main ideas may be unaware of themes if the text

does not point them out explicitly. Functioning at the level of themes requires that the set of

main ideas that make up an idea or notion be present in memory. Concept maps can make a

set of main ideas explicit to the leamer. Reflecting on the main ideas in a map serves as a

staning point for the generation of thematic links. Concept maps provide audents with the

"big picture" and thereby encourage the transfer of knowledge across multiple themes.

Kirby places the how of comprehension under two headings: comprehension

processes and comprehension strategies. Two types of comprehension process are

recognized. Successive processing forms discrete units from incoming information and

maintains them in working memory. Simultaneous processing seeks relationships between

t hese discrete units of knowledge. Kirby contends that sirnultaneous processing can be

bottom up or top down. In bottom up processing two units of information are integrated to

fonn a higher level unit. Top d o m processiny dismantles high level units into their

component parts.

Comprehensions are saved to long terni memory while reading occurs. If a concept

map simply regurgtates comprehensions that were acquired from reading then its expression

will be controlled by the very processes that control cornprehension. To explain. link data

items represent discrete units of information that fom in working memory during the reading

process. They represent the products of successive processing. The linking statement

represents a comprehension; it is the product of sirnultaneous processing. In seeking

 51

relationships between items of data, simultaneous processing is fùnctionally equivdent to

link-procedural knowledge.

Control of reading comprehension is assumed to operate on three levels.

Metacognition guides comprehension strategies and in tum cornprehension strategies guide

comprehension processes. The assumed inteopated arranyement of metacognition and

cognition indicates that complex metacognitive activity results in the application of complex

cognitive strategies and processes. This means t hat measurements of metacognition should be

related to assessments of link-procedurai knowledge. An analytical method is developed to

measure link-procedural knowledye. An i n t e ~ e wis developed to elucidate metacognitive

activity. A positive correlation between the i n t e ~ e wand Iink-procedural knowledge scores

will be taken as supportive evidence that metacognition and cognition are integated,

hierarchicd processes. The finding will also be taken as fùnher suppon for the thinking that

concept maps mode1 the properties of theories of cogition. These theones hold that concept

maps represent mental models, finding that concept maps emulate theories will be taken as

support for this notion.

 Chapter 5

An Introduction to Concept Map Anabsis

5.0 An example concept map

5.0.1 Knowledge to be expressed in the map

Figure 4 was created fiom Brock and Madigan ( 1988). It depicts the three stages of

fermentation in yeast. The ethanol and CO, by-products of this fermentation are useful to the

brewiny industry. Different products are produced when _glucosefermentation is camed out

by other prokaryotes (fermentation agents). For exarnple lactic acid bactena produce lactic

acid. Al1 fementing organisms store fermentation energy by adding phosphate to adenosine

diphosphate (ADP). The resulting adenosine triphosphate (ATP) subsequently releases its

energ to fuel metabolic reactions.

The fermentation of one molecule of gucose is depicted in Fiyure 4. The first stage of

fermentation is a preparatory stage in which glucose is split apan. The energ needed to

perform the reactions of stage I is provided by two ATP molecules. Two molecules of

glyceraldehyde-3-phosphate are formed by the end of stage 1. These products are represented

at the start of stage 2 on the ieft and rîght sides of Figure 4. Each glyceraldehyde-3-

phosphate molecule undergoes the same reactions in the second and third stages of

fermentation.

Stage 2 begins with the addition of phosphate to giyceraidehyde-Sphosphate to form

diphosphoglycerate. This reaction. Wte all the reactions of fermentation, is made possible by

enzymes. Er~ymeschange the rate at which a reaction occurs. Coenzymes are Ioosely

associated with a number of enzymes; they carry smaü molecules between them. in
 53

diphosphoglycerate production two hydrogens are added to two coenzyme NAD+ molecules

to fom NADH. In the second reaction of sta8e 2 phosphate is removed f'rom

diphosphoglycerate to fom phosphoglycerate. Two phosphates are used to fonn two

molecules of ATP. At this stage of fermentation there has been no net loss or gain of energy.

A positive energy balance is achieved in the last reaction of stage 2. The production of

pyruvate generates a net energy gain of two ATP molecules.

5.0.2 The map

Figure 4. The fermentation pathway

The first two stages of fermentation are collectively known as @ycolysis.The third

aage of fermentation is necessary to convert NADH back to NAD+. The yeast ce11 has a

limited nipply of NAD+, if al1 of it were converted to NADH, glucose fermentation would

stop. The third aage also produces ethanol and carbon dioBde (CO3
&a ,.
Concept map o f fermentation

Figure 5 is a concept rnap of the fermentation process. Two links emanate From the

These links terminate at nodes which anchor two regions of the

concept Ferrne~t~n~io~t.

concept map that are separate from one another. A concept map region contains a number of

intercomected concepts. Regions are discrete fiom one another in that few if any links pass

from one region to another. The first region of Figure 5 surnrnarizes the process of

fermentation. It consists of jun two concepts: Erwrgv and A P.These concepts are referred

to collectively as the Smmary region. The other region, the Example region, illustrates the

fermentation of glucose in

yeast . Within the Ewampk region, Stage I , Slagr 2 and Stage 3 and the concepts they

subsume fom t h e smailer regions The importance of regions is discussed in section 1.4.3

Figure 5 was created by the author of'this thesis with WordPerfecttsdraw feature. in

combination with Table I it represents a complete representation of the process of

 55

fermentation. WordPerfect's draw feature does not provide sufficient space for the inclusion

of complete link descriptions. Soflware that is specifically designed for concept map creation

(such as SemNet Academic') is required if complete link descriptions are to be generated

electronically alongside the map. Paper concept maps offer a cheap solution to the problem of

software that offers insufficient space for link creation.

Complete descriptions of Figure 5's links were developed by the author and are

presented in Table 1. The nght hand column of the table requires a little explmation. The top

line of each row in the right hand column is the linking staternent of that link. The numbered

statements beneath the Iinkins statement represent items of data that make up or are

comprised in the linking statement.

The first step the author engayed in when creating Table 1 was to generate linkins

statements for every link. In the next step. items of data were created for each link by

reflecting on the thousht processes that contributed to link creation. Each link was

considered in tum. Each datum represents a single factor that contributed to a link

relationskp. When the data of these links are reviewed. the logic behind link creation c m be

deduced. The product of this deductive process represents the inference or link-procedurd

knowledye that the author used to coordinate link data and yenerate a linking statement. For

example the data of iink 12 describes the reactions that lead to ATP production. Deductive

reasoning reveds that the author realited that if these reactions occur in stage 2 then this

nage m u a be a source of ATP. When marking student's maps assesson use the sarne

'SemNet Academic, intellimation, Dept. 5SCY 130 Cremora Drive, PO Box 1922 Santa
Barbara. CA- 93 1 16-1922
deductive reasoning to determine how the data of each link were combined to form the

linking statement. The explicated reasoning is thought to be reflective of the link-procedural

knowledge that was used to create the link.

Table 1

Descriotions of the Links o f Figure 3

Descriptian

Fcmcritation is a biwhcnical pathway that p r o d u ~ LW@

s
( 1 ) Cilucmc is osidinxi in
pathway
(2) Cstaboli~molosidizd mmpound rcIca.;c=sa m a I l amount of C'IICT~J
Substraic: Irvt.1 phosphonlation prduccs ATP
( 1 )Enqmatiç caiebolim of substrstc gmcrata ATP
(2) -lne prms is caiiai substrate level pho~iphonlation
Ycaist is one of scvcral prokaqotcii that pcrform-s f~mcntation.
( 1 ) Ycmt is one of w a a l prokaqota that pcrfom kmeatatioa.

The Iast stage of firmtatation is stage 3

( 1 )'The Iast stage of krmatation Ls stage 3

Link 6 Stage 1 fcrmtntation is a wrnponent of glywl~sis

( 1)Stage 1 is a Sçna of preparataq rrriction-s that phmphoqlate and 5,plit glucose
(2) No osidrition or duction ~ar;tioascw invohed
(3) No s n q is rcieased
Stage 2 is the last pbaw of g i y w ~ s i s
( 1 ) Stagc 2is an osidatiw ph&w
(2) NAD+ is d u d
(3)ATP is producd
s siage 1 consume 2 ATP molcxules
The preparatory ~ c t i o n of
( 1) The prqaratory s stage 1 c;onsumc= 2 ATP molxules
~ c t i o nof

Link 9
 --

NAD+ LS rduwci in stagc 2 ractioas

( 1 ) As g~cc~aldcbydc-3-pbosphatr:
is osidid NAIN is rc'dud
Stage 2 produçt?i ATP
( 1 )Formationof I .3-diphmq~hogly~ric
acid gtncratt5 orgiinic P
(2)ATP producd from conversion of 1.3-di-P-gl'çcrica d to 3-pho.iphogiyçc.rk acid
(3) ATP p r o d u d h m mautnion of ptro~ph~olp~nivate ta pjnivatr

Stage 3 rc'~torc5the d i s supp- OCNAD+

( I ) Acctaldch~dt.is d u c d bu clectrons d~rivcdtiom NAüt I

Etbanol is an cnd prduct of p!nitrak rcduction

(1)Ethmol is rui end product of p'matc duçtiori

5.1 Novak and Gowin (1984) and the structure of concept maps

Novak and Gowin ( 1984) based their concept map scorins technique on three

Ausubelian notions: progressive differentiation, inteqative reconciliation and hierarchical

cognitive stnicture. Progressive differentiation and integrative reconciliation derennine the

number and quality of links in a concept map. Progressive dxerentiation refers to the

thinking that concepts are never fùlly lemed and that propositional links are continuaily

added to them as learning occurs. Integrative reconciliation is a conscious reasoning process

that facilitates the discovery of relationships between concepts. Newly discovered links may

be unique to their author as integrative reconciliation is a creative process. Integrative

reconciliation aiso has a reflective component that can increase the quality of understanding

and identi@relationships that are inappropriate.

 Novak and Gowin also thought that progressive differentiation and integrative

reconciliation determine the hierarchical structure of concept maps. Dunng concept map

creation, progressive differentiation and integrative reconciliation cause general, inclusive

concepts to appear near the top of the map. Specific concepts are subsumed beneath inclusive

concepts. Figure 5 exemplifies these structurd trends. Inclusive concepts such as Glyco!v.si.s

appear near the top of the map. Specific concepts such as E f h a d appear near the bottom.

5.2 The Novak and Gowin scoring technique.

Novak and Gowin's sconng syaern accounts for the structural influences of

progressive differentiation and inteyrative reconciliation. It enurnerates links and hierarchical

levels. During concept map creation progressive differentiation and integrative reconciliation

cause conceptually related concepts to be placed in close proximity. In consequence concepts

are placed in clusters or regions. In yeneral concept maps regions are largely discrete fiom

one another. Most links are intraregional. they connect concepts that reside within the same

region. Only a few links pass fiom one region to another and as a result concept map regions

are largely discrete fiom one another. Novak and Gowin used the name crosslinks to refer to

these interregional links. They thought that crosslinks form when mappers aep back fi-oma

map and engage in reflective thinking or integrative reconciliation. The reflective infiuence

behind these links means that on occasion they contain examples of creative thinkiny. Novak

and Gowin also were aware that crosslinks may also be lemed Eom instruction. They

suggested that crosslinks be examined carefùlly for evidence of creative thinking. Crosslinks

that contain true original thinking were awarded a higher score than intraregional l i s . By
 59

awarding different scores to crosslinks and intraresional links. Novak and Gowin introduced

the idea that different types of concept map link have different qualities. Method I builds on

this thinkiny as it recognizes five types of link and five levels of link quality.

In Fisure 5 the concepts Stage 1, Stage 7 and Stage 3 are used to anchor three

regions. Unlike the Swnrnay and Fkampie regions. the stage regions are not completely

separate form each other. They are connected by links 1O and 16 which attach the Stage I

region to the Stage 2 resion and the Stage 2 region to the Shge 3 region. These links repeat

information provided in Brock and Madiyan ( 1988).They do not contain creative thinking

and they were not rated as crosslinks.

Novak and Gowin claimed that accurate assessrnent of concept maps requires they

have a hierarchical structure with clearly labelled and categonzed (cross or intraregional)

links. When these conditions are met they said that values can be assi~nedto links and

structures on the basis of reasonable judgement. The followinç pidelines are adapted form

Novak and Gowin ( 1 984). We will use them to score Figue 5 .

1. Give a score of 1 to al1 valid intraregional links. Subtract a score of 1 for each

invalid intraregional link. Enurnerate al1 vaiid crosslinks that do not contain creative

thinking. Award each non-creative crosslink a score of 1.

2. Count al1 vdid levels of hierarchy and allot a score of 3 to each level.

3 - Give valid crosslinks a score of 10. Crossiïnks indicate important integrative

reconciliation. They may be better indicaton of meanîngful learning than the number

of hierarchicd levels.
 60

4. If example links were asked for and provided. award a score of 1 to each level.

5.3 An example Novak and Gowin score

Table 2 contains a Novak and Gowin score for Figure 5 . Some explanation is

required. There are 16 links in total. and five hierarchical levels. A point is awarded for each

link and three points are awarded for each hierarchical level. The final score for Figure 5 is

3 1. Hierarchical levels were identified by counting concept layers. Link 3 /~'rrmri>ratiortto

Yenst) link connects two concept layers. The link fi-om Ymst to (i!tw/jsis provides another

layer and the C'lyco~vsis-StageI and Stage /-il TP links make up two more. Link 3 was

awarded a score of 1 as it is an example link. Links IO and 16 were also awarded a score of I

as they are not crosslinks.

Ta bie 2

A Concept M ~ Scorin
D Sytem lafter Novak and Gowin 1198411 for Figure 4

Structure Amount Value Score

Link 16 1 16
Hierarchical Ievels 5 3 15
Crosslinks iO
O O
,
Total Score 31

Novak and Gowin's sconng technique bundles the effects of progressive

Merentiation and integrative reconciliation to provide scores for entire maps. Only two

levels of link quality are recognized, intraregional links and crosslinks. Methods 1.2 and 3
 61

explicate the cognitive and metacognitive processes associated with individual concepts.

They acknowledge the individual contributions of inteyrative reconciiiation and progressive

differentiation and they recognize five levels of link quality Chapters 6 and 7 discuss these

methods in tùll.
 Chapter 6

The Next Ceneration of Assessrnent Methods

6.0 Concept map assessment methods and the goals o f the thesis

The developrnent of concept rnap assessrnent methods cm now be considered. Three

methods are developed. Method 1. Method 2 and Method 3. Method I assesses the quality of

structural knowledge, Method 3 assesses the quality of link-procedural knowledge and

Method 2 enumerates concept rnap links. The methods are used to meet the first and third

goals of the thesis. In regard to the first goal, ratinys from methods 1 and 3 are correlated to

investigate the thinking that structural and link-procedural knowledse evolve in parallel and

with comparable quality. A positive correlation is said to indicate that concept maps emulate

behaviours indicated by the integration of theories. These theories indicate that concept maps

represent mental rnodels. Therefore a positive correlation is said to be supponive of the

notion that concept rnaps represent mental models.

In regard to the third goal it is argued that by estimating link-procedurai knowledge.

Method 3 also estimates the cognitive activity of concept mappen. An interview is developed

to assess the metacognitive activities of concept mappers. Evidence that Method 3 scores are

proportional to interview scores will be interpreted as supportive evidence for the argument

processes of concept mapping have complementary

that the cognitive and rneta~o~gitive

qualities. Such a finding indicates that concept maps exhibit properties indicated by the

integration of theones. Once again, findins that concept maps emulate properties derived

from theory integation is said to support the thinhg that concept maps represent mental

models.
 63

In an aside to the main goals of the thesis. methods 1 and 2 are used to illustrate the

utility of concept maps. Method 1 is used to assesses the results of integrative reconciliation.

That is, it is used to detennine the quality of aructural knowledge. Method 2 is used to

account for progressive differentiation. That is it is used to enumerate the links associated

with individual concepts. All three methods are based in the work of Biggs and Collis ( 1982).

6.1 The SOLO knowledge assessrnent method

Biggs and Collis ( 1982) proposed a system for identifjmg the quality of students'

responses to questions. They called this system, the Structure of the Observed Learning

Outcomes (SOLO). A description of SOLO is presented here as it provides the reasoning

behind the evaluation strategy of Method I . SOLO assesses the understanding contained in

answers to questions. Assessrnent is achieved by considerin_uthe manner in which information

has been integrated. Method 1 adopts the same strateyy, it considers how items of data are

integrated to fom a linking statement.

Fiyures 6 to 1O illustrate the five levels of response qudity that are recognized by

SOLO. In each of these figures. responses to cues are generated using one of three types of

data: imelevant data not contained in the instruction relevant data contained in the

instruction, relevant data that is not contained in the instruction. SOLO assumes that

response quality is a reflection of the arnount and type of data in a link. Data that are not

contained in the instruction have to be interpreted by the learner. If the learner interprets

these data correctly. the response is said to be of high quality. Lower quality responses are

associated with relevant data contained in the instruction. In both cases the more data
 64

involved the higher the quality. Data may or may not be integrated in the response. A higher

quality response is sipified when data are integrated. Incorrect responses are associated with

irrelevant data. Even if data are integated in an incorrect response the quality is still

considered to be low.

6.2 The prestructunl SOLO response

Figure 6 illustrates the lowest SOLO level, the prestmctural response. Inappropriate

or irrelevant data are used to pnerate a response to a cue or stimulus. According to Biçgs

and Collis, this is the type of response senerated by a student with no prior knowledge of a

subject and who came to an instruction with no intention of learning.

Cue Response

- Hypothetical, not givai

- data

W r e 6. The prestmctural SOLO structure

The foiIowing question wiI1 be used to exempli@aii of the SOLO response levels
 65

"why is Iraq in a bad political situation?" It will be assumed that current news progams serve

as the source of instruction. A prestructurai response wouid be "it is bad because of Iraq's

gold mines"

6.2.1 Capacity, closure and consistency

Capacity refers to the amount of workiny rnemory needed to create a SOLO

structure. In a prestructural response there is no need to simultaneously retain the cue and

response in working mernory and the requisite capacity is minimal. Biggs and Collis also

considered the processes of consistency and closure. Closure refers to the need to corne to a

conclusion. Consistency alludes

to the need to consider al1 the data and ensure that conclusions are correct. In the low level

prestructural response forces of closure predominate and low consistency results. Hiçh-level

Cue

m u r e 7. The unistructurai SOLO structure

 Cue Response

Hypothttical, not @en

data

Figure 8. The multistmcturai SOLO structure

- -

Cue Response

ILb
, Hypothetical, not @en -

Figure 9. The relational SOLO structure

 Cue Response

-R

Hypothetical, not &en

data

Eipure 10. The extended abstract SOLO structure

SOLO structures place greater demands on capacity. These structures are less impacted by
closure than consistency. In fact at the highest SOLO level (extended abstract) closure has no

impact on the response and multiple conclusions may be reached.

6.3 The unistnictural SOLO response

Figure 7 illustrates that the unistructural SOLO response is generated by relating the

cue to a single relevant datum that was provided in the instruction. Biggs and Collis referred

to the integration of eue and data as the process of induction. The arain on capacity is still

low in the unistructural response as only the datum and cue need be manipulated in working

memory.

A unistnictural response to the question would be "it'sin a bad situation because of the oil."
 68

The question acts as a cue and "oil" acts as a datum. The response linked one datum to the

cue.

6.4 The multistmctural SOLO response

Creating the multistructural response of Figre 8 places a medium strain on capacity

as the cue and two or more related items of data are simultaneousiy manipulated in working

memory. A multistructural response to the Iraq question wouId be "it's bad because of the oit

and it is also bad because of Saddam Hussein." The data "oil" and "Saddam Hussein" are

associated with the cue through the conjoining phrase "and it is also." Conjoining phrases

typifi the multistructural response. In hisher level relational links the data and cue are

integrated to form single phrases.

6.5 The relational SOLO response

In Figure 9 induction coordinates most or al1 of the instructional data into a single

response. The strain on capacity is hish in relational responses such as Figure 9. At this

SOLO level the Iraqi question might stimulate the following response "Iraq's troubles stem

from Saddam's pan-Arabian ideoiogy which threatens the World's access to cheap oil." The

response is relational as the data "politics, oil and money" are integrated into a single phrase.

6.6 The extended absmct SOLO response

Capacity is strained to the full in the extended abstract response. Figure 10 typifies

extended abstract thinking It coordinates material fiom inside and outside of the instruction
 69

and it generates multiple logical deductions. These deductions may involve new hypotheses,

abstract principles and analo$es. The example question might elicit the following response at

the extended abstract level "Iraq's troubles stem corn 1923 when the Middle East was

divided to suit the political aspirations of Colonial powers. Recognizing that extemally

imposed land division attenuates Arabian power. Saddam Hussein sought to unite the Middle

East. A united Middle East could undoubtedly change the balance of global politics by

controlling the movement and price of oil. It is the çlobal threat imposed by Saddarn's

ideoloa that makes Iraq's politicai situation so bad."

The above answer is more detailed than that provided at lower SOLO levels but this

is not the reason for the extended abaract rating. Rather the rating is derived fi-om the

association of data such as "the current political situation. oil and money" to the land

divisions of 1923. Current news programs have neglected to repon the impact of Colonial

strategies on current events. Thus it cm be said that the response associates the cue with a

daam that was taken from outside the instruction (current news prograrns). The use of non-

instructional data is typicd of the extended abaract response.

6.7 Trends in SOLO structures and cornparisons to concept map components

The general trend in Figues 6 to 10 is that more data has to be considered alonyside

the cue as the SOLO structure becornes more cornplex. Furthemore, it seems that creatiny

SOLO structures of increasing complexity requires greater consistency and a progressively

less powerfùl need to reach clonire. SOLO armures cm be compared to the components

of concept maps. For example. the cues and responses of SOLO structures compare to the
two concepts involved in a concept rnap Iink. Also the connections between cues and

responses cm be equated to concept map links. The SOLO structure-concept rnap

cornparison suggests methods for concept map assessment. These methods are descnbed in

the next chapter.

 Chapter 7

Three Analytieal Methods and an Interview

7.0 Method 1: An adaptation of SOLO

Method 1 is an adaptation of Biggs and Collis' SOLO knowledge assessment method.

It rates concept map links against the SOLO response scale. Links consist of one or more

items of data and a linkins statement. Each item of data expresses one element of the

relationship between two concepts. Dunng Iink creation, simultaneous processing (link-

procedural knowledge) forms an inference that connects items of data. The understanding

contained in this connection is expressed in the linking statement. Method I determines the

quality of this understanding. It assesses the consequences of the actions of iink-procedural

knowledge.

The SOLO-Method 1 relationship is demonstrated by descnbin~Fi y res 6 to 10 in

terms of CIP and concept map links. Al1 these figres associate a cue and response through

data. The cue and response can be thought of as the two concepts of a Iink. The lines

co~ectingthe cue. the data and the response represent the cognitive processes of link

creation. Each item of data represents a discrete unit of information that has entered working

memory in response to a problem. The lines between cues and items of data represent the

actions of successive processiny. Lines comecting items of data to the response represent the

actions of simultaneous processing or link-procedural knowledge. The representation of

simultaneous processing syrnbolizes the inference that connects items of data and thereby

enables the formation of a Linking statement.

The inference of Figure 7 (unistructurai Iînk) concems one item of data that was
 72

provided to the mapper. The two concepts of the Iink are connected by a single item of data.

Figure 8 (multistructural response) contains a number of discrete information units (items of

data) that were forrned by successive processing. The linkins statement of this link conjoins

each datum with phrases such as "as well as" and "and also." No attempt is made to integrate

this data and form a contiguous expression of understanding. Figures 9 and 10 portray higher

level links. Their linkiny statements are expressions of understanding that were forrned by

integrating data. Figure 1O difTers fiom Figure 9 in that its inference connects one or more

items of data that were not provided in the instruction.

Figures 6 to 10 illustrate that linking staternents differ in quality. The factors that

affect link quality are:

1 . whether data items are correct.

2. whether data items were integrated in the linking statement.

3 . the number of data items.

4. whether the information involved was provided in the instruction.

These factors suggest a method for assessing concept maps. This method. Method 1.

is explored below by examining the linking statements of Figure 5. The method is applied in

four aeps:

I . examine each link to ascertain that its linking statement and datum (or data) are

correct

2. count the data items.

3. ascenain whether the data was presented in the instruction.

4. determine whether the linking staternent has associated its data with
multistnictural. relational or extended abstract logic.

The factors that affect link qudity are the amount of data the manner of its integration and

whether it was provided in the instruction. SOLO sugsests that these factors generate five

levels of understanding: prestnictural, unistructural, unistructural. multistructural. and

extended abaract. A sconng system that recognizes the five SOLO levels is presented below.

7.1 Prestructural links

A prestnictural rating is pven when the inference in a Iink is incorrect. In Chapter 6.

responses were provided to the exarnple question "why is Iraq's political situation bad?" In

tems of concept maps these responses miyht be thought of as linkiny statements that connect

the concepts Iraq and Polirkal Si~i~atiwi.

The prestnictural response of Chapter 6 would

appear in a concept map link as per Table 3. Table 3 represents links in the same manner as

Table 1. The lefi hand column of the table contains the link's number and SOLO rating. The

top line of the right hand column contains the linking statement. Subsequent lines in the nght

hand column represent items of data. There are no preaructural links in Figure 5 .

Table 3

A Prestructural SOLO Respnse Exoresseci in Link Format

1 Link 1 Data 1
x It is bad because of iraq's gold mines
Prestnictural It is bad because of Iraq's gold mines
7.2 Unistructural links

When concepts are associated through a single relevant datum their linkiny statement

is classified as unistructural. The exarnple unistructural response to the Iraq question would

appear in a link as per Table 4.

Table 4

A Unistructural SOLO Response Ex~ressedin Link Format

1 Link 1 Data 1
11 xUnistructural 1 It's in a bad situation because of the oil 1
1 ( I )Ittsin a bad situation because of the oil 1

Table 5 presents three of the unistructural links of Figure 5. In some unistructural

links the data item and linking statement are identical. This is thought to be a reflection of the

fact that these Iinking statements are influenced by just one item of data. Relational linking

statements are not identical to any one datum. Their construction is influenced to varying

degees by each item of data in the link. Some of Table 5's links contain a singe item of data

that differs corn the Iinking statement. These links are thought to be intermediary to the

unistructural and relational stages. Perhaps these linking statements are influenced by

additional items of data that the mapper is incapable of expressing or is only unconsciously

aware of Link 13 is an exarnple of an intermediary link.

Table 5.

Exam~leIlnistructural Links in Fieure 5

1 Link 1 Data 1
1 Link 8 1 The preparatory reactions of stage 1 consume 2 ATP molecules
-- - - - - - - --

1 Unistructural 1 ( 1 ) The preparatory reactions of stage I consume 2 ATP molecules 1

I L i n ki O 1
Unistructural
Link 13
Glyceraldehyde-3-Phosphate is oxidized as Stage 2 begins
( 1) Glyceraldehyde-3-Phosphate is oxidized as Stage 2 begins

Stage 3 restores the cells' supply of NAD+

>
Unistructural ( I) Acetaldehyde is reduced using electrons derived From NADH

7.3 Multistructural links

Linking statements are thought to have multistructurai quaiity when they connect

component data with phrases such as "and." "or." and "and also". At this level of

understanding the leamer cannot generate the inferences that would synthesize data into a

single relational phrase. The exarnple multistructural response of Chapter 6 mizht appear in a

map as per Table 6.

Table 6

A Multistructural SOLO Res~onseExpressed in Link Format

1 Link 1 Data 1
1 1 It's bad because of the oil and it is also bad because of Saddam Hussein 1
I ( 1 ) It's bad because of the oil
(2) It's bad because of Saddam Hussein

Figure 5 contained no multistructural links.

7.4 Relational links

Table 7 contains the relational response of chapter 6. The link's relational status

refers to the fact that its linking statement encompasses two or more items of data. Thus far

this description of Method I has made use of the SOLO strate3 of rating quality against data

integration. The second factor that influences Iink quality is the number of data items. This

numerical factor can be incorporated into Method 1 by adding it to Iink ratinys as a suffix.

Thus relational links containiny two data items become relational-î links. Links with three

data items are relational-3 links and relational-4 Iinks contain four items of data. The number

of data items can also be applied to multistructural and extended abstract links. Link quaiity

is also influenced by whether the relevant data is provided in the instruction. This factor is

considered in section 7.5

Table 7

A Relational SOLO Reswnse Ex~ressedin Link Format

Link Data
x Iraq's troubles stem from Saddam's pan-Arabian ideolog which threatens
the World's access to cheap oil
( 1 ) It's bad because of the oil
(7) It's bad because Saddam's political desires seek to increase Arabian
control over oil supplies
i
(3) Increased Arabian control would increase Global oii prices

Table 8 contains three Figure 5 links that were judged to have relational quaiity.
Table 8.

Exam~leRelational Links in Fieure 5

1 Link 1 Data 1
Fermentation is a biochemical pathway that produces energy
( 1 ) Pathway oxidizes glucose
(2) Catabolism of oxidized compound releases a small amount of energy
Link 7 Stage 2 is the last phase of glycolysis
Relational-3 ( 1)Stase 2 is an oxidative phase
(2) NAD+ is reduced
(3)ATP is produced
Link 16 NADH is recycled to NAD+
Relational-3 ( 1 )Ce11 has lirnited supply of NAD+
(?)Fermentationrequires NAD+ to proceed
(3)As acetaldehyde is reduced NADH is oxidized

7.5 Extended abstract links

The use of matenai fiom outside the instruction is the third factor that influences

SOLO quality. Only hi* quality inferences can incorporate extemal data. Rumelhart and

Norman ( 1978) indicate why this is so. They recogize three leaming modes: accretion.

stnicturing and tuning. Accretion occurs when new knowledge is added to long t e n

memory. Stnicturhg involves the addition of knowledge to existing schema. Tuning requires

the retooling of schema structure. It represents expen performance. Links that simply repeat

an instruction may represent the accretion or the stnicturing mode. Links that upon

explanations provided in an instruction represent the tuning mode. Tuning links incorporate

the learner's their existing knowledge into an instruction. The result is a finicino aatement that

improves upon the instruction's description of the co~ectionbetween two concepts.

Extended abstract links represent the tuning mode. They contain one or more items of
 78

data that were not presented in the instruction. Inferences at this level of cognition should be

exarnples of creative thinking. There are no extended abstract links in Fiyure 5 . ï h e extended

abstract response presented in chapter 6 would appear in link format as per Table 9.

Table 9

An Extended Abstnict SOLO Resoonse Exoressd in Link F o m

- - -

Link Data
Iraq's troubles stem fi-om1923 when the Middle East was divided to suit
the political aspirations of Colonial powers. Recognizing that extemally
imposed land division attenuates Arabian power, Saddam Hussein
sought to unite the Middle East. A united Middle East could
undoubtediy change the balance of global politics by controlling the
movement and pnce of oil. It is the global threat imposed by Saddam's
ideology that makes Iraq's politicai situation so bad.
( 1 ) It's bad because of the oil
(7) It's bad because Saddam's political desires seek to increase Arabian
control over oil supplies
(3) increased Arabian control would increase Global oil prices
(4) The above problems stem from the division of the Middle East in
I 923

Method I takes the different leamhg modes into account. In the empirical component

of this thesis, students read a text and then map it. If a student knows absolutely nothing

about the contents of the text then leaming will occur in the structuring mode. if the text

adds to students' existing knowledge then the accretion mode of leaming will be at work.

When midents map leaming that occurred in the stmcturing mode they will likely regurgitate

the text. Students for whom the text provides accretion type lemirtg may also choose to

simply reproduce the text. If two sudents produce maps that reproduce the text then their
 79

unistructural links will be of the same quality. Similarly their relational links will involve

comparable inferences t hat reproduce the text.

However, students learning in the accretion mode need not simply reproduce the text.

They cm provide structures that indicate how they associated the text with their existing

knowledge. Their maps may contain a representation of t he text as well as links that illustrate

how the text was connected to their existing knowledge. Links that reproduce the text

contain inferences that are of the same quality as the structuring mode. Links that associate

the text with existing knowledge contain inferences of a higher quality. Their superiority

stems from the fact that the text is not used to guide their creation. Inferences created in the

tuniny mode are of higher qudity still. Learners operatin%in this mode reason as to how their

current understandinç of a link fits in with the description provided in the tea. Then they

integate the two underaandinss to generate a superior representation. The tuning mode is

associated with extended abstract thinking. . The structuring mode is represented by

the unistmctural. multistructural and relational SOLO type links descnbed above. A more

cornpiete scde is needed to account for the tuning and accretion modes. Such a xale is

presented beiow .

Prestnictursl links: the linking statement was fonned by associating concepts

through an incorrect inference or an inference that while correct involves an irrelevant

item of data.

Unistructural links: the linking statement was formed by associating concepts

 80

through a correct inference that involves a single item of data. The linking statement

repeats the contents of the text.

Extended unistructural: the linking statement was formed by associating concepts

through a correct inference that involves a single item of data. The matenai is relevant

to the text but was not presented in it.

Ni ultistructural: the Iinking statement was formed by conjoining items of data to

associate the two concepts of the link. That is the Iinking statement is a conjoined list

of the data items. No inferences were formed. the material was presented in the text

Extended multistnictural: the linking statement was fomed by conjoining items of

data to associate the two concepts of the iink. No inferences were formed. The

materid is relevant to the text but was not presented in it.

Relational: the linking statement was formed by associating concepts through a

correct inference or inferences that involve relevant items of data. The linking

statement repeats the contents of the text.

Extended relational: the linking statement was forrned by associating concepts

through correct inferences that involve relevant items of data. The matenal is relevant

to the text but was not presented in it.

 81

Extended abstmct: the linking statement was formed by associating concepts

throuyh correct inferences that involve relevant items of data. The statement improves

upon a relationship presented in the text. It incorporates one or more additional items

of data.

7.6 Assessing the stnictuml knowledge of entire maps and individual concepts

Method 1 can be applied to entire maps or individual concepts. In either situation.

structural knowledge is quantified by counting Method 1 Iink types. Table 10 quantifies the

structural knowledge of Figure 5 .

Table IO.

The Structural Knowleds Links o f Fieure 3

( Link Type 1 Total 1

1 Relational-2 1 4 Links: 1,2,4,9 1
1 Relational3 1 4 Links: 6,7,12,16 1
If another map of fermentation was found to contain:

A. 8 unistnrcturai

B. 3 relationai-2

C. 5 relational-3

it would be considered to exhibit slightly supenor structurai knowledge.

Cornparisons of individuai concepts involve sirnilar comparisons. A concept having

the following links:

A. 3 unistmctural

B. 3 relational -2

would be considered superior to one with links:

A. 3 unistructural

B. 3 relational-3

7.7 Weaknesses o f Method 1

Problems anse when iMethod I is used to compare individual concepts or entire maps

that do not have the same number of links. Method 1 considers the diference between

unistructurd and rnultistnictural quality to be the same as the difference between relational

and extended abstract quality. Method 1 also considers that links of the same type are of the

same quality. In actuality, links of the same type do differ in quality accordiny to the concepts

they connect and the data they contain. Further. even when two links are identical their

bearing on understanding wiil likely Vary according to the abilities of students who created

them.

Two hypothetical maps (not shown) help to illustrate that specific cornparisons are

not possible when maps contain different numbers of links. Suppose that map A contains 15

[inks

a. 10 unistructurai

b. 4 relationai-2

c. 1 relationai-3
and that map B contains 10 links

a. 4 unistructuraI

b. 4 relational-3

c. 3 relational-3

Method 1 cannot determine which map is better as it is incapable of companny the coynitive

worth of different link types. In the case of maps A and B a scale is needed that is capable of

specifjm~the value of unistmctural links in terms of relational links.

When maps (or concepts) do not contain the same number of links. Method 1 permits

only a yeneral commentary about link quality. It is possible to say "this map (or concept)

contains Say five links of unistructural quality. and four of relational-2 quality " while another

contains "four of unistmctural and two of extended abstract quality." These commentaries

may facilitate rough rankings but unless the differences between maps (or concepts) are

obvious. Method 1 cannot compare two maps (or concepts) directly.

Method 1 cornparisons may also suffer when assessinç generalized concepts. Links

h m these concepts summarize the lower level concepts they encompass. In makins a

surnmary the student has to decide which data to include and exclude. For example. link 1 of

Figure 5 summarizes the process of fermentation with N O data items. It is conceivable that

another author would use more data to generate a summary. The additional data may be

unnecessary to the link or it may be repeated at lower points in the hierarchy. The links

associated with generalized concepts rnay receive infiated Method 1 scores unless the

examiner takes time to i d e n e and exclude redundant data.

 84

Despite the weaknesses identified above. Method I is well suited to Our purposes.

The goal is not to compare the quality of structural knowledg across maps or concepts.

Rather we are interested in cornparhg the quality of the structural knowledge of individual

links to the link-procedural knowledge that created it.

7.8 Method 2: Progressive differentiation and individual concepts

Intesrative reconciliation is a reflective process t hat improves the quality of existing

relationships and identifies new ones. It also identifies erroneous links. By assessing the

quality of linking statements. Method 1 reflects the results of inteyrative reconciliation.

Proyressive differentiation is based on the reasoning that concepts are never hlly understood

(Novak and Gowin. 1984). It assumes that links can always be added to concepts as leaming

is never complete. Method 2 accounts for progressive differentiation by quantifjing the links

associated with individual concepts. It assumes that more links signiQ more progressive

differentiation.

In Method 2- concepts are considered to have Nurnber-Link structures. Each of these

structures comprises a particular concept and its links to superordinate and subordinate

concepts. Number-Link structures are readily compared in different maps. In Figure 5 the

concept Ethat~oihas a I-Link structure as it has oniy one link to its superordinate concept

S i q y 3 (link 15). Multistructural Number-Link concepts have two or more links to

superordinates and subordinates. Glyco!vsis is linked to two subordinates (links 6 and 7) and

one superordinate (Li& 4). It is a 3-Link structure. The concept Stage 2 is a CLink structure
(links 7, 10, 1 1 and 12) .

Unlike Method 1. Method 2 does not require lenghy consideration of the data.

Number-Link structures are quickly identified in maps by visual inspection. A complete list of

the Number-Link ratinys for Figure 3 is provided in Table 1 1.

When the links of Table 1 I are counted a total of 32 is attained. There are only 16

links in Figure 5. The discrepancy arises because Table 1 1 registers each link twice. once for

the concept at the head of a link and once for the concept at the tail. Method 7 can be used to

compare the

Table 11.

Number-Link IMethod 2) Ratines o f Fipure 5

-

1 Concept 1 Number-Link rating

- - - . .. - . - -

1 Fermentation
1 ATP
Glycolysis 3-Link
Stage 1 3-Link
Stage 2 4-Link

1 ATP (stage I region) I

1 NADH
1 ATP (stage 2 region) 1 1-Link I
1 Ethanol 1 1-Link 1

progressive differentiation of entire maps provided one remembers that each link is registered

twice. Met hod 2 is better suited to the cornparison of individual concepts. For example.

suppose the representation of a concept changed from 2-Link to 5-Link as a result of

instruction. It can be said that the instruction stimulated progressive differentiation to the

extent that three new links were generated. If the concept changed from 1 -Link to 3-Link in

another map the progressive differentiation would be considered inferior as only two new

links were generated. Unlike the situation in Table 1 1. cornparisons of individual concepts

enumerate links just once.

7.9 Limitations of methods 1 and 2

Method 2 is included in this thesis to illustrate some of the uses to which concept

maps can be put. Progressive differentiation and integrative reconciliation occur

simultaneously duhg the learning process. Methods 1 and 2 can be used to reflect learning

as they measure progressive differentiation and integrative reconciliation. If two maps

associate a concept with the same sub and superordinates then the understanding they express

can be compared through Method 1 (the need for the wne super and subordinates is

expiained in the next paragraph). The concept with the highea Method I rating is said to

exhibit supenor integrative reconciliation (struc~ralknowledge). iMethod 2 adds to the

 87

cornpanson as it quantifies the progressive differentiation (number of links) associated with

each concept. Table 12 illustrates the information that is needed to compare the

understanding associated with individual concepts.

In Figure 5 , krmertrutiort is linked to î i j t r r ~and Yec~stby a relational-2 link (link 1)

and a unistructural link (link 3). Another author creatinç a rnap of fermentation fiom Brock

to L i t e r ~and Yrast with a relational-:! and a

and Madigan's text may link fikrmtnmio~~

relational-3 link. Cornparison in this case is a simple matter. the second author exhibits

superior understanding. If a third author links Frrmrt~rario~~~

to E w r m and Lactic Acid

Bacterirr by a relational-2 link and a unistructural link the cornparison is more cornplicated.

The question of how much more difficult it was to link Lactic dctd Hïcc~rriuas opposed to

Yrast has to be considered. Method 1 cannot answer this question as it fails to consider the

impact of different concepts on link quality.

Table 12

Method 1 and Number-Link Ratiag~for Conceots in Fieure 5

Concept Number- Super Subordinate SOLO Rating of

Link ordinate concepts Link
rating Link
Fermentation 2-Link- N/A Energy Relational-2
Yeast Unistructural

1
f
Glycolysis 3-Link Yeast Stage 1 Relational-3
Stage 2 Relational-3
 Method 1 and Method 2 are both limited by the fact that their results can't be

integrated into a sinyle expression that ponrays the learning exhibited in a map. Single

expressions of learning make for easy comparisons as the map with the hi&est score can be

assumed to portray supenor understanding. Intqrating rnethods is considered important as

progressive differentiation and integrative reconciliation occur in an integrated fashion dunng

the leaming process. Future studies wishins to compare concept rnap knowledge would do

well to provide an integrated rating aiongside Method I and Method 2 scores.

7.10 Method 3: Link procedural knowledge

Method 1 assesses the structural component of knowledye. It considers the

understanding generated by integrating data to form an inference. Method 3 assesses the

procedural component of knowledge. it considers the inferences themselves The difference

between rnethods 1 and 3 is a subtle one. Both assess the sarne thinç. knowledge. but they do

so fiom different perspectives. Method 1 assesses results of the actions of link-procedural

knowledge. Method 3 assesses link-procedural knowledye itself

Biggs and Collis ( 1982) suggest that students' responses to questions have five levels

of quality. This infers that links are created by five levels of inference quality. The following

descriptions of the five inference levels were developed by integrating Biggs and Collis'

SOLO levels with the learniny modes (accretion, amauring, and mning) of Rumelhart and

Norman.

Level 1. At this level inferences involve incorrect reasoniny. Two scenarios are possible. In
 89

the first instance the inference connects two data items that are not truly associated with the

concepts of the link. Althouyh the inference is correct the data involved render it irrelevant to

the link. In the second situation the data is appropriate but the inference is incorrect.

Level7. Simple listings of facts that fail to connect data throu@ inferences.

Level 3. Inferences based on fact (inferences that replicate descriptions found in the

instruction)

Level4. Inferences based on facts that were not ovenly linked to the concepts in the

instruction (correct reasonins that builds upon information presented in the instruction).

LeveI 5 . inferences based on facts that were taken from outside the instruction (correct

reasoning that could not have been inferred From information presented in the instruction).

Levels 1.2 and 3 represent the structuring mode of learning. They involve text replication

and encornpass the following links of Method 1 : prestructural. unistructurai. multistructural

and relational. The accretion mode of leaMng is represented by level4. Level4 inferences

build upon information presented in the text. Extended unistructural and extended relational

links contain level4 inferences. Level 5 accommodates creative thinking, it is reserved for

extended abstract links and the tuning mode of leaniing. The inference levels represent five

levels of link-procedural knowledge. Method 1 and the works of Biggs and Collis and
 90

Rumelhart and Norman susgea that the five inference Ievels of link-procedural knowledge

will be measurable on the following eight stage scale.

Stage 1. Inappropriate reasoning. An inference is made that fails to consider information

presented in the te*. The result is a link that connects concepts through inappropriate items

of data.

Stage 7. Concepts are linked through a listing of one or more items of data. No inferences

are present to indicate how or why the concepts are connected.

Stase 2aa.One item of data is lined to link the concepts. No inferences are made.

Stase Ib. Two items of data are listed to link the concepts. No inferences are made.

Stase Sc. Three . . .

Stage 3. One item of data links the concepts through an inference based on facts (the

inference replicates descriptions presented in the tes).

Stage 4. Two or more items of data are linked through inferences based on facts. The

inferences replicate descriptions of information presented in the text.

Stase 4a. Two items of data are linked through an inference based on facts.

Stage 4b. Three items of data are linked through two inferences based on facts.

Stage 4c. Four ...

Stage 5. One item of data links the concepts though an inference based on facts. The

inference was not overtly discussed in the text.

Stage 6. Two or more items of data are linked through inferences based on facts. The

inferences were not ovenly discussed in the text.

Stage 6a. Two items of data are linked through an inference that was not presented in

the text. The inference is not based on inappropriate reasoning. It buiids on

information presented in the instruction. Some explanation of the inference may be

presented.

Stage 6b.Three items of data are linked through an inferences. The inference is not

based on inappropriate reasoning. It builds on information presented in the

instruction. Some explanation of the inference may be presented.

Stage 6c. Four . . .

Stage 7. An item of data connects the concepts through an iinfrence based on fact. The

inference does not build on information presented in the text. It is an example of creative

thinking.

Stage 8. Two or more items of data are comected through inferences based on facts. Rather

than building on information presented in the text the inferencets) represent creative thinking.

Stage 8a. Two items of data are linked through an inference that was not

presented in the text. Rather than building on information presented in the text
 the inference is an example of creative thinking

Stage 8b. Three . ..

Stages 2. 3 and 4 involve text replication. They represent the stmcturing mode of

learning. The building process that typifies the accretion mode is accomrnodated by stases 5

and 6 Stage 7 and 8 are desiyned to represent link structures of the tuning mode. The

incorrect thinkiny of stage 1 links may occur in al1 three leaming modes. The association

between the inference levels and methods I and 3 indicates the author's expectations

regarding the structural-link-procedurd knowledge relationship. It is anticipated that

structural knowledp of unistructural. rnultistructural and relational quality will be affiliated

with link-procedural knowledge of stages 2 . 3 and 4. Enended unistructural and extended

relationai links are thought to be associated with stases 5 and 6. Finally. it is expected that

link-procedural knowledge of stages 7 and 8 will create extended abstract links.

Table 13 contains Method 1 and Method 3 ratings for the links of Figure 5 . These

ratings conform to the author's expectations. Hi&er level structural knowledge is associated

with higher levei link-procedural knowledge. Unistructural links were created by stage 3

inferences. Stage 4a inferences created the relational-? links and relational-3 links are

associated with inferences of the stage 4b level. Figure 5 replicates the description provided

in Brock and Madigan (1988). Consequently the map contains no extended unistructural.

extended multistructurai or extended relational links. These link types would be found in

maps that contain Iinks that are representative of the accretion mode of leaming. Links that

represent the accretion mode connect a concept from the text to a related concept that the
 93

could be linked to Krspiratiom.

text fails to mention. For example. the concept Fèrntenta~io~~

Respiration is related to fermentation as it is an energy producing rnetabolic process. Brock

and Madigan do not mention respiration in their description of fermentation.

Table 13

Method 1 and Method 3 Ratines o f the Links of Fieure 5

1 Link Number 1 Method I Score 1 Method 3 Score 1

3 Unistructural 3
4 Relational-? 4a

- - - - - -- -- -- -- -- -

6 Relationai-3 4b

7 Relational-3 4b

8 Unistructurai 3

10 Unistructural 3
11 Unistructurai 3
12 Relational-3 4b
I

13 Unistructural 3
14 Lrnistructurai 3
7.1 1 The interview

Link-procedural knowledye is functionally equivalent to the cognitive processes of

link creation. Therefore in assessing link-procedural knowledge Method 3 also assesses the

quality of cognitive processes. An i n t e ~ e wscript was developed to mess the metacognitive

activity of concept mapping. It is used alongide Method 3 to investigate the cognitive and

metacognitive processes of concept mapping. The format of the i n t e ~ e wwas modeiled after

Myers and Paris ( 1 978).

Method 3 scores were used to select six students from the gifted grade 9 Biology

class. Three of these students exhibited hi@ Method 3 scores or hi@-level cognitive activity

Two exhibited medium-level cognition and low-levei coynitive activity was displayed by the

remaining nudent. The students were interviewed individually over three consecutive lunch

hours in the week followins the mapping exercise. During the interview the interviewer (the

author) and student sat side-by-side at a table with a tape recorder between them. Students

were told that there are no right or wrong answers and that the object of the exercise was to

find out what they think. The i n t e ~ e wquestions were presented in a conversational manner.

If a audent failed to understand a question it was repeated. If repetition of the question failed

to elicit a response it was rephrased. The entire interview was tape recorded.

7.1 1.1 The interview questions

The i n t e ~ e wquestions and the example answers are represented below. They are

adapted from Park and Jacobs ( 1984). During the i n t e ~ e w

the questions were read out in

numericd order. Students were prompted with example answen if they codd not respond to
a question or if it was obvious that they had rnisunderstood it. Questions 7, 12 and 13 were

presented differently. Here, students were told to select the example answer that they felt was

most appropriate for the question. They were also asked to explain why they selected that

particular answer. In each of the questions presented below, the first answer typifies

responses that would receive a score of zero. Answers that score zero represent inadequate

metacognitive activity. The second answer is typical of responses that would receive a score

of one. .4 score of one signifies adequate metacognition. Answen that would receive a score

of two are represented by the third answer. Superior metacognitive activity is represented by

answers that would receive a score of two. There are 13 i n t e ~ e wquestions and the

maximum score for the interview is 26.

1. What is the hardest part about concept mapping for you?

Decidins on the overall stnicture of the map
Creatins the linking statement of the links
lntegrating data to form a linking statement

2. What would help you become a better concept mapper?

Better texts
More practice
An increased abiiity to see the big picture

3.How do the concepts at the top of a concept map differ corn those at the bottom?
There is no difference between them
Concepts at the top represent the beginning of the story
Concepts at the top are general, they encompass specific concepts that reside at the
bottom of maps

4. Which concepts in a map are the most important?

Al1 concepts have equal importance
The most interesting concepts are the most important
The ones at the top as they contain the most information
5. Suppose you dont have the time to create a complete map. Which concepts do you include
to convey the most information?
The concepts in the middle of the rnap
The most interesting concepts
The concepts at the top as they include the most information

6. Before you stan to create a rnap what plans do you make?

I make no plans. 1 jun start placing concepts in the rnap
I make sure I will not be disturbed for a while
Ithink about where concepts should be placed in the rnap

7. Which of the following do you think would be easier to rnap and why?
A passage with a lot of pictures
A passage that is easy to understand
h passas you have read before

8. Why do you reread the text before you linli some concepts together?
Because you forgot what you had read
Because it is sood practice to do so
Because you don't understand the concept's relationship

9. What do vou do if you come to a link that you don't understand?

~gnoieit and go on to the next link
Ask someone else to expiain it
Use the context in the text to figure it out

10. Which parts of the text do you omit tiom the concept map?
You include everythiny
The parts you don't understand
The unimponam parts that don't add to the concept map

1 1. While you read the text what do you do to help you remember it?
1 don't do anything to help me remember
Concentrate hard and try to remember
Ask purself questions about the important parts

12. If you were creating a rnap in preparation for a test, whîch of the following would help
you the most and why?
Saying the concepts over and over
Read the rnap as many times as possible
Discuss the map with someone else to make sure you understand it
13. If you are creating a map ffom a technical passage. which would help you the most and
why?
Sounding out the words you don't know
Skipping the parts you dont understand
Wnting the passage out in your own words before starting the map

The students' tape recorded responses were transcribed. Lengthy responses were

paraphrased and summarized. The transcribed responses were rated against the example

answers. Pans and Jacobs ( 1984) claimed t hat this marking system reflects metacogitive

awareness. Therefore the higher the intewiew score the higher metacoyitive activity of' the

mapper.

7. I f A summary of Methods 1,2 and 3

This section summaizes the sconng systems of the three rnethods. In Metbod 1 link

rankings are based upon three factors: the amount of data in a link. the extent to which that

data are synthesized. and the novelty of the data. Links that simply repeat the text are

labelled prestructural. unistructurai. rnultistnictural or relational. Prestmctural links connect

concepts with irrelevant data. Lhistructural, multistmcturai and relational links are identified

by counting relevant data items. Unistructural links contain jus one item of data.

Multistructural and relational links contain two items of data or more. Unistructural and

relationai links synthesize data into a single phrase (the linking statement). Multistructural

links fail to exhiiit data imegration. UniaructuraL multistnictural and relational links

replicate the text. They may represent leaming that occurred in the aructuring or accretion

modes.
 98

Links that build upon information provided in the text represent a higher b e l of

structural knowledge. They represent the accretion mode of Ieaming. These Iinks are Iabelled

extended unistructural. extended multistructural and extended relational links. Extended

abstract links contain knowledge from outside the instruction. They represent the tuning

mode and they irnprove upon the tefi's description of links. Extended abstract links integrate

data and may provide multiple conclusions, they represent the hi&est level of structural

knowledge.

Met hod 2 enurnerates the links associated with individual concepts. Once enumerated

the Iinks are used to provide each concept with a Number-Link score. Method 2 is well suited

to the identification of progressive differentiation associated with individual concepts. It is

not well suited to the assessrnent of entire maps. Funhermore it cannot distinpish between

concepts that have an equal number of links. Methods 1 and 7 can be used in conjunction to

assess the progressive differentiation and integrative reconciliation associated with individual

concepts. If concepts have identical Nurnber-Link and SOLO ratinys. a finer scde is required

for differentiation. Methods 1 and 1 would benefit greatly if future studies locate a means by

which the cognitive wonh of one link type cm be expressed in terms of another. Such a

system would make it possible to compare maps whose links differ in number and quality.

Method 3 recognizes the same factors as Method 1. Links that replicate the text

contain either structunng or accretion inferences. They are accorded the lowest score. A

higher level of inference occurs in accretion links that build upon the text. Links that

represent the tuning mode contain the highest level of inference. Regardless of the leamhg

mode. the more data involved in a link the higher the quality of the link-procedural
knowledge.

Table 13 indicates that the quality of structural knowledge is related to the quality of

link-procedural knowledge. This finding is encouraging as it suygests that empincal data will

generate a similar finding. Comparable structural and link-procedural knowledge supports the

thinking that the structural and procedural modalities are isomorphic phenomena.

Link-procedural know ledge is functionalIy equivalent to the simultaneous processing

of comprehension. I t is a copitive process. Link-procedural knowledge and interview scores

permit a comparison of the cognitive and metacognitive processes of mappers. Findiny that

Method 3 and intewiew scores are comparable is interpreted as beiny supportive of the

argument that the cognitive and metacognitive processes of concept mappins are related

processes.
 Chapter 8

A Description of the Ernpiricd Component of the Research

8.0 The research

The main aryument developed in this thesis has been that the quality of structural

knowledge is isomorphic to the Iink-procedural knowledge t hat generates it . This supposition

is exarnined here by applyiny methods 1 and 3 to links that exhibit high and low quality

knowledge. Given the need for hi@ quality links, a gifted grade nine Biology class was

chosen for this snidy. It was felt that their high academic abilities would enable the necessary

production of high quality links. Students of lesser ability may be incapable of producing high

level links. A second reason for choosing a bioiogy class is that the researcher has some

experience in this domain. This experience is needed to gauge the accuracy of students' links.

The research was nin over five school days. In the first session the class' 27 students

were introduced to concept mapping through a twenty minute lecture (section 8.7) that was

delivered by the researcher. On completion of the lecture the researcher ran the students en

masse through a concept map that depiaed the making of a cup of tea. It was thouçht that

this farniliar activity would introduce mapping with a minimum of cognitive strain. For the

remainder of the class (and for homework) the students were asked to pnerate a concept

map that depiaed their understanding of the causes and consequences of oil spills. They

were provided with three key concepts Tattkws, Oil Spilis and C'lran Cl'. They were aiso

told to think of 7at1ker.sand CIem [Jp as anchor concepts for two regions subsumed beneath

the concept Oil Spills.

Asking students to map their understanding of oil spills had three purposes. First it
 101

provided students with an opportunity to practice concept mapping. Second it allowed the

researcher to identiq and correct problems with student's mapping techniques. Incorrect iink

structure proved to be a cornmon problem. It was remedied through one-on-one instruction.

The third purpose of the mapping exercise was to prime students' thinking so that they could

easily incorporate the target text "Water Pollution and Oil Slicks" into their existing

knowledge. The target text was taken from the "Environmental Database for use in SchooIs"

which is on-line at: http://www.soton.ac.uk/engenvir/environmenwater/. This web site is

affiliated with the University of Southampton in the United Kingdom. The target text is

presented in Appendix A.

In the second class students read the tarset text and began to map it. The class ran on

a Fnday and students were yiven the weekend to complete the maps. M e r markiny the maps

the researcher selected six students for metacognitive interviews. Two of the six subjects had

Iow Method 3 (cognitive process) scores, two had medium scores and two had hi@ hlethod

3 scores. Students with low. medium and high Method 3 scores were selected in response to

the reasoning that Method 3 and interview scores would increase in parallet. In other words a

range of cognitive abilities were selected for i n t e ~ e wso that the cornparability of cognitive

and metacognitive processes could be demonstrated.

8.1 Methodology and data collection

Before the midy began students were asked to sign a letter of consent. The letter is

presented in Appendix B. It explained the nature and purpose of the snidy and it assured

participants that al1 Somation would be kept contidential. Students who wished to opt out
 1Oz

of the study were given the option of reading the target text and continuing with the regular

curriculum. Altematively students could simply continue with the regular cumculum. Letters

of consent were also presented to the teacher who nomally teaches the class and the school

principal. Copies of both letters are presented in Appendix B. .

Running the research successfilly requires participants who are farniliar with concept

rnapping and the research pals. The followiny section presents the written instructions

provided to the teacher and nudents.

8.2 Written instructions presented to the teacher

Concept maps are created by identifying relationships that link one concept to

another. In most maps the concepts are arranged hierarchicdly. General concepts tend to

appear at the top of the hierarchy and specific concepts tend to appear at the bonom. The

completed map reveais students' understanding of a topic. White and Gunstone ( 1992)

explain the benefits and educational uses of concept mapping. Their explmation is

surnrnarized below.

Oflen as students l e m a topic they become wrapped up in its minutiae. Their

cognitive resources are so fûlly taken up with knowledge acquisition that they never engage

in the reflective processes that cari Link a topic to other subjects. Concept mapping presents

midents with the opponunity to step back, see "the big picture" and create connections

across topics or themes. Creating links forces students to confiont their cornprehensions.

When a comprehension is seen to be hadequate, remedial action has to be taken. Concept

maps are also a tool of rnernory that requires the retrieval and reprocessing of knowledge.
Reprocessing strengthens the links between concepts. It also increases the potential for future

recall.

Mapping is also of benefit to teachers. It increases their awareness of their

comprehension processes. Arming teachers with a better understanding of the operations

behind their comprehension enables them to better expiain the mechanics of comprehension

to their students. Each time a teacher maps a topic a different aspect (or perspective) of it

may be emphasized. Providing students with multiple perspectives leads to more cornpiete

understanding than the presentation of a single perspective (Spiro, Jacobson. Feltovich and

Coulson. 199 1a).

Eigurei 1. An example concept map.

Figure 1 1 is an example concept map. It ponrays the making of a cup of tea.

Concepts such as Wuter, Tra Bug and Tra C i p are presented in rectangles. The rectangles

cm be thought of as iconic representations of the concept narned within. Links express the

mapper's understandimg of concepts. They are the hart of concept maps. Each link reveais

somethlng of the understanding surnmarized by the rectanqular icon. The nurnber and quality
 1O4

of links reflect the mapper's understanding. Table 14 contains descriptions of the links of

Figure 1 1.

This research asks students to create concept maps. Three methods have been devised

to assess these maps. The methods are desiged to estimate the quality of student learning.

They also reveal something of the cognitive processes that students engage in as they acquire

comprehension. Mer the concept maps have been assessed. six students will be inte~ewed

to

Table 14.

escriotions o f the Links of F b r e 11

Link Number Linking Statement and Data

1 Add boiling water to tea cup
1 .Add boiling water to tea cup
3
I Add tea bag to tea cup
1. Add tea bas to tea cup

3 Allow the tea to steep to desired concentration

1. Boiling water dissolves components of tea
2. Convection currents in hot water distribute dissolved components of tea
3. Stirring speeds up the distribution of dissolved components
4. Repeated visual checks are needed to prevent over steeping
4 When desired concentration is reached remove the tea bag
1 When desired concentration is reached remove the tea bag
5 Add milk to the tea
1. Add rnilk to the tea
6 Add sugar to the tea
1. Add sugar to the tea

elicit the metacognitive processes they used to guide their thinking while mapping. It is
 1 O5

suspected that high level metacognitive activity will be displayed by students who create high

quality links.

If the methods are to work they require that students provide complete descriptions

of links. Complete Iink descriptions reveal something of the cognitive processes that students

engage in as they leam. A link is considered complete if it contains a general linking

statement and a listing of the items of information that were used to yenerate the statement.

Linking statements stipulate the relationship between link concepts. Each piece of

information concems one aspect of the link relationship. For example, link 3 of Figure I 1 has

the linking statement "Allow the tea to steep to desired concentration." Four pieces of

information are provided beneath the linking statement.

1. Boiliny water dissolves cornponents of tea.

2. Convection currents in hot water distribute dissolved components of tea.

3. Stirring speeds up the distribution of dissolved components.

4. Repeatedly check concentration to prevent over steeping.

Each item of information explains one aspect of the steeping process. It is these pieces of

information that reflect the mapper's cognitive processes. Some encouragement may be

needed to engender students with the confidence they need to e.xpress their cognitive

processes in links.They should be informed of the fact that if' items of information are

relevant they cannot be incorrect.

Linking staternents vary in qudity. Any linking statement that comprises a single piece
of information is cailed a unistructural link. Links 1, 2,4. 5, and 6 of Figure 1 1 are

unistructurai (see Table 14). Their single piece of information is identicai to the linking

statement. Such links are pefiectly leyitimate. In other unistructural links the single piece of

information may differ from the linking statement. These links are also legitimate.

Multistnictural links contain two or more pieces of information. Their linking

statements conjoin pieces of information with phrases such as "and," "or" and "also."

Relational links differ from multistnictural links in that their linking statements inteyrate

pieces of information into a single phrase. Link 3 of Figure 1 1 is a relational link. Its linkiny

statement "allow the tea to neep to desired concentration" represents an integration of four

items of data. If link 3 were a multistnictural link its linking statement might be phrased

"boiling water dissolves cornponents of tea and convection currents distnbute them. The

dissolving process is enhanced by stimng. In additioa repeated checking may be needed to

prevent oversteeping."

Linking statements increase in quality from unistructural to multistnictural to

relational. Extended abstract links represent an even higher quality. These links contain at

least one piece of information that was not presented in the instruction. They represent an

attempt to add extant knowledge to descriptions provided in the instruction. There are no

extended abstract links in Figure 1 I but several could be created. For example. Iink 3 could

incorporate the production of any beverage whose taste depends on concentration.

Sometimes links will contain invaiid linking statements andor pieces of information. These

links are cded prestructurai links. They are not included in the assessrnent procedure.
8.3 lntroducing students to concept maps

Research with concept maps requires students who are competent concept mappers.

White and Gunstone ( 1992) provide a method that can produce competent mappers in shon

order. The following points represent an adaptation of t heir method. The researcher used this

adaptation to introduce concept maps and the research. The adaptation was presented to

students as a shon ( 15 to 20 minute) lecture.

1. Prior to class select a simple topic that is familiar to the students. Choose a few key

concepts and senerate a concept map. Insert linking statements and appropriate data into the

map.

2. The map must contain examples of unistructural and relational links.

3. At the beçi~ingof class present the typical hierarchical structure of a concept map on an

overhead projector or blackboard. Point out the significance of hierarchical levels. regions,

and seneralized and specific components. Inform students that creating maps can improve

their recall and help them or~anizematerial for essay writing.

4. Display the key concepts from point 1 on a blackboard or overhead projector. Arrange the

concepts according to the concept map created before class. DO NOT connect the concepts

at this stage.

5 - Reveal the title of the map.

6. Ask students to explain how the concepts are related to each other. Place these

explanations on the blackboard and comect the appropriate concepts with an arrowed link.

These explanations constitute Linking statements.

 1 O8

7. After drawing each link. ask the students for the rationale behind the positioning of the

arrow's head and tail. If no answer is forthcornhg provide one. Invariably the explmation will

be that the direction is expressed explicitly or implicitly in the linking statement. An explicit

directional cue might be "concept A performs action X on concept B." An implicit cue such

as "action X impacts on concept B " indicates t hat the arrow's head should be adjacent to

concept B. The direction of arrows is not a difficult concept to yrasp.

8. Ask the students to identify how they fonned the linking statements. One way to do this is

to ask what "information" they used to make the linking statement. If this open-ended

question fails to elicit the desired response, provide clues usin3 the concept map created in

point 1. Table 1.1 contains the linking statements and data of Figure 1 1 . It illustrates correct

linking procedure. each item of information is numbered and liaed beneath the linking

statement. Be sure to model correct link structure when creating links from information that

audents provide.

9. Once data items are identified and the map is cornpiete the audents should be made aware

of the following:

a) items of information that match the linking statement are acceptable when there is

ody one item of information in a link

b) no item of information from multistructural or relational links should match the

linking statement exactly

10. Reiterate the point of the exercise using the following as a guide.

a) Tell students what they cm expect to set from the exercise. Perhaps the most

tangible benefit that students wil1 latch ont0 is that concept mapping has the potentid to
 1O9

improve their comprehension and retention. tnform them that concept mapping is a good

habit to develop.

b) Note that the aim of t he research is to leam about the cognitive processes that

students enyage in as t hey lem. Make students aware t hat the success of the research

depends on their generating a linking statement and data for each link.

I I . Ask nudents to generate a practice concept map. inform them that the subject will be the

causes and consequences of oil spills. Provide the followiny key concepts Od .Sp/Ls, 7a11ker.s

and ('iean tlp and tell them to use T~mkersand <'lean Up to anchor regions beneath the

concept 0 1 1Spdls. Hand out the written instructions for concept mapping. These instructions

are provided in appendix C.

13. Provide each student with a sheet of bristol board and a set of post-it notes. Students will

arrange concepts on the bristol board as they see fit. When students are satisfied with their

concept arransements they will stick the post-it notes onto the bristol board and create

linkine statements and data for the links.

14. Check to ensure that linking staternents and data are entered correctly. Provide corrective

instruction as necessary.

15. Hand out the target text and another sheet of bristol board.

16. Ask students to read the target text and map it. They should use the three key concepts

Oïl Spills, Tankers and C'lm Clp. In addition they should incorporate information fiom the

practice map whenever they think it appropriate.

17. Collect and examine the concept maps.

18. Select six subjects for the metacognitive i n t e ~ e w

 Point 1 1 notes that wrinen instructions were provided to the students. These

instructions were designed to reinforce the researcher lecture. They describe the vaiious steps

that have to be followed in order that a concept map can be created. The instructions are

provided in appendix C. Appendix C also contains written instructions that explain how to

apply methods 1.2 and 3 to the students concept maps. These marking instructions were

provided to the teacher.

8.4 A map o f the target text

Some assurance is needed that the taqet text is suited to concept rnapping and the

goals of the thesis. To provide this assurance a concept map of the target text is provided in

Figure 12. Table 1 5 describes the links of Figure 12. Table 16 contains method 1, and 3

Fimire 12, A concept map of the target text

 111

ratings for Figure 12's links. it also contains Method 2 ratings for the key concepts Tmrkrrs.

Oil Spills and ('lean ilp. Methods 1.2 and 3 are applied to Figure 12 to meet the first and

second p a l s of the thesis. The third goal was not assessed as it requires multiple i n t e ~ e w

scores.

Table 15.

The Links of Fieure 12

1 Link Number
-- -- --

Description
The Gulf was is an example of an oil spill
1. The Gulf was is an exarnple of an oil spi11
- - -

The E.xxon Valdez is an example of an oil spill

1. The Exxon Valdez is an example of an oil spi11
Braer is an exarnple of an oil spi11
1. Braer is an example of an oil spi11
Tankers are a source of oil spill
1. Tankers are a source of oil spill
Tanker leakage rnay cause a slick
1. Tanker leakage may cause a slick
- - -

Tanker accidents may cause a slick

1. Tanker accidents rnay cause a slick
Spills have to be cleaned up
1. Spills have to be cleaned up

Spills have to be cleaned up

1. Spills have to be cleaned up

SpilIs have to be cleaned up

1. Spills have to be cleaned up

Legislation lead to reduced leakage

I . The U.S. Oil Pollution Act 1990 called for double huils
2. Second hull provides additional protection against leakase
 Tankers fined with hydroaatic controls to reduce leakage
1. If mpture occun controls sets intemal pressure to environmental pressure
2. When pressures are equalized leakage slows or stops
- - - - - -- -- - - - - - - -- -

Weather forecasts used to plot path of tanker and reduce accidents

I . Forecast good for three days
7. Tanker slow to respond to direction changes
3. Better long range forecasts would reduce accidents further
Tanker routes are planned to avoid collisions
1. Tanker routes are planned to avoid collisions
- - - - --- - - -

Clean up can be achieved by leaving the oil to disperse naturally

1. Heavy currents disperse the oil
2. Bad weather disperses the oil
3. Evaporation disperses the oil
4. Oil remains in environment but local pollution is reduced
The best clean up method is collection
1. Booms corral the oil into small area
2. Skimmers suck the oil out of the corral
3. Or. skimmers add adhesive to oil.
4. Oil is removed from the environment
Clean up can be achieved by spraying chernicals to disperse the oil
1 Chernicals break up the oil so that it is diluted in the water
2. OiI remains in the environment
3. The dispersants may be toxic

Table 16.

Method 1.2 and 3 Ratines for the Kev Conta

1 Key Concept Method 1 1 Method 2 1 Method 3 1

1 Link 1
1 Link 2
 Link 3 Unistructural 3
Link 4 Unistructural 3
I

Tankers 3-Link
Link 4 Unistructural 3
Link 5 Unistructural 3
Link 6 Unistructural 3
Link 10 Relational-2 4a
Link 1 1 Relationai-2 4a
Link 12 Relational-3 4b

.Link 13 Unistructural 3
Clean Up 6-Link
Link 7 Unistructural 3
Link 8 Unistructural 3
1

Link 9 Unistructural 3
Link 14 Relational-4 3c
Link 15 Relational4 4c

Link 16 Relational-3 4b

The links of Figure 12 repeat descriptions of concept relationships that are provided

in the text. ïhey represent leaming in the amturing. mode. Al1 unistructural links were rated

at the stage 3 Ievel of Method 3. Higher quality relational links were rated among the

divisions of stage 4. This finding is encouraging as it supports the thinking that structurai

knowledge is comparable to the link-procedural knowledge that generates it. Ir emulates

behaviour predicted by theory integration. It also suppons the reasoning that concept rnaps

have a psychological vaiidity that is comparable to that of the theones of cognition.

 The Novak and Gowin method reported the combined efKects of progressive

differentiation and integrative reconciliation. It provided scores for entire maps. Method 1

and Method 2 cm be used alongside the Novak and Gowin rnethod, they provide assessments

of the progressive differentiation and integrative reconciliation that is associated with

individual concepts. Unlike Novak and Gowin they provide scores for individual concepts.

Îut~krrswas classified as a 3-Link concept. It is associated with three unistructural

links. (-%cm I/p was rated as a 6-Link concept. Three of its links were rated at the

unistructural level, one link was a relational-3 Iink and two were reIational-4 links It cm be

said that ('femz iIp exhibits superior pro~ressivediflerentiation than Ta~~kers.

(('leunUp is

associated with more concepts. It can also be said that ( ' I ~ C I ~ilp
I exhibits hiyher level

integrative reconciliation than Tui~kers.On average the ('lem iIp links have a higher Method

1 ratiny than the Iinks associated with Ïàtzker-S.Oil Spilis exhibits intermediary levels of

progressive differentiation and integrative reconciliation. It is a 4-Link concept and al1 of its

links are unistructural links. The links associated with OilSpilLs are fewer in number and on

average are of Iower quality than the links associated with C'ka11 Ilp. The links associated

with Oil Spills have the same quality as links associated with Tcnlkers. but there are more of

them.

The above analysis demonstrates that the target text is suited to this thesis as it is

arnenable to concept rnapping. Furthemore methods 1.2 and 3 can be applied towards an

analysis of a map of the target tes. The next chapter presents the concept maps of three

students. The remainder of the thesis concems the analysis of these maps.
 Chapter 9

Assessing Structural and Link-procedural Knowledge

9.0 An overview of the chapter

Each of the grade 9 biology class' 27 students produced a map of the target text.

Data analysis began with the application of methods 1 and 3 to these maps. The Method 3

ratin~swere used to select six subjects for the metacoynitive interview. The intent when

applying Method 3 was to find two interviewees from each of the followinç categories. low.

medium and high cognitive activity. Only one subject exhibited low level cognitive activity

(low Method 3 ratings), so a third subject of hi& level coyitive activity was selected for

interview. The cognitive activity of the interviewees is presented below:

Subject 1: high level cognitive activity

Subject 2: medium level cognitive activity
Subject 3: medium level cognitive activity
Subject 4: high level cognitive activity
Subject 5: high level cognitive activity
Subject 6: Iow level cognitive activity.

This chapter discusses an example of each of the Method I link types contained in the

rnaps of subjects 1, 2 and 6. Assessments of ail six subjects are presented in appendix D.

Subjects 1.2 and 6 were selected for discussion as they represent examples of high, medium

and Iow level cognitive activity. Discussion of their links reveals the logic behind Method 1

and Method 3 ratings. It aiso reveals the nature of link-procedural knowiedge. In discussing

the links reference is made to the accretion, nmcturing and tuning modes ofleaming. The

discussion also considers the inference levels that govem link quality. These inference Ievels

are represented below to serve as a reminder.

 Level 1. At this level inferences involve incorrect reasoning. Two scenarios are
possible. In the first instance the inference connects two data items that are not truly
associated with the concepts of the link. Although the inference is correct the data
involved renders it irrelevant to the link. in the second instance the data involved is
appropriate to the link but the inference is incorrect.

Level 2. Simple listings of facts that fail to connect data throuyh inferences.

Level3. Inferences based on fact, (inferences that replicate descriptions found in the
instruction)

Level4. Inferences based on facts that were not overtly linked to the concepts in the
instninion (correct reasoning that builds upon information presented in the
instruction).

Level 5 . Inferences based on facts that were taken corn outside the instruction
(correct reasoning that could not have been inferred from information presented in the
instruction).

9.1 Interrater reliability

Interrater reliability was established for metbods 1, 2 and 3 by comparing the ratings

of the researcher and two additional raters. The maps involved in this cornparison were the

maps of subjects two and three. Subjects two and three produced a total of 47 links. Six of

these links were general knowledge links. GeneraI knowledge links represent points at which

the learner's enant knowledge has been associated with the target text. They add context to a

map rather than replicatinç or building upon it. For example, link 1 of subject 2 states that the

"shipping industry manufactures tankers." This fact, while accurate?is not mentioned in the

text. nor is it an example of creative thinking. The link represents a point where the mident

associated the text with eisting knowledge. That is, it places the contents of the target text

within the larger context of the knowledge that the student considers to be associated with
oils spills.

General knowledge links were handled in the following way.Raters were provided

with representations of the maps of subjects two and three. The general knowledge links of

these representations were preidentified by the researcher. The critena for identification were

that a link should not replicate or build upon the text, not should it involve creative thinking.

Raters were told to rate the remaininy links by the scales of methods I and 3. They were not

asked to identiG geeoeeral knowledge links as it was felt that they would find it too difficult to

distinguish them From other link types. They were toid to be on the look out for links that

represent generd knowledge. They were also told to brins such links to the researcher's

attention.

For Method 1 the three raters agreed on 14 of the 4 1 none yeneral knowledp links.

Expressing the agreed upon number of links as a percentage of the total generated an

interrater reliability of62%. Interrater reliability was cdculated for Method 3 in the same

way. It was found to be 5 1%. These low-Ievel interrater reliabilities may be due to poor

training of the raters? inadequate descriptions of the marking scales. and lack of rater

communication. The issue of training is considered fira. The raters were teachers who ageed

to mark the concept maps in the last week of tem. They had no time to partake in practice

marking exercises. One or more durnrny nins with exemplary links would no doubt increase

interrater reliability Ideally these practice runs would garner feedback from the raters that

would be used to irnprove the descriptions of the marking scaies of Method 1 and Method 3 .

It would dso be preferable if the raters were to work together during the dummy mns. Their

communications could be focussed toward reaching a consensus on link types that are
 118

difficult to identifi (general knowledge links, links with fiapented data etcetera). Intenater

reliability was 100% for Method 2. The three raters reached agreement on number-link

ratings for the three key concepts of the maps of subjects 2 and 3.

9.2 The nature o f link-procedural knowledge

Link-procedural knowledge foms inferences. Detenining how these inferences are

formed will reveal the nature of link-procedural knowledge. Methods 1 and 3 use two

assumptions to explicate the nature of link-procedural knowledge. The first assumption is

that concept mappers express items of data in ways that reveal how inferences are formed.

The second assumption is that linking statements describe relationships between concepts. If

these assumptions are not met it is hoped that linking aatements and data will be worded in

such a Nay that simple reasoning can be used to divulee how inferences were formed. Most

of the links created for this thesis failed to render the reasonins behind inference creation

explicit. It is suggested that a more extensive training program (for the students) would

increase the incidence of links that are stmctured so as to reveal the manner in which their

inferences were fonned.

9.2.1 Subject L

The paper concept rnap of subject 1 is reproduced in Figure 13. The links of this rnap

are described in Table 17. By Method 1 there are: nine extended relational-7, links. eight

extended unistmctural links, three relational-2 links and two prestructural links. By method 3

there are nine aage 6a links, eight aage 5 links, three aage 4a links and two aage 1 Links.A
number of links could not be classified. For example. link I was not suited to analysis by

Method 3 as it was deemed to be a general knowledge Iiink. Method 3 cannot account for

general knowledge links. It recognizes inferences that replicate or build upon the text. It also

recognizes inferences that contain creative thinking. Links 22, 26 and 27 were excluded from

the analysis as they contain imprecise references to inferences contained in other links. It was

not possible to detemine the amount of data compnsed in these links.

Tankm 7 10 Cltgi Up
3 *-
-r-->-13 --T-F14
.--A- _Y - -

44 7 - - Ai?
DoubleHuil Am A-
N a v i g a t W \ Pawerful wulaJ
- - - -. .
- -- -- - -

En&e *ph Chanhl

Oil Radar - --- - -
c e L -- -
Mspersion
-- - - -

m r e 13. Subject 1, an example of high level cognitive activity

Table 17

The Links of Subkct 1

Oil q d l s m u t be clciuid up
a) to prwtrit ~ w i r a n m m t a impact
l
b) to siilvsgc oil

A doublc hiill is a safcn; p r d u r c

a) crashe?; musi bc harder to puncturc boih hulls
b) oil containcm arc ILXS 1ikcIy to bc b w ç h d
- - -

Oil mntaincrs with the sarnc inna prwsurr: as outsidc pnwurc will
spill ILM oil
a) I w oil siills
b ) the dis;i';ttx is ILSS p r i t

Mon: poivcrful mgincs w il1 makc. stiips dtr

a) ~ S I L Ttumin3 u iil bc possible
b) obstacles can bc brtttx avoidd
- - .- ------ -- -

Tankers wt $
a) hypothctiçally: morc S. bcttcr ship
Satkh p r d u r c s cost $
a ) hpothetically : morc S. bctttr prwedurt-s
b 1- S. w o w p r o d m s
iank~rsrcquih' w o f k ~ r ~
a ) more workm. l a s r;hanc;c.of probiems
b ) l a s workm- morc chrince of probltms

W o r k m cost S
a) more S. bcita qliiiiity of w o r k m
b) las S. Iowx quafi5 of worktxs
Workcxs app- de' p r o d m
a ) b s t t a quriity w o r k m = bettm quatic p r d m
b) 1 0 % quality
~ ~ work~rs= low-tr qualin. pror;cdm?i
A clmn up requins worktm
a) higher qualie w o r k m = higher qualie dem up
Iowm qualie w o r k m = Iowa quaic çlem up
 Warktxs must apply p d m
a) highm qualie worktrs = bighm quality appliçaiion of p d u r e
b) lowm quaiity worktrs = Iowa quality application of p r d u n . ;

Naîural d i s i i o n is a clean up p d m
a) it is 80N cfkwtiuc
b) somctim~?ibas sidc ctlitxts

- . -- - - - - - - -. .-- - -. - -.

Chtmical disicrsion 1s a clcm up prwdurr:

a) it is samçtimtir ctktivc
b) it a1way-i hm sidc etTtcts
Worticrs arc not n d d for natural dispersion

T h t x wst mon.;

This does noi wst mon-

Baausc of mm! abovc: reasoas. workcrs arc n~ponsiblt.for oil sptlls

Corporations detamine hou-much M a p worktz-s pt.so corporritiot~

detemine workrrs

From if 22-26 above. wc c m con~ludcthat the g o v ~ m ~kn t

ultimstcb mponsiblr: for oil spilis

Link 17 is an example prestructural link. It refers to the relationship between

CMrction and ('km Up PrucrJiires. Its data stipulates that collection gets al1 of the oil that

is spilled. It also concludes that collection is a 100% effective clean up method. The text does

state that oil is completely removed from the environment once it has been collected.
 122

However it does not imply that al1 the oil that spills frorn a tanker can be collected. Thus

collection is not 100% effective as some of the oil that is spilled will not be collectible. The

erroneous reasoning in link 17 lead to its prestructural rating.

Link 18 is an example relational-2 link. It connects the concepts (%ati I l p P r o c d ~ ~ r r

and Chernid Dispersim. The linking statement stipulates that chemical dispersion is an

example clean up method. The data accurately relates the teds description of the

effectiveness of chemical dispersion. One datum States that chernical dispersion "is sometimes

effective" the other says that "it always has side effects." The data of this link does not elicit

how the linking statement was foned. Furthemore reasoning about the data and linking

statement fails to determine the rnethod of inference formation. This link was said to be

disconnected.

A nurnber of the links exhibit the disconnected structure of link 18. They were rated

by f i r ~assuming that the mapper would have described the reasoning behind their formation

if they had been exposed to a more extensive training progam. Next, data were counted and

it was determined whether the link replicated. built upon or restructured the text. Links that

repeat the text represent the strucniring mode of leaniing. The tuning mode is represented by

links that improve upon descriptions of relationships as provided in the text. That is. tuning

links integrate the mapper's extant knowledge into the text's description of a relationship.

Accretion links coruiect a relevant concept that is not mentioned in the text to a concept that

the text does mention. That is mappen add extant knowledge to the text when creating

accretion links. The amount of data and the learning mode constitute the structure of the link.

Once link structure is known it is rnatched againa the scales of Methods 1 and 3. Table 18
 123

indicates the association between the leaming modes and the scales of Methods 1 and 3. Link

18 integrates two items of data and it represents the structunng mode of leaming. This

structure corresponds to the relationai-2 rating of Method I and stage 4a of Met hod 3.

Interpretiny links in terms of the theories of reading comprehension requires an

understanding of the inference that lead to link creation. Link 18 was not interpreted in terms

of the theones of readiny comprehension as its disconnected data failed to elicit the reasoniny

that lead to the creation of its inference. Ascertaining the method of inference formation also

elicits the functioning of link-procedurai knowledge. In links were the rationale behind

inference formation was identified. link-procedurai knowledge was Found to mimic the

actions of simultaneous processing.

Table 18

The Relationshh Between Learning Modes and Methods 1 and 3

Strucniring Mode Aç~mtionMode Tuning Mode

Mshd I Methad3 Msthod 1 Methoci 3 Mcthod 1 Mcthod 3
Mu1tistruc;ttirai stage 2 Exiendd Multistructtual Strige 2 Esttadd Abs&act Stage 7

Untstniçtural Stage 3 Etrad& U n i s ~ c ~Sl t i p 5 J - ~ . Absaact

. ~ ~ ~ ~ Stage 8
Relationd Stage 4 Extendcd Rt.hional Stage 6

Link 6 can be interpreted in tems of the Kirby ( 1991 ) and Jacobs and Paris ( 1987)

theories. Successive processing held the concepts Nmigationul Radar and Safey Procedztres

in working memory as discrete units. Simultaneous processing sought to descnbe the

relationship between the concepts. It deterrnined that d e t y procedures seek to lower

accident rates. It also detemiined that the use of better navigational radar can reduce

accidents. This knowledge lead to the conclusion that the goal of Scrftiiy PruceJ11re.s(lower

accident rate) cm be met by Brrrrr Navigutiu~talRadar. Conclusions represent the result of

cognitive activity, they are expressed in linking statements.

Link 6 is an example extended unistructural link. Its single datum (less crashes will

occur) is used to generate a linking statement (better navigational radar will allow obstacles

to be avoided). The linking statement irnplies that navigational radar should be adopted as a

safety procedure. It represents the accretion mode of leaming as it associates pnor

knowledge of navigational radar with the text's admonition for the adoption of safety

procedures. Thus link 6 involves a singe item of data and an inference that is relevant to. but

was not presented in the text. This stnicture matches Met hod 1 's description of ex~ended

unistructural links. it also corresponds to stage 5 of Method 3.

Link 7 connects the concepts More Puwrrfi<lDtgirws and .Sufep P r ~ ~ e c h ~Like

r~s.

link 6. link 7 represents the accretion mode of leaming. The concept More Pu~tv~$d
E~zgines

is relevant to the text but was not presented in it. The inference of this link involved the

knowledge that safety procedures seek to lower accident rates. It also involved the

understanding that more powerful engines can increase tuming speed and thereby increase

the tanker's ability to avoid obstacies. Thus it was concluded that more powerfbl engines can

reduce the number of accidents and rneet the goal of d e t y procedures. When simple

reasoning was applied to the data and linking statement of this link it provided an indication

of how the inference was forrned. Sequential processing held both concepts in working

memory (as discrete units) while simultaneous processing sought an inference to comect
 125

them. Thus reasoning indicates that the formation of this link emulated the comprehension

strategies of the theones of readiny comprehension.

Link 7 is an example extended relational-2 link. The extendecl component of this

rating was derived h m the fact that the link is an accretion link. It took extant knowledge

about engine power and associated it with the text's description of safety procedures. The

relationaI-2 component was derived From the link's structure. Two items of data were

integated to fonn a Iinking statement. The structure of link 7 matches the Method 3

description of stage 6a. Stage 6a refers to link-procedural knowledge that connects two items

of data in a manner that builds upon information presented in the text.

It is suspected that Subject 1 ran out of time when creating Figure 13. This

supposition is based on the fact that links 19 to 25 are incomplete. They contain linking

aatements but no data. Methods 1 and 3 were associated with these links by assuming that

they contained one item of data that is identical to their linking statements. The iinking

statement of link 19 "workers are not needed for natural dispersion" builds upon information

provided in the text. Thus link 19 was said to represent the accretion mode of learning.

Accretion links that contain a single datum are extended unistructural or stage 5 links. Links

20 to 25 also received extended unistructuraYstage 5 ratings. they involve the accretion mode

and they integrate one item of data.

Twenty-three of Fiyure 13's twenty-seven links were marked by Methods l and 3 . Of

the Method 3 (cognitive activity) scores, 39% were of stage 6a 39% were of nage 5. 13%

were of stage 4a and 9% were of stage 1. These results can a h be interpreted in tems of

inference levels. The inferences of stage 6a and stage 5 involve correct reasoning that builds
 126

upon information presented in the text. They constitute 78% of the links in Figure 13 and

they contain level4 inferences. The inferences of the stage 4a links replicate descriptions

found in the text. They contain level3 inferences and represent 13% of Figure 13's links.

Stage 1 links contain incorrect reasoniny. They contain level 1 inferences and make up 9% of

Figure 13's links. The predominance of level4 inferences in Fiwre 13 lead to the hi$ level

cognitive activity ratiny for subject 1.

When Methods 1 and 3 were applied to the map of subject ? they revealed a

predominance of level3 inferences. Level3 inferences were the only inferences produced by

subject 6. .411 of subject 6's links were unistructural links. Subject 2 produced higher quality

links than subject 6. Subjects 2 and 6 were said to exhibit medium and low level cogitive

activity respectively. The link-procedural knowledge of these subjects is discussed in the

following sections.

9.2.2 Su bject 2

The paper concept rnap of subject 2 is reproduced in Figure 14. The links of Figure

14 are described in Table 19. By Method 1 there are: one extended relational-3 link. one

extended relational-2 link, one relational-4 link, six relationai-3 links. six relational-2 links

and seven unistructural links. By Method 3 there are one aage 6b link, one aage 6a link one

stage 4c link six stage 4b links, six stage Ja links and seven stage 3 links. Link 1, was

classified as a general knowledge link as were links 3 and 4. Link 9 was deemed unmarkable

by Method 1. The problem is that the data of link 9 are overly tiagrnented. At least two of the

data items are surplus to requirements as they are part and parcel of ideas presented
elsewhere in the data. For example. the datum "preventing oil spills" encompasses "if

accident occurs. stop them from leaking." The data "stop them from haviny accidents" is

repeated in the datum " avoidiny oil spills." Furthemore this datum encompasses the notion

of "plotting routes to avoid accidents." Let's take a closer look at these data.

a. Preventing oil spills

c. If accident occun, stop them from leaking

The "preventing oil spills" refers to actions that can be taken to prevent oil spilis. The text

sussested that tankers be fitted with hydrostatic controls and double hulls to inhibit oil

leakage &er a collision occurs. These control measures are referred to in the daturn "if

accident occurs. stop them tom leaking." Raters should be wary of mappers who overly

frament their ideas and knowledge. Lhdetected. this situation cm lead to erroneously
ci

eievated Method 1 scores. For example, if al1 the data of link 9 were considered appropnate

it would have gamered a "relational-5" score. Ideally the data items "preventingoil spills" and

"if accident occurs. stop them from leakin&' would have been presented as a single datum.

This datum might have been "measures have been implemented to prevent oil spills once a

collision occurs."

The followin%(link 9) data items also suffer from too vigorous a fkagnentation.

d. avoiding oil spiils

b. stop them from havins accidents

e. plotting routes to avoid accidents

The data "avoiding oiI spills" and "stop them tom having accidents' are essentially redundant.

They both refer to the measures that Maritime Authonties recommend should be taken to
 128

avoid tanker accidents. "Plotting routes to avoid accidents" is actually part of the datum

"avoidiq oil spills," it refers to one of the Maritime Authonties' prevention measures. ldeally

dl three data items would have been presented as a single datum. This datum miyht have

been "Maritime Authorities have introduced measures to avoid accidents. One of these

measures is the plotting routes to avoid other vessels and inclement weather." A follow up

interview with subject 9- would be needed to determine how many lqitirnate data items are

present in link 9.

Figure 14. Subject 2, medium Ievd cognitive activity

Table 19.

The Links of Subiect 2

TIink~mare iwgc boats

a) dt.sipcd 10 hold lots of oil
b) biggm ship. h d io rnancuvw
C) long t i m to slow do~vn

Oil suppiy providd bu comprulics

a) b q i d m
b) till thcm with oil
G ) m d to ohm compmics

Oil qdb cause pollution

a) q m a d through cwimnm~rit

Pollution kills plants md ;uitmds

a) Animal's habitat is dc.stroyd
b) Animal hm nowhm to survive

Pollution causes deiith and hea1th problems

a) air pollution Ica& IO brcathîng probltms
b ) w t a bccomc3 poison

Oil spills nrxri to be prevmted and avoided

a) prevtmting oil spius
b) stop thcm h m having accidents
ç) if accident occurs- sqop thsm h m l&g
d)avoidiag spilIs
c) plotting routes to avoid accidents

Maritime authorities help with avoidanw

a) rrcommend keepmg at 1-t 10 d e s away h m m a t
b) tanli.mtiie route plans with maritime authontir=s
c) k ~routes
t ~.separate
1.aws hdp with prwcation
a ) oil pollution açt
b ) n w tankm built with double hull
c) hyhstatic pW4iw wntrols
Mort: media will bclp with phwmtiou
a) m d i a att~niioncan convincc sbipping indus. tn u.w pnl-cntion
mcthds and tc~hnolo@
b) increa.~pubtic awrucricx!
- .- --

Oil spills n d ro bt. clcancd up siiikly

a) t'ind the d i t mcthd
b) u.w wbat is avaiIablc
c ) start dean up
- -

Ont. techniqut. of c l c a up is natwal dispersion

a! to c l a n up the oil is d i s p d nriturally

NaîuraI diqxxsion c m only bt: u s d in minin w a t h t r conditions

a) oil di'ip~r'rwiby waws and harsh conditions
b) wava break up the ail
- -

hotbtr trx-hniqur: is mllc~tion

a) 0111s wlIcctd and n m o v d fmrn waicr md ~ w again
d

Oil t'irst containai by booms

3 ) surround oil. stops h m spreading

oil LS wUatd uing skimmm

a) hvo methai of skimming to mll~wtoil

Ont. method of skimmiag is suction

a) suck od otYwatcr
b) also m c k up lob ofivatcr: 90 %
- .

Otha method of skimming is adhesion

a) swaq material t h u @ 9piU
b) oil sticks to mataid
c ) oil carid away h m spi1
d) oil squcczcd out of mataiai
Sorbenîs usai to absorb oil coll=ctd bu booms
a) c m bt. natural or >inthetic
m i e h of shapes- sheeîs. mops etc.
2 ) used h r ~maii pwkets usuaiiy

Last methoci of clan up is chtanical d l p m i o n

a) chtmicals useci to dis- oii

Rate of rfisprsion depends on waier t m p

1) wld water- oil changes chmct&tics a d c m o t be dispersai my
more
 Dispzrsluits brcak up oil sIick
a) b d up oil slick
b) slick diluies into water
c) cm only bt. uscd an w i a i n oils

- - -- --

k i c n t ~ s t sarc working on wsys ta dc~elopaew ~&niques of chmical Rclational-2

dissrsion
a) a n w b n d of oil di.c,p:rs;uits- living argiinkms that f k d on oil
b) tving to tïnd Icxs tosic ways ol'disptming

Fiyures 13 and 14 have several linking structures in common. For example. they both

contain relational-', links. Link 25 of Figre 14 is a relational-2 link. It is considered here to

illustrate that link-procedural knowledge operates in a similar manner across different

subjects. The sequential processing of link 35 held the concepts, d i r p l m s and Bisp~.r.wm.

in working mernory as discrete units Simultaneous processing sou& an inference to connect

the concepts. The second datum of this link "Airplanes spray dispersant on the oil" refers to

the manner in which airplanes are used in clean up. The first datum "Responsemust be fast"

reveais that dispersants have to be applied quickly. The inference of link 25 generates the

realization that airplanes meet the needs of chemicai dispersants. They are the fastest means

of delivering chemicals to an oil spill. The Iink-procedural knowledge of link 25 associates

two items of data in a manner that replicates the text. fts actions correspond to those of the

link-procedural knowledge that generated relational-2 links in Figure 13.

Uniike Figure 13, Figure 14 contains links that were rated as unistructural/stage 3.

relational-3/aage 4b and relational-4/stase4c. A description of the inferences of each of

these link types is provided below. The description omits reference to the theories of reading
 132

comprehension. In al1 cases it is assumed that link-procedural knowledye fùnctions as per the

description of simultaneous processing that was provided for the links of Figure 13. That is it

is assumed that concepts are held in working memory as discrete units while simultaneous

processing generates an inference to connect them.

Link 22 is the example unistructurai/sta~e3 Iink. It contains one item of data and it

represents the structuriny mode of leaming. Its linking statement notes t hat ('hrmical

Bispzr.si~nis an example < 'ka11I/p procedure. Only one item of data is present in this link. It

states that "chernicals are used to disperse oil."This information reveds that chemical

dispersion is a clean up method for once oil is sufficiently dispersed it is said to be cleaned up.

The single datum of link 22 does not match the Iinking statement. Page 74 of this

thesis discussed unistructural links whose datum differs from the linking statement. They

were said to involve an item of data that the mapper is unaware of or unable to express. The

understanding of these links was said to be intemediary to the unistructural and relational

stages. Each unistructural link in Figure 14 possesses this intermediary structure.

Link 24 is the example relational-3/ stage 4b link. It differs from relationai-2stage 4a

links in that it contains an additional item of data. The concepts Dzspersu~rtsand Chrmtc~af

Disperrsa~sare comected in link 24. Two items of data are used to indicate how dispersants

work. They "'break up oil slicks" and the "slick dilutes into water." Both of these data items

are integrated in the linking statement which states that "dispersantsbreak up oil slicks."

Using these data and the linking statement it is possible to reason as to how link-procedural

knowledge comected the concepts of the link. If it is known that "dispersants cause oil to

disperse" and that "dispersed oil is diluted in water." then it is possible to conclude that
"dispersants break up oil slicks."

The third datum of link 24 stipulates that dispersants "can only be used on certain

oils." It is difficult to reason how this datum was associated with the data items "break up oil

slick" and "slick dilutes into water" by the inference of this link. The problem is not fatal

however provided we assume that subject 2 did integrate the three items of data but lacked

the traininy to express the relationship in full. This assumption is not unreasonable for a

relationship associatins al1 three data items is described in the text. Subject 2 would have

been aware of this relationship. Links which use three items of data to replicate the text

match the Method I description of relational-3 links. They also correspond to the 4b stase of

Method 3 .

Link 20 is the example reiational-4 link. Like link 24 it represents the structurin~

mode of leaming. Stnicturing is indicated whenever a link repeats descriptions that are found

in the teM. Text replication suggests that the mapper possessed no additional knowledge

regarding the link in question. Link 20 connects the concepts Adhrsioti and Skimmrrs. Its

linking staternent "other method of skimming is adhesion" indicates that skimming is an

example of the adhesion clean up method. The data of this link is presented below. It

describes the skimmins operation and in doing so it indicates that skimming is an adhesion

method.

a) sweep material through spi11

b) oil sticks to material

C) oil carried away from spill

d) oil squeezed out of material

The data of Iink 20 exemplifies the link structure that was envisioned during the creation of

methods 1 and 3. Link 20 is encouraging as it indicates that students are capable of

explicating their lin k-procedural know ledge. Further training wouid increase the production

of this calibre of link.

Link 13 is the exarnple extended relational-3 link. The text describes a number of

clean up methods and hints at the need for safety. Link 13 improves on the text as it correctly

assumes that before clean up cm beçin, some time should be spent selecting the clean up

method that will minimize environmental impact. Link l 3 is not an extended abstract link as

the improvement foilows logically from the text. it does not represent an example of creative

thinking. Like link 20. the data of link 13 is worded to reveal the actions of its link-

procedural knowled~e.

Figre 14 represents medium level cognitive activity. Twenty two of the twenty six

links were marked. Of the Method 3 scores, 5% were of stage 6b. 5% were stage 6a 5%

were stage 4c. 27% were stage 4b. 27% were stage 4a and 32% were stage 3. The links of

stages 6b and 6a build upon information presented in the text. They conaitute 9% of the

rnap's links and represent level four inferences. The remainin_glinks repeat information that

was presented in the text. They represent 91% of the map links and contain a predominance

of level3 inferences. Subjects whose maps contained a predominance of level3 links were

rated as medium level cognitive activity subjects.

 9.2.3 Subject 6

The paper concept map of subject 6 is reproduced in Figure 1 5 The links of this map

are described in Table 20. Each link is a unistmcturall stage 3 link. Link 3 was selected as an

example

m u r e 15. Subject 6, low level cognitive activiîy.

unistructural link. It connects the concepts Cases and Politicx The linking statement of this

Link is exactly the same as its datum "Politicsare a cause."One item of data and a correct

inference matches the unistmctural iink of Method 1 and stage 3 of Method 3. The links in

Figure 15 contain level 3 inferences, they repeat descriptions of concept relationships that are

provided in the text. The predominance of unistnictural Iuiks lead to the low level CO-@ive

activity rating.
Table 20

The Links o f Subiect 6

1 'Thtn. arc w e r a l criuses to oil hpills I Jnisiniçtural 3

a) there are s c m l M L -to oil spills
2 'îiinkcr accidm~sWC a cauw ( !nistniciural 3
s1 tanktr awidtnts a m u n t for 1 2% of oil in the WGIII
3 Politics iin:a cauw t Jnistructural 3
a ) politics iin: a cause

I Kwage is a political tactiç

a) wcrigc is a politicni tactic

I Sdkty is a larp conwrn

a) thc Oil Pollution Act

9 CIcan up wsts $ CJuisuuctural 3

a) clcm up costs S

1O Naturc is destroyed 1 hsmctural 3

a) nature is dcstro_vd

9.3 A summary of the chapter

Subjects 1,2 and 6 were selected for the above analysis as their maps exhibit high,

medium and low level cognitive activity. A number of problems were encountered when

trying to assess the link-procedural knowledge of the midents maps. The sarne problems
 137

arose during the assessment of the maps of subjects 3 . 4 and 5 . These assessment problems

were caused by students' failure to structure links correctly. Methods 1 and 3 require that

linking statements describe relationships, they also require that data reveal how the

understanding contained in reiationships was formed. Although most linking statements

described concept relationships. data oflen failed to elicit inference formation. When the

requisite link structure was not produced. raters had to determine the quality of link-

procedural knowledye through an interpretive process that considers possible associations

between data and linkiny statements. Links such as link 13 and link 30 of subject 2 did

contain the requisite structure. The presence of these links lead to the suggestion that a more

extensive training can reduce the problem of inadequately structured links.

The interpretive rating method elicited the nature of Iink-procedural knowledge. Its

actions were said to compare to those of sirnultaneous processinç. That is, it sought to

identiS, relationships that connect discrete units (items of data) in workiny memory.

In closing it is noted that the aructurd and link-procedural knowledge of subjects 4.

5 and 3 were also exarnined by the above methods. Subjects 4 and 5 exhibited high level link-

procedural knowledge (high level cognitive activity). Most of their inferences were of leveI4.

Subjecî 3 produced a predominance of level 3 inferences. The cognitive activity of this

subject was rated at the medium level.

 Chapter 10

Assessing the Psychological Validity of Concept Maps.

10.0 An ovewiew of the chapter

This chapter examines the concept maps of the six subjects who were chosen for the

metacognitive interview. Methods 1 and 3 are used to investigate the daim that concept maps

represent mental models of problems. Further evidence for this argument is presented by

examining the students' Method 3 and metacognitive interview scores. The results of both

investigations support the suppositions derived from theory integration. Structural and link-

procedural knowledge were found to have comparable quality. Metacognitive activity was

found to be isomorphic to the cosnitive activity that it suides.

The concept maps of the six subjects are also subjected to an e:wnination by Methods

1 and 2. The methods were used to assess the progressive differentiation and inteyrative

reconciliation associated with individual conceptS. The investigation met the second goal of

the thesis in that it demonarated that methods 1 and 3 cm be used to assess concept maps. It

also indicated that Methods 1 and 2 provide assessments of individuai concepts that

complement total map scores that result fiom the application of Novak and Gowin's ( 1984)

method .

10.1 The relationship behveen structural and link-procedural knowledge

Methûd 1 and Method 3 ratings can be used to connrua a picture of the relationship

between structural and link-procedural knowledge. Table 2 1 illustrates the link types that

were produced by subject 1. Two thinçs are apparent in this table. Fust, links with the same
Method 1 rat@ are associated with the same stage of Method 3. For example. each

prestructurai link had a stage I rating and each relational-:! link had a stage 4a ratiny.

Second, the higher the quality of the met hod 1 rating,, t he higher the Method 3 stase.

Table 21

The Structural- Link-Proceûural Knowledw Relationshio o f Figure 13

1 Method 1 (Structural KnowIedge) 1 Method 3 (Link-Procedural Knowledge) 1

Prest ructural Stage 1
1 Relational-2 1 Stage 4a 1
1

1 Erctended Relational-2 1 Stage 6a 1

.An analysis of Figure II adds to our understanding of the relationship between

structurai and link-procedural knowledge. The analyses of Figures 13 and 14 are combined in

Table 72. The findings of Table 22 are not unique to subjeas I and 2, al1 six subjects

exhibited this comparable quality of knowledge across methods I and 3. Every link of a

particular Method 1 rating was associated with the same stage of Method 3.

Table 22 illustrates thar concept maps behave as per the predictions of theory

inteyation. The quaiity of structurai knowledge increases aiongside that of link-procedural

knowledge. Ascertaining that the two knowledge forms exhibit comparable quality indicates

that concept maps exhibit properties that correspond to the predictions of theory integration.

Concept rnap theory has it that maps represent mental models of problems. Consequently the

finding that concept rnaps behave as per theoretical prediction is taken to be supportive of the
notion that mental models are represented in maps.

Table 22

The StructuraCLink-Procedural Knowledee

- Relationship

1 Method 1 (Structural Knowledge) 1 Method 3 (Link-Procedural Knowledge)

1 i~nistructurai 13

1 Relational-4 1 4c
1 Extended Relational-2 1 6a
1 Extended Relational-3 1 6b

10.2 Progressive differentiation and integrative reconciliation

Novak and Gowin ( 1987) recognized the roles that progressive differentiation and

integrative reconciliation play in the learning process. They also recognized that these

processes influence the nature and form of concept maps. Progressive differentiation affects

the number of links. Inteyative reconciliation determines link quality. Novak and Gowin's

scoring technique provides a total score for each map. Their score bundles the effects of

progressive differentiation and integrative reconciliation. Methods 1 and 7 complement the

Novak and Gowin technique. They mess individual concepts rather than entire maps. They

provide separate scores for progressive Werentiation and integrative reconciliation.

Table 23 provides the information that is needed to compare the key concepts (Oïl
 141

SpilLs, Tat1kers and ('Itratz I l p ) of subject 1 to the key concepts of ot her subjects. Table 24

and Table 25 provide similar information for the Oil SpiI1.s concept of subject 2 and subject 6

. The progressive differentiation associated with key concepts c m be compared across

different maps by Iink enurneration. Oii SpilLs was presented as a 4-Link concept by subjea

1. a 5-link concept by subject 2 and a 2-iink concept by subject 6. Subject 7 created the rnoa

links and can be said to exhibit supenor integrative teconciliation for the Oil Spiils concept

Table 23

The Kev Concepts o f S u w t i

1 Suborûinntc Concepts
I SOLO Rating of Luilr

Clem CJp 1 Estcndrd Relational-2

Oil SpilLs
Table 24

he Oii Spiils Concept of Subject 2

SOLO Ratine ot' L i d

I

1 Ckm IJp

Table 25

The Oil S ~ i l l sConce~tof Subiect 6

Comparing the integrative reconciliation associated with individual concepts is a more

complicated flair. The complications derive from the weaknesses of Method 1. Section 7.7

covers these weaknesses in full. It notes that maps can be compared through their integative

reconciliation when a concept is associated with the same superordinates and subordinates

across different maps. For example, suppose a concept has three unistruchird links in one

map and three relational-2 Iinks in another. If the concepts involved are identical it cm it be

said that the integative reconciliation and the leaming of the latter map is supenor to that of
the former.

When super- and subordinates differ across maps so does the progressive

differentiation. Cornparisons of learning in this situation have to consider the effects of

integrative reconciliation and the effects of progressive differentiation. Suppose a concept is

linked to the same number of subordinates in two maps. but the concepts involved are

different. For example. Tunkers might be linked to ('ompatiirs and Oil SpIlls in a hypothetical

rnap A and to (.'ompat~ie.sand Nm~@uiot~in a hypothetical rnap B. The Tntlkers to

('omparrie.~llnk can still be compared through Method 3. The rnap with the highest Method 3

rating can be said to exhibit superior learning for that link. In the Tutrkers-Oil SpilI-sand the

links. progressive differentiation is quantitative- identicai. but

îcu~ker.s-Nc~c'ig~~~iot~

qualitative differences exia between the proyressive differentiation of the two links. Method

2 is insensitive to these qualitative differences. Progressive differentiation is said to be

identical across the maps and rankings are provided throu& Method 3 ratings. Suppose both

link of Map B is
links of Map A are unistructurai. Suppose also that the Tn,~ker~-C*umpcit~ks

unistructural and the Tatrkt.rs-iVc~vigatiotrlink is relational-3 link. Map B is said to exhibit

supenor leanllng. The qualitative progessive differentiation differences in Tat~kers-

Nuviga1iot1and Ïankers-Oil Spiils are ignored.

Cornparison may dso be possible when a concept contains more links in one rnap than

another. For exarnple, suppose a concept has three uniamaurd links in one map and two

unistructural and two relational-2 links in another. The combination of more and bener

quality links in the second rnap means that it exhibits supenor progressive differentiation and

superior integrative reconciliation. The second rnap is aiso said to exhibit superior Iearning as
 144

leaming manifests in concept maps as the combined effects of progressive differentiation and

integrative reconciliation.

Methods 1 and 2 are of little use when the number and qudity of links are in

opposition across different maps. This limited utility is related to the fact that the methods

cannot compare the cognitive wonh of one link type against another. For example. suppose a

concept is linked to five unistructural links in one rnap and to three relational-3 links in

another. Methods I and 2 cannot detemine which map is superior as they cannot express the

cognitive value of unistructural links in tenns of relational links. That is they cannot

determine how many unistructural links equal one relational link.

The following paragraphs compare the prog-essive differentiation and the integrative

reconciliation of subjects 1. 3 and 6 . To begin subject 6 is compared to a hypothetical

subject. Subject 6 links Oz/ Spifis to (la.se.s and ('ieat~I l p through two unistnictural links

Suppose the hypothetical subject links Oil Spilfs to C1asc.sand CIea~tI l p through a

unistnictural link and a relational-2 Iink. These maps differ only in their link quality and it can

be said that the hypothetical subject exhibits superior interyative reconciliation and supenor

learning.

Subjects 1. 2 and 6 presented Oil Spiiis as a superordinate concept. They al1 linked it

to a different number of subordinates. Only one subordinate, C11ra7r Up. was cornmon to al1

three maps

Subject 2 presented Oïl Spills as a 5-Link concept

Link 6. Oil Spiiis-Pollution. A unistructural link

Link 9. Oil Spills-Prevention and Avoidance. An unmarkable link

Link 5. Oil Spills-Wars. A relational-3 link

Link 3. Oil Spills-Tankers. A general knowledge link

Link 13. Oil Spills-Clean Up. A relational-3 link

Subject 6 presented Od Spiils as a N i n k concept

Link 1. Oil Spills-Cases. A unistructural link

Link 7. Oil Spills Clean Up. A unistructural link

Subject 2 produced 3 markable links. The general knowledye link (link 3 ) could not be

marked as Method 1 is insensitive to the inferences of this link type. Link 9 was unmarkable

as it contains redundant items of data (see Chapter nine). Subject 6 produced two markable

links. Two of subject 2's links were of higher quality than the links produced by subject 6 .

The combination of more and better quality links leads to the conclusion that subject 2

exhibited superior progressive differentiation and superior integrative reconciliation. The

extent to which the integrative reconciliation is superior is unknown. The progressive

differentiation was superior by two links in the map of subject 2.

Subject 1 presented Oii SpiMs as a 4Link concept

Link 1. Oil Spills-Tankers. A generai knowledge link

Link 2. Oil Spills-Clean L'p. An extended relational-2 link

Link 22. Oil Spills-Workers. An unmarkable link

Link 27. Oil Spills-Govemment. An unmarkable link

Subject 1 linked Oil Spilh to four concepts. Only one of those links was markable. Link 1

could not be marked as Method 1 is insensitive to the inferences of general knowledge links.

Links 22 and 27 were not marked as they refer to inferences in other links. These references

failed to reveal the manner of data integration (see Chapter 9). The links of subject 2 exceed

the links of subject 1 in nurnber and quality. It can be said that subject 2 exhibited supenor

integrative reconciliation and superior progressive diflerentiation. The extent to which the

integative reconciliation is superior is unknown. Quantitatively speaking, the progressive

differentiation of subject 2 was superior by two links.

Method I cannot compare the integrative reconciliation of subjects 1 and 6.

Differences in the nurnber and quality of links are in opposition across the two maps.

.4lthough subject 6 produced more markable links. they were of lower quality than the

extended relationai-2 link of subject 1. It is not obvious which of the two maps is supenor. A

scale that considers the cognitive wonh of the different link types is needed for this

cornparison. It can be said that subject 6 exhibited superior progressive differentiation.

The second goai of this thesis was to develop methods for concept map anaiysis.

Methods 1 and 2 meet the needs of this goal. Method 2 can readily assess the progressive

differentiation associated with a concept across different maps. If the same superordinates

and subordinates are associated with a concept across different maps then Method 1 can be

used to compare integrated reconciiiation.

10.3 Cognitive and metacognitive activity

Table 26

The Co~nitiveActivitv of the Six S u b i e c ~

Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 1 Subject 6

Method 3
6b 5% 6% M%
I

6a 39% 5% 33%
5 39% II%
4c 5%
I

4b -97% 36%
4a 13Y0 18% 3 1% 11%
L.

3 36% 19%

1 9% 5% 10%

Table 27

s Hieh. Medium and Low Level Cwnitive Activity

Levels o f Inference in M a ~ of

Inference Subject I Subject 3 Subject 3 Subject 4 Subject 5 1 Subject 6 1

L
Level High Medium Medium High High 1 Low
4 78% 10% 6% 88%

3 13% 90% 84% 1240

1 9% 1O?h

Table 26 reveds the percentages of different link types that were found in the maps of
 148

the six subjects. It summarizes their cognitive activity. Table 27 collapses the Method 3

rating of Table 26 into the inference levels of link-procedural knowledye. For example. the

stage 5 and stage 6a links of subject 1 have been collapsed into level4. A high percentage of

level four inferences is exhibited by subjects 1. 4 and 5. The predominance of high level

inferences lead to a high level cognitive activity rating for these subjects. Level 3 inferences

predominate in the maps of subjects 2 and 3. They were said to exhibit medium level

cognitive activity. AI1 the links of subject 6 were of level3 quality. Unlike subjects 7 and 3.

subject 6 did not produce level4 inferences. The absence of level4 inferences in the map of

subject 6 lead to the low level copitive activity rating.

In chapter 4 it was reasoned that the higher the cogitive activity of map creation. the

higher the level of the accompanyiny metacognitive activity. The validity of this reasoning is

examined in Table 28. Table 28 presents cognitive activity and metacognitive interview

scores for the six subjects (see appendix E). The mean interview scores of the low. medium

and high cognitive activity groups were 15.0, 18.5 and 37.0 respectively. The general trend is

that the higher the cognitive activity level, the higher the level of the metacognitive strategy.

Furiher analyses with many more subjects are needed to confinn this trend but it is

encouraging. It corroborates suppositions taken from the integation of the theories of Kirby

( 1991) and Jacobs and Paris ( 1987). That is it indicates that the cogitive and metacopnitive

processes of concept mapping have comparable quality. Agreement between the suppositions

of theory integration and the results of methods 1 and 3 is taken to mean that concept maps

have a psychological vdidity that is comparable to that of the theories of Kirby ( 199 1 ) and

Jacobs and Paris ( 1987).

Table 28

itive Interview Scores

Subject Cognitive Activity InteMew Mean Interview Score by Group

I

1 High 30 22
4 High 33 22
5 High 23 22
7
II Medium 18 18.5

3 Medium 19 18.5

6 Low 15 15

In closing this chapter it is noted that grade 9 students failed to produce

multistructural (stage 3) and eaended abstract (stage 7 and stage 8) links. Severai lines of

reasoning are submitted to explain this finding. First. it is possible that the sample size of the

study was too small to ensure the production of al1 link types. If more maps were produced

the potential for producing more link Npes would increase as weIi. Second, it might be that

the mapping exercise was not conducive to the production of extended abstract links. More

thou@t may be needed to provide mapping exercises that encourage creative thinking. Third

it is possible that humans are loathe to express the multistructural stage of comprehension.

Multistructurai links would be created by someone who has mastered the successive

processing but not the sirnultaneous processing stage of comprehension. That is. sorneone

creating a multistmcturd link can hold concepts in working memory as discrete units but is

not be capable of integrating them. Complete comprehension requires the identification of

 150

relationships by sequential processing. In essence then, the production of a multistnictural

link represents an admission of incomplete or failed comprehension. In response to question

10 of the metacopitive interviews. students suggested that they would omit or gloss over

links that they did not understand.

 Chapter I l

Conclusions

11.0 Conclusions drawn from the data analysis

Conclusions drawn fkom the data analysis are limited by the fact that only six subjects

were involved in the analysis. The low level interrater reliability gamered for Methods 1 and

3 reduces the reliability of conclusions. It is important to note that an anaiysis involving a

significant sample of the student population may yield results that differ considerably from

those presented here. At best interpretations derived in this study may point the way for

future study into the cognitive alsorithms that senerate concept maps. At worst the

interpretations may simply be misieadiny and erroneous. it is with these caveats in mind that

the following conclusions are drawn up.

Methods 1 and 3 were developed under the assurnption that items of data would

reveal the simultaneous processing that led to the creation of a linking statement. Most links

fàiled to meet this expectation. They listed data but failed to explain its interconnections and

its association to the linkins statement. These connections contain the nature of link-

procedural knowledge. Inadequately aructured links were rated by assurning that the mapper

comected data in the rnanner described in the target text. A few links, such as links 13 and 20

of subjea 2. met the structural expectations of Methods 1 and 3 . Their data were

interco~ectedand they explicated the thinking that lead to link creation. It was suggested

that a more extensive training program would increase the incidence of appropnately

structured links.

ïhree modes of leaming are recognized by methods 1 and 3, structuring, accretion

 152

and tuning. Structuring links connect concepts d e r the logic presented in the target text.

Accretion links connect concepts rnentioned in the target text to related concepts that the text

fails to mention. That is, they associate extant knowledge with the taryet text. The

production of accretion links does not involve creative thinking. Creative thinking is

associated with the tuning mode. Tuning links improve upon the text by adding extant

knowledye to its descriptions of relationships. On occasion a link will connect two concepts

that the text fails to mention. if it contains creative thinking, it is a tuning link. lf creative

thinking is absent it is an accretion link. The different leaming modes generate links of

different quality. The tuning mode generates links of the highest quality. Medium quality links

are produced by the accretion mode and the lowest quality links are ~eneratedby the

structunng mode.

In reyard to Method 1. the tuning mode is associated with extended abstract links.

Extended relationai. extended multistructurai and extended unistructural links are produced

by leamers operating in the accretion mode. The structunng mode produces relational.

multistructural and unistmctural links. In regard to Method 3, the tuning mode generates

links of stage 7 and stage 8. Stage 5 and stage 6 links are senerated by the accretion mode.

Stages 2,3 and 4 are associated with the stmcturing mode. Al1 three leaming modes may

generate inaccurate links. Links are considered inaccurate if their linking statement is

incorrect or Zone or more items of data are inappropnate. Regardless of the leaming mode,

inaccurate links receive prestrucniral, stage 1 ratings.

Predictions derived by integrating theories of cognition and concept map theory were

supported by the analyses of methods 1 and 3. Structural and link-procedurd knowledge

 153

were found to have comparable quality in individual links. This finding supports the reasoning

that concept maps represent mental models. More specifically it supports the reasoning that

concept maps exhibit properties that compare to those of mental models as predicted by the

integration of theories of cogition. Predictions about the properties of mental models were

further supported by the analysis of data from Method 3 and the interview. Students' scores

in the two techniques were positively correlated. This finding supports the aryurnent that the

cognitive and metacognitive processes that govem reading comprehension are also active

durin$ concept map creation. It also implies that concept maps emulate the properties of

mental models as predicted From the integration of theories of reading cornprehension.

Method 1 and Method 3 exhibit a positive correlation because they are sensitive to

the same factors. They rate link quality against three aspects of link data: extent of synthesis

in the linkiny statement. novelty of the thinking and the number of data items. Aithouçh both

methods assess knowledge. they do so from different perspectives. Thus they are not

redundant. Method 1 assesses the quality of understanding expressed in linking statements.

Method 3 assesses the quality of inferences that generate a linking statement.

There are a number of problems associated with methods 1 and 3 . The most

important of these was introduced at the begming of this chapter. When audents fail to

provide proper descriptions of their link-procedural knowledge, the rater is left to reason as

to how a link was formed. In this situation it is impossible to gauge how closely the rater's

deductions match the thinking of the mapper. The problem is best resolved by providing

training that reduces the incidence of inadequate link descriptions by audents. Training could

also be appiied to reduce the incidence of other linking problems such as overly fiagrnented
data and references to the inferences of ot her links.

General knowledge links were also a problem for methods 1 and 3. In this thesis the

author identified these links and excluded them from the rating process. The identification

process was subjective. The subjectivity of this identification process will likely hinder the

general acceptance of methods 1 and 3. An altemate approach would be to restrict mapping

exercises so that students simply reproduce the text. Althouyh this approach will eliminate

general knowledge links, it is also likely that it will stifle creativity. A better tactic rnay be to

accept general knowledge links as a necessary (if inane) component of concept maps. They

would then be rated in the same manner as other links.

In this thesis. links were rated at the prestnictural level if they contained just one

incorrect item of data. When a student's thinking is incorrect. the teacher is required to

provide feedback that informs students why they are incorrect. The identification of

preamcturd links can assist in the feedback process as it pinpoints the source of students'

rnisconceptions.

Method 2 enurnerates links. It was used aiongside Method I to assess the progressive

differentiation and integrative reconciliation associated with individuai concepts. If a key

concept is associated with the sarne superordinates and subordinates in different maps, it

exhibits the same progressive differentiation. In this situation, ody Method 1 ratings need be

considered when cornparhg understanding across maps. Maps exhibitiny higher Method 1

ratinss c m also be said to exhibit superior understanding.

Comparîsons of understanding becorne more complicated when superordinates and

subordinates dBer across maps. In this situation, both progressive differentiation and
 155

integrative reconciliation have to be considered. If one map contains more and better quality

links than another it can be said that it exhibits superior progressive differentiation. superior

integrative reconciliation and superior understandiny. lf link number and link quality are in

opposition across maps, simple comparison is no longer vaiid. In cases such as this. the

cognitive worth of one link type has to be compared to the copitive wonh of another. For

example, suppose maps A and 6 (not show) contain five unistructural links and two

relational-2 links respectively. It can be said that the progressive differentiation of map A is

superior by three links. Quantitative cornparisons of progressive differentiation can always be

conducted by link enurneration. In contrast, quantitative comparison of integative

reconciliation requires a system that compares different link types in terms of their cognitive

wonh. In the case of maps A and B. unistructural links have to be expressed in t e m s of

relational-2 links.

I l . 1 Implications for pedrgogy and student learning

The students' answers to the metacognitive i n t e ~ e wquestions have import for

pedagogy and student learning. Implications for pedagogy are considered first. Five of the six

inte~eweesrealized that concepts become more specific fiom the top to the bottom of the

map. Most inte~eweesthought that the concepts at the top of the map were the most

important and they said that they would select these concepts to summarize a topic.

interestingly, two students suggested that concepts in the rnidde of the map are the most

important. Two more subjects said that the rniddle concepts should be used to summarize a

topic. It is felt that these students answered the questions incorrectly because they interpreted
 156

thern differently than their peers. They recognized that the Iinks from middle concepts to

superordinates explicate understanding of generalized concepts. That is. they recognized

that an understanding of middle concepts is required to surnmarize a topic. They also

recognized the importance of links From middle to low level concepts. These links detennine

the quantity of understanding as they are superior in number to the links to superordinates.

They are also the links that explicate the specifics of a topic. Thus the students ernphasized

the importance of middle concepts as they are the seat of understanding Th- affect the

abiiity to summarize a topic and understand its specifics.

There are additional pedasogcal implications to be derived fiom the students'

answers to the metacognitive i n t e ~ e wquestions. For example. audents requested and

should be given opportunities to discuss their understanding with others. The importance of

discussion is well known. Verbalizing provides opportunities to consolidate conceptions and

to compare them to the understanding of others. Question six of the interview revealed that

students failed to make plans before aarting their maps. Discussion could be used as an

alternative to planning.

.A second pedagogical implication is that students should be asked to create two

concept maps. The first map would be created in a free-form fashion that contains the

student's complete understandins of an instruction. The second map would be a polished

attempt at summarizing. The fim map might serve as a set of plans that a student cm use to

guide creation of the second. Reflecting on the first map provides an opportunity for snidents

to exercise their metacognitive skills. In a conventionai classroom, the alternative to this

double mapping approach might be to get audents to read a text twice and take notes during
 157

the second reading. The mapping approach is preferable as it forces students to connect

knowledye. Connected knowledge is preferable as it has geater potential for recall.

Two implications can be denved fiom the interviews to benefit student learning. The

first concems the issue of audents askiny for clarification when they corne upon a link that

they do not understand. Students in this study said that they would gloss over or omit the

links that they did not understand Fully. The matter is important as the interviewees are

confident. accomplished students. Less able students, it is felt. would have less confidence

and feel even less inclined to ask the teacher for clarification. It is likely that this matter will

best be resolved in classrooms where the teacher takes time to develop an environment that

encourases enquiry.

The second implication for student leaming is that the production of two maps can

increase cognitive flexibility. Dunng production of the first map students should be

encouraged to look for dl anyles and potential links. They should also be provided with an

opportunity to discuss this initial map. By comecting the concepts of maps across many

angles?students link knowledge across multiple perspectives. Students who can apply

knowledge in multiple perspectives are said to exhibit cognitive flexibili~.

11.2 Future research

Ruiz-Primo and Shavelson ( 1996) make note of the wide variety of concept rnapping

techniques. They express concem that despite dserences in task expectations and assessment

protocols, each technique is thought to be an expression of cognitive structure. To identifi

the most appropriate technique they c d for research that makes a comection between "how
 158

we assess knowledge structure" and "our conception of how knowledge is organized."(Ruiz-

Primo and Shavelson 1996, p. 585). They feel that such a connection will allow for

assessments whose empirical evidence evaluates the cognitive theones of knowledge

organization.

Methods I and 3 represent a preliminary attempt to connect assessrnent to theories of

knowledge organization. Method 3 explores the thinking that links are yenerated by the

actions of iink-procedural knowledge. It also assumes that link-procedural knowledge

determines how knowledge is structured. Method 1 assesses knowledge stmcture and

assumes that its quality is a refiection of the link-procedural knowledye that generated it. The

latter assurnption was denved from the integration of the theories of cognition. It found

support in the determination that ratinys For Method 1 and Method 3 were comparable across

the concept maps of the six subjects. Thus. Methods 1 and 3 do not simply provide evidence

to suppon the comparable development of structural and link-procedural knowledg. They

also support the aructural assumptions of the theones of cognition. The preliminary findings

of this study require funher exploration in hture studies. These studies would do well to

involve a significant number of participants. They would aiso benefit from training programs

designed to generate links whose data elicits the actions of link-procedural knowledge. These

training programs would concentrate on ways in which data can be worded so that the

connections between items of dara and the linking statement are rendered eqlicit.

The problem of inadequately structured links is a serious one. Liu ( 1994) touches

upon the subject noting that link correctness is a variable that depends upon assumptions

made by the rater. The assumptions Liu was refemng to are the assumptions made while
marking maps with techniques that enumerate links. Liu ( i 994) notes the popularity of

enurneration techniques and States that the following workers have used them to score maps:

Cleare ( 1983), Novak and Gowin ( 1984). Schreiber and Abegg ( 1991 ), Vargas and Alvarez

( 1992). and Wallace and Mintzes ( 1990). In enumerating a link, raters assume that the stages

of its production were completed correctly. This is a considerable assumptive leap for correct

linking statements can still be produced when one or more items are incorrect. By addins

data to concept map production. methods 1 and 3 eliminate the assumptive leap of

enurnerative methods. When data are available, the accuracy of data and their connections are

readily observed.

Future research should also seek to improve Methods 1 and 3. Both methods are

limited by the fact that they use ordinal assessrnent scales. Links of the same type are

indistinguishable. Differences in quality between link types are thought to be uniform across

the range of links that the methods recognize. For example. the difference in quality From the

unistructural to the multistnictural tevel is assumed to be the same as the difference between

the relational and extended abaract levels. In actuaiity. links with the sarne Method 1 or

Method 3 ratinys differ according to the concepts involved. Improvements to Methods 1 and

3 sbould seek to ascertain how different concepts affect link quality.

This work has made some prorgess toward clarifjmg the terminology of cognitive

theones. Integrating theories of cognition exposed redundancy in rems such as concept

knowledge and declarative knowledge. Integration also lead to the suggestion that

knowledge has two moddities, structural and procedural. These modalities were said to CO-

evolve and be comparable in quality. It is thought that the ciaim for bimodai knowledge and
the constmct of link-procedural knowledge are unique to this thesis. Although the

interpretations of the six students' maps lend weight to the idea of bimodal knowledge that

has a link-procedural subcomponent. funher research involving more students is needed to

bolster this claim.

The impetus for this theory was the intellectual challenge of dernonstratins that

concept maps represent mental models. Althoueh this work provides indirect evidence that

concept maps represent mental models, it is unlikely to impact on the use of concept maps in

schools. The restrictions on greater use are two-fold. First. the three methods and the

interview reguire extensive time commitrnents on the pan of the rater. Second, considerable

skill is required to rate links accurately. An extensive training period would be needed to

impan the requisite skills to teachers. Similarly, extensive training would be needed to

provide students with the requisite link structurhg skills.

The time commitments associated with methods 1.7 and 3 are unlikely to be

overcome. A workable alternative requires a method that allows for the rapid generation of

links and has a rapid marking synem. Novak and Gowin's ( 1984)method fits the bill. Liu

(1 994) compared scores generated by Novak and Gowin's technique and item response

theory and found them to be comparable-The result is encouraging as Liu ( 1994) proposed

that item response theory provides valid assessments of student ability. Future study that

compares Novak and Gowin scores to a method that combines Method 1 and Method 2

might be particularly productive. Determining that these methods exhibit a positive

correlation could be seen as fùrther evidence for the contention that concept maps reflect

mental models.
 16 I

Bibliognphy

Anderson. J. R. ( 1983). The architrctwe rfcogrritiot~New York: Harvard University Press.

Ashcrafl, M. H. ( 1989).H I ~ Urnrrnory

N ard con>rtiotr Glenview. IL: Scott. Foresman and

Company.

Barenholtr H. & Tamir. P. ( 1992). .4 comprehensive use of concept mapping in design

instruction and assessment. Reseorch i ~ Scierm

r c c r d ~'t)ch~ro/og~c'd
Ediicutio~~,
JQ

11),37-52.

Biggs, J. B. & Collis. K. F ( 1 982). Evcrliia~urgthe pinIiy uflrar,ri~ig:The SOLO tmorromy

of the obsrrved Iearr1it1gmfcome)-New Y ork : Academic Press

Is»~(ctrire

Incorporated.

Brock. T. D. and Madigan. M. T. (1988). The Riokogv r , f ~ ~ i c r r t ~N.J.

t . sPrentice
s Hall.

Chi. M. T. H. & Glaser. R. ( 1985). Problem solving ability In Sternberg, R.J. (Ed.).

Hl~rnrurahilitirs ru1 ry5rrmalio11processwg ~~pprmch

(pp 227-250). New York:

Freeman and Company.

Collins, A. M. & Loftus. E. F. ( 1975 j. A spreadins activation theory of semantic processing.

Revirw.
P~ychc~lo~c*ai a ( 6 ) . 407-128.
Collins. A. M.& Quillian ( 1972). How to make a l a n g a ~ user.
e In E. Tulving, & W.

Donaldson (Eds.), Orgw~itu~io~i

of >lmernory(pp. 383-35 1 ). New York: Academic Press.

Edmondson. K. M. (1 993). Concepts for the development of medical cunicula. (Repon No.

TM 020 1 19) Atlanta. GA: Paper Presented at the Annual Meeting of the Arnencan

Educational Research Association. (ERIC Document Reproduction Service No. Ed

360 322).
 162

Edmondson. K. & Smith, D. F. ( 1996). ( Ù I I c c mappit~g

'~~ tofacilitate veleri~zavstt~dem'

ofJl'crid am4 riecrro@t! disorJt.rs. (Report No. TM 025 698). New

rrr~ders~a~tdi~~g

York, NY: Paper Presented at the Annual Meeting of the Amencan Educational

Research Association. (ERIC Document Reproduction SeMce No. ED 400 3 15).

Fisher. K. M. ( 1996). Supporting knowledge construction and reflection. Ilwk 9 (2L 1-34.

Gordon. S . E . & Gill, R. T. (1989). ;Chufr)rrnaiio~~

ofthe r~sec~krmvie.JKr
itlprohirm

I ~ I . ~No. AFOSR-88-0063 ). Washington. DC: Boiling Air Force

soivi~zgJ O ~ C I(Repon

Base.

Goldsmith. T. E.. Johnson. P. I. & Acton, W. H. ( 199 1 ). Assessiny structural knowledge.

. J O I I ~ I Ir.Qf ~Eclt~criiiu~zai
Psychuiogy. 83 ( 1 ). 88-96.

Home, A. J. & Goldman. C. R. ( 1994). L i r n m i o ~(2nd ed). New York, NY:McGraw-

Hill.

Jacobs. J. E. & Paris. S. G. (1987).Children's metacognition about reading issues in

definition. measurement and instniction.~Fd~~catio~~c~i

P.~ychdugisi,22 13 &41255-

178.

Jonassen, D. H., Beissner, K. & Yacci. M. (1993).-Stnic1tiraIkr~owIeJge:Tech~~iy~rrs

for

trpreseriir~igam./ acpiri~îgsiruc~~irui
k~owirJRe.NJ: Ehrlbaum.

Kirby, J. R. ( 1991). Reading to leam: Toward an applied psychology of reading

cornprehension. In I. B. Biggs (Ed.), Trachtzgfor kamiizg: ThL. virwfrom cognitive

P~ycbhoiogy.(pp 103- 125) Hawthom, Australia: Aunralian Council for Educational

Research
 163

Koneman. P. A. & lonassen, D. H. ( 1994).H~wrtextirrrerfac desigtr utrd stnrcrz~ral

kr~owicdgt!ac~piisirior~
(report N0.R 016 737). Nashville. ïN: National Convention

of the Association for Educational Communications and Technolog (ERIC

Document reproduction Services No. ED 373 727).

Liu. X.( 1994). 7ne validity ami reliabilip rfcomepi mappi~gas un ulter,wrivr scietrcr

(Report No. TM 02 1 588).

tz.s.sc.s.smer,>rwhrn item re.spot>.serheory i s ~ s t d f i scuri~rg
~r

New Orleans, LA: Paper Presented at the Annual Meeting of the Amencan

Educational Research Association (ERICDocument Reproduction Service No. ED

370 992).

Mandler. J. ( 1 983). Stones: The fiinciion of structure. Paper presented at American

Psycholo~cdAssociation annual convention.

Marshall. S. P. ( 1 995).3i.hemu.s hl Problem Solvi~~g.

New York: Cambridze University

Press.

Minsky. M. ( 1975). A Framework for representing knowledge. In P. Winston (Ed.), The

pychoiogy ofcornputer visiurr (pp. 3 1 1-28 1 ). New York: McGraw-Hill.

Morine-Dershimer. G.( 1993). Tracing conceptual change in p r e s e ~ c eteachers. Troching

ami Teacbher E&rcafiir,r 9 9 1 1). 1 5-26.

Myers. M. II & Pans. S. G.( 1 978). Children's metacogitive knowledge about reading.

Journal: of Ecluc*a~iot~ai
Psychoiogy. 70 (5). 680-690.

Novak. J. D.& Gowiq D. B.( 1984). Lramit~ghuw to ieczr~~.

New York: Cambridge

University Press.

Paris. S. G., Cross D.H.& Lipson, M. Y. (1984). informed strategies for leaniing: A
 164

program to improve children's reading awareness and comprehension. .Jocirm/ of

Wi~catio~ial
P.~yc~hoIogy.
76 (61- 123 9-12 5 2.

Paris, S. G. & lacobs. J. E. (in press). The benefits of infomed instruction for children's

readins awareness and cornprehension skiils. Child Development .

QuiIlian. M. R. ( 1966). Semantic information processiny. Unpublished doctoral dissertation.

Carnegie Institute of Technolosy, (Reprinted in part in M. Minsky (Ed.), Srmn~itic

N,fumafiot~pr~~cc!.s.~i~ig
Cambridge MA: M . 1. T. Press).

Robenson. W. C. ( 1990). Detection of cognitive st nicture wit h protocol data: Predicting

performance on physics transfer problerns. (hg!Ntivr Sckrice. 14, 253-280.

Ruiz-Primo. M. A. 7 Shavelson. R. J. (1996). Problems and issues in the use of concept maps

in science assessrnent. Juwrioi of Rrsearch itr Scierice Trctchir~g.33 16). 569-600.

Rumelhart. D. E. & Norman, D. ( 1978). Accretion, tuning and restnictunng: Three modes of

learning. In J. W. Cotton & R. Klatzky (Eds.). Sematiricfmrors /ri c40~iiriott.

Hillsdale, NJ: Erlbaum.

Rumelhart. D.E. & Ortony, A. ( 1977). The representation of knowledge in memory. In R.

C. Anderson. R. J. Spiro. & W. E. Montague (Eds.), SchooiirgariJ fhr ucq~isitio~i

of

ktiuwirJgr (pp. 99- 13 5). Hillsdale, NJ: Erlbaum.

Schank, R. C.(1975). the structure of episodes in memory. in D. Bobrow and A, Collins

(Eds.)?Rrprrstittntion m d iiticlerstatirliing (pp. 23 7-272). New York: Academic

Press.

Shavekon, R. I. (1972). Some aspects of the correspondence between content strucniml and

cognitive structure in p hysics instruction. Juwnai of UucafiotiuiPsyhlogy. 63 (3 1.

 325-234.

Spiro, R. J.. Feltovich. P. J., Jacobson, M. J., Br Coulson. R. L. ( 199 1 a). Random access

instruction for advanced knowledge acquisition in ill-structured domains. Educ*urtic~~a/

3 1 (51, 24-3 3
Tec~ht~oiugy

Spiro, R. J., Feltovich. P. J.. Jacobson. M. J.. & Coulson, R. L. ( 1991b). Knowledye

representation content specification, and the development of ski11 in situation-specific

knowledge assembly: Some constructivist issues as they relate to cosnitive flexibility

t heory F:J,rcatiomd Trchriohgy L (91,24-3 3

Thro. M. P. ( 1975). Idivicl~~ai

Jfferet~cesarnorigsl cdiege srrrclewi 111 cognitive srncctlrrr

a,rJphysicv yer_/i)rmc~~~ce.
Paper presented at Amencan Educational Research

Association annual meeting.

Winogad. T. ( 1977). A frarnework for understanding discourse. In M. A. Just & P. A.

Carpenter (Eds.), Cognitive processes in comprehension (pp. 63-88). Hillsdale. NJ:

Erlbaum.

White. R. t Gunnone, R. ( 1992). Probit~gZ Jm/ersra~zdi,>R.London Fdmer Press.

Yacci, M. ( 1990).Asscisi~igK~towlrJXrStnicrzrres (report No. IR 0 14 580). Proceedings of

Selecred Paper Presentations at the Convention of the Association for Educational

Communications and Technolog. (ERIC Document Reproduction Services No. ED

323 957)
 Appendix A

The Target Text

Oil Spills

-
Everyone knows about oil slicks the huge media attention surrounding accidental

spills wch as the Exxon Valdez and Braer, and the huge spills caused by Saddarn Hussein in

the Gulf War has made "oil slick" a househoid word. Although measures have been taken to

prevent spillage from oil tankers, there will still be accidents as long as the world uses oil, and

there is always the possibility of oil being spilled in war or by terronst activity. Oil also sets

into the sea fiom rnany other sources. There are already techniques for clearing up oil slicks,

these al1 have problems. There is no simple solution to the problem of oil slicks.

Exxon Valdez

On the 24th of March 1989' the Exxon Valdez oil tanker ran aground in the Prince

William Sound off Alaska spilling 232,000 barrels of oil. The effects of the incident are not

really known, but it is known that only 25% of the migratory salmon population retumed to

the area the following season, thousands of otters were poisoned, and thousands of birds

died. The scaie of the incident prompted the development of the Valdez Pnnciples, which are

slowly being adopted by industry. It also highlighted jus how expensive a major oil spill can

be - Exxon spent some 3 billion dollars cleaning up the spill, and a fùrther 1 billion to settle

civil and criminai charges related to the case.

 167

The Braer Oil Spül

On the 5th of January 1993, the oil tanker Braer ran aground off Sumburgh Head in

Shetland. At 5.30 in the moming, the tankers engines failed, leaving the ship drifiing in

stormy seas. Emergency rescue seMces were called out, and by 9.00 the last of the 34 crew

had been airlifked o f t h e ship. At 10.55, two tugs began trying to pull the tanker away from

the rocky coastline, but twenty minutes later the Braer with it's 6 19,300 barrels of oil (about

130,000 tonnes) ran aground and began leaking oil.

The Shetland Islands Council already had plans to deal with such an evenr, but

because of the harsh weather none of the plans could be carried out. Six aircraft were sent

out to spray chernical dispersants on the slick, but were dnven back by the high winds. The

action plan also involved using booms and skimmers, but the waves prevented their use.

However. the very weather which prevented the clean-up operation actually helped it - the oil

the Braer was canying was a Light crude oil which is easily dispersed by large waves, and

70% of the oil either evaporated or dispersed naturdy.

The GuIf War

One of the greatest ecological disasters in history happened when Saddam Hussein

used oil as a weapon in the Gulf War. The environmental attack came in two parts: the

release of oil into the Gulf, and the burning of Kuwait's oil wells.

The Water

This prong of the attack was the release of a huge arnount ofoil into the Gulf-
 168

estimates range from 0.5 million to 1 1 million barrels, and average at about 6 million barrels.

Compare this with the tremendous consequences of the Exxon Valdez oil spill, which was

"only" 230,000 barrels. The oil formed a slick 30 miles long and 8 miles wide. which is about

the same area as the Isle of Wight. The slick is thought to have killed between 1 5,000 and

30,000 birds.

The problem was not just the vast amount of oil in the Gulf, but also the nature of the

Gulf itself This inland sea is at most only 35x11 deep. and only has a narrow comection to the

Indian Ocean through the Straits of Honuz, which means that it takes between three and

five and a half years

for al1 the water in the Gulf to be changed, so there was no hope of the oil simply dispersing,

as it did in the Braer spill.

The Air

Although the buming of Kuwait's oil wells may have been intended as an economic

weapon, the environmental consequences were no less severe than those of the oil slick. It is

estimated that about 67 million tonnes of oil were bmed in total, which produced about 2.1

million tonnes

of soot and 2 million tonnes of sulphur, one of the main causes of acid min. Fortunately, the

soot particles did not go any higher than 5,000 metres, which meam that they were washed

down with the min over the next two weeks. If the soot had got above the rainclouds, it

would have remained up there, with severe consequences to the world's cbate. As it was,

the consequences for Kuwait were still harsh. The soot cloud made the daytime almost as
 169

dark as the night, and the World Health Organisation estimated that death rates in Kuwait

rose by 10% over the following year because of associated breathing difficulties and skin

problerns.

Prevention of Oil Slicks

There are two main methods used to prevent oil spilling fiom tankers. The first is to

stop them having accidents, and the second is to stop oil leaking if they do have an accident.

Avoiding Accidents

Maritime authorities recommend keeping at leaa 10 miles away from any Coast

(unless they are docking of course!). Tankers also plot their routes according to the weather

forecast. but the weather cm only be predicîed with any accuracy for the next three days,

which is of limited use to a tanker which can take a whole day to stop. Oil tankers and other

sea vessels avoid collisions by filing route plans with maritime authorities who keep the

routes well separated in the same way that air t r a c cornrollers keep aeroplanes well apart in

the sky.

Preventing SpiiIs

The Oil Pollution Act was passed in the USA in 1990 in response to the E n o n

Valdez oil spill, and means that al1 new oil takers must be built with a double hull so that if

they do hit anything the outer hull cm be pierced without causing a leak. Another meanire

used to prevent oil spilling ifthe oil tanks are pierced is to fit the tanker with hydrostatic
 1 70

controls to make the pressure inside the tank the same as outside, so no oil will come out of

the tank. However, both these measures are very expensive to implement, and as they are

only voluntaiy (as yet), they have

not been widely taken up by the shipping industry.

Other Sources o f Oil

The oceans are impacted by sources of oil other than oil slicks. These sources are:

Tanker Accidents 12%

Atmosphere 9%

Industrial discharges and urban run-off 37%

Natural Sources 7%

Exploration and Production 2%

Vesse1 Operations 33%

HI
IUI
-.-
.--
---
.--
---
---
--

Oil Slicks- Cleanup Methods

There are currently only three ways to ciean up oil slicks:

Naturd Dispersion

Collection

Chemical dispersion

Naturai Dispersion

Oil slicks cm simply be lefi to disperse and break d o m nahidly, which can be
 17 1

helped by certain weather conditions, such as in the case of the Braer. Most of the oil is

dispersed this way even when other clean-up methods are used. It is estimated that in the

Exxon Valdez spill, no more than 8 percent of the oil spilled was recovered, despite the huge

sums spent on the clean-up the rest was dispersed by the fierce Arctic storms of the region.

Collection

The best way to clean up an oil slick is to actudly collect the oil. This means that not

only is the oil completely removed from the environment (unlike chernical and natural

dispersion techniques), but it can also be used again.

The first stage in collecting the spilled oil is to contain it. This is usually done using

booms, which corne in two types. The first boom is inflatable, with a chain almg the bottom

to keep it the right way up. The second boom uses solid floats, like those you used when you

leamed to swim.Inflatable booms are easier to store, as they take up much less space, but are

not as resistant to damage. Where such specialist equipment is unavailable, booms can be

improvised fiom locaily available materials, such as fences and bafes of hay:

Once the oil has been contained by a boom it cm be collected using skimmers or

sorbants. Skimers can be broken down into two types: suction and adhesion. Suction

skimmers operate Sie giant hoovers, sucking the oi1 off the water. However, these suck up

lots of water as well - oflen more than 90 percent of the liquid collected is water. Adhesion

skimmers sweep a material

through the spiIl. The oil sticks to the matenal. and is carrieci away from the spiu. The oil is

then squeezed out of the material

 1 72

Sorbents are materiais which absorb oil. These can be either natural, such as straw, or

synthetic, such as polypropylene. They can be used in a variety of shapes, such as sheets and

mops. They are not really suitable for collecting large volumes of oil, so tend to be used in

the final stages of cleaning when only small pockets of oil are left.

There are ships which integrate these methods to make a complete oil collection

systern, such as the Al Waasit used to clean up afier the Gulf War. This ship lays inflatable

booms around a section of the slick, and then sucks the oil off the water using vacuum

pumps. The mixture of oil and water sucked in is then separated, and the water is then

pumped back out into the sea, while the oii is kept in storage tanks on the ship, from where it

cm be off-loaded to a tanker or a refinery. The Al Waasit can clean up to half a square mile

per day.

Chernical Dispersion

One of the main problems with oil slicks is the way they stick together and float on

the surface of the water, so chemicais have been developed which cm be sprayed on the oil

to break the slick apart.

Once it has been broken up, the slick dilutes into the water without causing any major

problerns. However, because of the way these dispersants work they cannot be used on

viscous oils, which includes some crude oils and dl heavy fuel oils. Also, most crude oils

which can be dispersed change their characteristics within a few hours in cold water, and cm

no longer be dispersed. This means that any dispersant response mua be fast, so

dispersant-spraying aeroplanes and helicopters are usually used.

 173

Scientists working in Amenca have developed a new breed of oil dispersants which

are actually living organisms which feed on oil. These have b e n very successtlll in laboratory

tests, but they have not been tested on real spills because of concems about releasing

genetically engineered organisms into the environment.

Although modem dispersants are much less toxic than earlier types, they are still a

pollutant in themselves, and before they are used it must be decided whether they may cause

more environmental damage than the spill would if it was left untreated. Whether dispersants

should be üsed or not depends on the depth of the water, how strong the currents are, and

what lives in the water there. For example, spills in the middle of the Atlantic Ocean could be

sprayed with chernical dispersants without much worry, but a spill in a salmon f m off

Scotland wouid probably not be sprayed.

 Appendix B

Letters of Consent

Principal
School
Hamilton, Ontario

Dear Principal,
Following Our first conversation 1 would like to take this opponunity to funher
infonn you of the research t am conducting involving science education. 1 am most grateful
and appreciative of your cooperation and enthusiasm in this research endeavour. 1 am a
doctoral student in Curriculum, teaching and leaniing at the lnstitute for Studies in
Education. My research concerns concept maps and their ability to elucidate the processes of
comprehension. The objectives of the study are: to provide evidence that concept maps
reflect the structuring of concepts in memory, to test the utility of three methods of concept
map analysis, and to elucidate the cognitive and metacognitive strategies of comprehension.
A mernber of your staff has agreed O participate in this audy and will be invited to
offer opinions and advice that might further improve the three analytical methods. Data for
the audy will be collected in two ways. Over a period of one week concept maps will be
collected both before and after a text is read. At the end of the week a number of tape
recorded interviews will be collected. If further cornmentaries are needed to clahfi maps or
interviews they will be negotiated on a case-by-case basis with the teacher and mdents.
1 look fonvard to visiting your school and working with a member of your faculty. If
you have any questions please do not hesitate to contact me at (905) 383 0735. Thank you
for your support, and for your interest in concept mapping.

Thank you
Sincerely,

Graham Passrnore
Department of Cumculum Teaching and Learning
OISE
 Principal Consent Form

1, . agree to participate in the study "Concept Maps and the

Processes of Cornprehension" as outlined in the attached letter.

Signature

Date
Dear Participating Teacher,
I would like to take this opportunity to formally invite you to participate with me in a
research project with the Curriculum, teaching and Leanting Department at the Ontario
Institute for Studies in Education from March 9 to March 13, 1998. This research project will
be submitted as partial fulfillment of the requirements of the degree of Doctor of Philosophy
at the University of Toronto. Your agreeing to join in this work will allow us to explore the
cornprehension processes that students engage in as they lem.

The Research
The project requires that students create two concept maps. One pre-instructional
map will be created before reading a target text. One post-instnictional map will be created
after reading the text. The pre- and post- maps will be assessed by three interpretation
methods that have been developed for this project. Six of the nidents will be selected for a
short i n t e ~ e won the basis of the cognitive change they demonstrate. Your participation is
invited so that the content of the target text may be better matched to the interests and
abilities of the students. You are also invited to participate in the interpretation of the maps
created by the six students selected for interview. You experience, feelings and ideas are
needed to furiher improve the utility of the interpretation methods.

Et hical Considerations
A number of neps shall be taken to ensure the anonymity and confidentiality of the
research. Al1 proper names and identifjmg details wiil be changed on the maps and on the
i n t e ~ e wtapes. The maps and tapes will remain confidential (with the exception of my thesis
cornmittee) and will be locked away once the project is completed. if further commentaries by
students are needed to elaborate on the maps or interviews they will be discussed on a case-
by-case basis with the teacher and midents involved.
i will share my findings and conclusions with you and provide interpretive accounts
prepared fom the data as the audy progresses. You are invited to read and respond to these
accounts and suggest corrections and differences of opinion. These responses will be use do
Uiform further drafts for the purpose of the thesis. A final surnmary of the study will be made
available to you.
You are f kto delete any data that you feel compromises your position in any way.
Fially 1 acknowledge that you have the right to withdraw fiom the study at any time without
reason.
If you are willing to participate in this research project, please complete the attached
form. Thank you for your interest and cooperation.

Sincerely,

Graham Passmore
 Teacher Consent Fonn

1, .
agree to participate in the study "ConceptMaps and the
Processes of Comprehension"as outlined in the attached letter.

Date
 Student Consent Form

You are invited to take part in a study conducted with the Curriculum, teaching and Leaniing
department at the Ontario Institute for Studies in Education. The study will run from March 9
to March 13, 1998. By agreeing to join in this work you will allow us to explore how
students lem.

The research
the project requires that students m a t e maps of their knowledge. One map will be
created before a short text is read and one will be created afler the text is rad. Both maps
will be assessed to identiQ thinking and leaminy processes. Some students will be selected
for a shon interview on the basis of the leanllng that they demonstrate.

Ethical Considerations
A number of steps shall be taken to ensure the anonymity and confidentiality of the
research. Ail proper narnes and identifjing details will be removed from the maps and
interviews. The map and interview data will remain confidentid and will be locked away once
the project is completed.
If funher commentaries by students are needed to elaborate on the pas or intewiews
they will be discussed on a case-by-case basis with the tacher, guardians or students
involved. Findinys and conclusions will be made availabie to the student on request.
Finally 1 acknowledge that you have the right to withdraw from the audy at any time
without reason.
If you are willing to participate in this research project, please complete the anached
form. Thank you for your interest and cooperation.

Sincerely,

Graham Passmore
 Student Consent Form

1, . agree
to participate in the
Processes of Comprehension"as outlined in the attached
study "Concept Maps and the
letter.

Signature

Date
 Appendix C

Written Instructions Provided to the Teacher and the Studeats

Marking instructions provided to the raters

Rate each link ayainst the scaies of methods 1 and 3. Once a link has been identified record

its score in the appropriate column of Table 1. In regard to Method 2, enumerate the links

associated with the key concepts and place the Number-Link score in the Method 2 column

of Table 1.

Links are created by inferences. Four levels of inference are recoynized by Methods 1 and 3.

a) incorrect inferences

b) inferences that repeat info in the text

C) inferences that add existing knowledge to information in the text. Example. tea describes

only links from concept A to concepts B and C. An inference that adds to the text would be a

link fiom link from concept A to concept D.

d) inferences that take existing knowledge and incorporate it into the text. Example, text

says that concepts A and B are linked via points X and Y. An inference that incorporates

information into the text would include point Z as well as X and Y in the link from A to B.

Method 1-

Purpose: to identfi the quality of linking statements

Look at the iink and rate it against the following descriptions.

Place the rating in the column marked Method 1

Incorrect inferences

1. If the linking aatement or one or more items of data are incorrect, mark the link as

prestructurd

Correct inferences that repeat information in the text

2. If the linking statement was formed from one item of data it is unistructural

3. If the linking statement was formed fiom two or more items of data it is either

multistructural or reIational.

3a In multistructural links, data is listed in the linking statement

Example: Suppose the rnapper has to respond to the question "why is Iraq's political

situation bad?" by linking the concepts "Iraq" and "Political Situation." A multistructural link

would appear as

Linking statement: It is bad because of the oil and it is also bad because of Saddarn"

Data: a)The oil makes it bad

b)Saddam Hussein makes it bad too

Each item of data is listed and conjoined in the linking statement through phrases such as

"and it is also", "as weiI as" etcetera.

The number of data items shodd be added to the nomenclature. Thus links with two data
items would be multistructural-2, Links with three data items are multistnictural-3.

Whatever the number of data items, add it to the link rating as a suffix. The example link

descnbed above is multistnictural-2. The number of data items should also be included when

rating relational and extended abstract links.

4 Relational links differ from multistructural links in that their linking statements are fonned

by integrating data. Even if the integration process is not made explicit in the link, the rater

should be able to rationalize as to how data integation took place. Relational links compare

to multistnictural links in that al1 their data was denved fiom the text.

An example relational link (responds to the Iraq political question mentioned in point 3a).

Linking Statement: Iraq's troubles stem from Saddam's pan-Arabian ideolog which threatens

the World's access to cheap oil.

Data a) It is bad because of the politics

b) It is bad because of the oil

c) It is bad because of the money

NB. Here the data is integrated into a single phrase. The result is a relational-3 link

Links that take existing knowledge and incorporate it into the t u t .

1 . Extended Abstract links. Like relational links these links htegrate data in the linking
 183

statement. Unlike relational links they include at least one item of data that was obtained corn

outside the instruction.

Example (also considers the Iraq question)

Linking Staternent: Iraq's troubles stem frorn 1923 when the Middle East was divided to suit

the political aspirations of Colonial powers. Recognizing that extemally imposed land division

attenuates Arabian power, Saddam Hussein sought to unite the Middle East. A united Middle

East could undoubtedly change the balance of global politics by controlling the movement
and price of oil. It is the ylobal threat imposed by Saddam's ideoiogy that makes Iraq's

political situation so bad.

Data: a) I t is bad because of the oil

b) It is bad because of the oil

c) It is bad because of the money

d) It is bad because of the land divisions of 1923.

Suppose the multistructural and relational links were created fiom information provided in

current news programs. Current news programs have neglected to report the impact of

Colonial strategies on current events. Thus it can be said that this link associates the concepts

with a datum that was taken tiom outside the instniction (current news programs). The use

of non-insmictional data typifies the extended abstract response.

Links that add existing knowledge to information in the text

1. Extended Multistmcturai links. Same as multistructural links except that they link a

concept mentioned in the text to one that was not mentioned in the text. OR they concem a

Iink that has taken severai points mentioned in the text and used it to form a linking statement

that was not overtly discussed in the text

2. Extended Relational links. Same as relational links except that only one of their concepts

was mentioned ovenly in the text.

Method 2

Purpose: to identify the number of links associated with the key concepts.

1. Enurnerate the links associated with the three key concepts Oif Spilis, Tankers and C'lem

Ilp. Inciude links to superordinate and subordinate concepts. If there are two links the

concept is a 2-Link concept. If there are three links the concept is a 3-Link concept etcetera.

Method 3-

Purpose: to identie the cognitive activity (link-procedurai knowledge) of the mappers

Look at the link and rate it against the following descriptions.

Place the rating in the colurnn marked Method 3

Stage 1. Inappropriate reasoning. An inference is made that f d s to consider information

presented in the text. The result is a link that connects concepts through inappropnate items
of data.

Stage 2. Concepts are linked through a listing of one or more items of data. No inferences

are present to indicate how or why the concepts are connected.

Stage 2a. One item of data is listed to link the concepts. No inferences are made.

Stage 2b. Two items of data are listed to link the concepts. No inferences are made.

Stage 2c. Three . . .

Stage 3. One item of data links the concepts through an inference based on facts (the

inference repiicates descriptions presented in the text).

Stage 4. Two or more items of data are linked through inferences based on facts. The

inferences replicate descriptions of information presented in the text.

Stage 4a. Two items of data are linked through an inference based on facts.

Stage 4b. Three items of data are linked through two inférences based on facts.

Stage 4c. Four . . .

Stage 5. One item of data links the concepts though an inference based on faas. The

inference was not overtly discussed in the tes.

Stage 6. Two or more items of data are linked through inferences based on facts. The

inferences were not overtiy discussed in the text.

 186

Stage 6a. Two items of data are linked through an inference that was not presented in

the text. The inference is not based on inappropriate reasoning. It builds on

information presented in the instruction. Some explanation of the inference may be

presented.

Stage 6b. Three items of data are linked through an inferences. The inference is not

based on inappropriate reasoning. It builds on information presented in the

instruction. Some explanation of the inference may be presented.

Stage6c. Four . . .

Stage 7. An item of data connects the concepts through an inference based on fact. The

inference does not build on information presented in the text. It is an example of creative

t hinking.

Stage 8. Two or more items of data are co~ectedthrough inferences based on facts. Rather

than building on information presented in the text the inference(s) represent creative thinking.

Stage 8a. Two items of data are Iinked through an inference that was not

presented in the text. Rather than building on information presented in the text

the inference is an example of creative thinking

Stage 8b. Three ...

Concept mapping instructions provided to the student

The following instructions were provided to the students before they created the practice

map.

1. Arrange the KEY concepts on the bristol board as you see fit.

2. Write out any additional concepts on post-it notes and add them to the map.

3. Make as many links as possible by drawing arrowed lines between the concepts. Number

the links and write a description of them beneath the map.

4. Remember ALL L M S must have a linking statement--do this as you draw the lines-

NOT LATER

5. When dl the links are described go back and add the pieces of information that you used to

create the linking statement.

6. Each piece of information should concem oniy ONE aspect of the linking statement.

This is often the most difficult aspect of concept mapping. When you cannot think of the

information you used try breaking the linking statement down into its components parts. It

should be quite easy to expand these components to form the pieces of information.

7. Some links will have just one piece of information and others will have several.

8. When just one piece of information is involved remember that it can be the same as the

linking statement or it can be dEerent. If severai pieces of information are involved in a Iink,

they should not match the linking statement exactly.

Table 1

1 LUik Number 1 Method 1 1 Method 2 1 Method 3 1

 188

oir Spilfs

Tankers

ïlean (/p

.
 Appeodix D

Method 1, Method 2 and Method 3 scores for the s i x subjects

Subject 1

L-

Tsikas w-.

m r e D l Subject 1

Table D 1
.
ethod 1 and Methpd 3 Scores for S u e e t I

Link Description Method 1 Method 3

1 Tankers cause oil spills General Generai
a) tankers crash Knowledge Knowledge
b) 03 containers are comprornised
c) oil spills out
2 Oil spills must be cleaned up Extendeci 6a
a) to prevent environmental impact Relational-2
b) to salvage oil
 I
- - -

Safety procedures are needed for tankers to be effective Extended

Unistructural
A double hull is a safety procedure
a) crashes must be harder to puncture both hulls
b) oil containers are less likely to be breached
Oil containers with the same inner pressure as outside
pressure will spi11 less oil
a) less oil spills
b) the disaster is less great
Better navigational radar will allow obstacles to be Extended 5
avoided Unistructural
a) less crashes will occur - - - - - -

More powerfùl engines will make ships safer Extended 6a

a) faster turning will be possible Relational-2
b) obstacles can be better avoided
Tankers cost %
a) hypothetically; more $. better ship
Extended
Unistructurai I
Safety procedures COS $ Extended 6a
a) hypothetically: more $, better procedures Relational-2
b) less $, worse procedures
tankers require workers Extended 6a
a) more workers, less chance of problems Relationai-2
b) less workers, more chance of problems
- .

Workers cost $ Extended 6a

a) more %, better quality of workers Reiational-2
b) less $, lower quality of workers
-- -

Workers apply safety procedures Extended 6a

a) better quality workers = better quality procedures Relational-2
b) lower quality workers = lower quality procedures
-- - - - - - -

A clean up requires workers Extended 6a

a) higher quality workers = higher quality clean up Relational-2
lower quahi workers = bwer qualiity clean up
To clan up, we must follow a procedure Extended 6a
a) good procedure = good clean up Relational-2
b) bad procedure = bad clean up
Workers must apply procedure Extended
a) higher quality workers = higher quaiity application of Relational-2
procedure
b) lower quality worken = lower qudity application of
procedures
-- -- - -- -- - - - -

Natural dispersion is a clean up procedure

a) it is 80% effective
b) sometimes has side effects
Collection is another procedure
a) it is 100% effective
b) it gets al1 oil
C)oii can be reused
Chernical dispersion is a clean up procedure
a) it is sometimes effective
b) it always has side effects
Workers are not needed for natural dispersion Extended
Unistructural
These cost money Extended
Unistructural
This does not cost money Extended
Unistmctural
Because of many above reasons. workers are responsible Unmarkable
for oil spills
- -- - -- - - -

Worken are determined by how much rnoney they keep Extended

Unistructurai
- - - - - - - -

The government determines how rnuch of their wages the Extended

worken keep, so government determines workers Unistructural
Corporations determine how much wages workers get, so Extended
corporations determine workers Unistructural
The governrnent detemines how much of their profit
corporations keep, which the wages corporations pay, so
the governent ultimately detemiines the workers
From # 22-26 above, we can conclude that the
governrnent is ultimately responsible for oil spills
Table D2

Method 2 Scores of Sribject 1

1 ~ e concept
y 1 Nurnber-Link Score 1
Oil Spills 4-Link
Tankers 4-Link
-

Clean Up 3-Link
 Sobject 2

--
m
w
L
se-Tmmi-16
-
, 1 CtlmbI \--j~
- --7
L ----
Dhpcahi -23!l
Colfactioo -
w=-
A

- --
*c - - - 7--
---Ji---
19 - z l ï

W r e D f Subject 2

Table D3

ethod 1 and Method 3 Scores for S&iect 2

Link 1 Link Description

-
Method 1 Method 3
Shipping indu- manufactures tankers General
a) get materials Knowledge
b) put parts together
c ) sell to cornparies
2 Tankers are large boats
a) designed to hold lots of oil
b) bigger ship, hard to maneuver
C) long time to slow down

3 Tankers cause oil spills General

a) tanker crashes Knowledge
b)oil leaks out
Oil supply provided by companies General
a) buy tankers Knowledge
b) fil1 them with oil
c) send to other companies
Wars cause oil spills (Gulf War)
a) oil used as a weapon
b) oil releesed into Gulf
C) burn oil wells

Oil spills cause pollution Unistnictural

a) spread through environment
Pollution kills plants and animais Relationai-2
a) Animal's habitat is destroyed
b) Animal has nowhere to suMve
Pollution causes death and health problems
a) air pollution leads to breathing problems
b)water becomes poison
Oil spills need to be prevented and avoided
a) preventing oil spills
b) stop them fiom having accidents
C) if accident occurs, stop them from leakîng
d)avoiding spills
e) plotting routes to avoid accidents
Maritime authorities help with avoidance
a) recornrnend keeping at lest 10 miles away from Coast
b) tankers file route plans with maritime authorities
c) keep routes separate
- --- -- -- --- - - - -

Laws help with prevention

a) oil poilution act
b)new tankers built with double hull
c) hydrostatic pressure controls
More media will help with prevention Extended
a) media attention can convince shipping industry to use Relational-2
prevention methods and technology
b) increase public awareness
Oil spi& need to be cleaned up safely 1 Extended 1 6b
a) find the safest met hod
b) use what is available
c) start clean up
One technique of clean up is natural dispersion
a) to clean up the oil is dispersed naturaily

Natural dispersion can only be used in certain weather

I Unistructural

Prestructural
I 3

1
conditions
a) oil dispersed by waves and harsh conditions
b) waves break up the oil
Another technique is collection Unistructural 3
a) oil is collected and removed from water and used again
Oil first contained by booms Unistnictural 3
a) surround oil, stops From spreading
oil is collected using skimers Unistructural 3
a) two method ofskimming to coliect oil
One method of skimming is suction Relational-2 4a
a) suck oil off water
b) also suck up lots of water: 90 %
Other method of skimming is adhesion
a) sweep matenal through spi11
b) oil sticks to material
C) oii carried away from spiii
d) oil squeezed out of material 1 1
Sorbents used to absorb oil collected by booms
a) can be natural or synthetic
variety of shapes- sheets mops etc.
C) used for smalf pockets usually

Last method of clean up is chernical dispersion 1 Uni~chiral1 3

a) chemicals used to disperse oil 1 1
Rate of dispersion depends on water temp Unistructurai 3
a) cold water- oil changes characteristics and cannot be
dispersed any more
 24 Dispersants break up oil slick Relational-3 4b
a) break up oil slick
b) slick dilutes into water
c) can only be used on certain oils I

25 Airplanes are used to spray dispersants Relational-2 4a

a) response must be fast
b) airpianes spray dispersants on oil
I

26 Scientists are working on ways to develop new techniques Relational-2 4a

of chernical dispersion
a) a new breed of oil dispersants- living organisrns that
feed on oil
b) trying to find less toxic ways of dispersing

Table DS

ethod 2 Scores o f S-ect

. 1

Key Concept Number-Link Score

Oil Spills 5-Link
Tankers 4-Link
F i i u n D3 Subject 3

Table D5

ethod 1 and Method 3 Scores for Subiect 3

Link Link Description 1 Method 1 1 Method 3

1 Tankers create oil spills
a) tanker crashes
b) oil spiils
2 Dunng the Gulf War, Saddarn Hussein put oil in the Gulf Relational-3 4b
and lit Kuwait's oil wells on tire in anempt to win war 1 1
a) oil put in water I I
b) burnt oil weUs I I
When oil spills, people try to clean it up to the best of their Unmarkable
abilities
a) oil spills
b) clean up using bea technique
C) environmental changes occur

The manufacturers make the tankers and are supponed by General

suppliers Knowledge
a) make tankers
b) supplied by supplier
The Exxon Valdez was a tanker that ran aground and
spilled a lot of oil
a)tanker ran aground
b)23200 barrels of oil spilled
c ) environmental changes
Tne Braer ran aground and the engines failed causing an
oil spi11
a) tanker ran aground
b)6 19300 barrels of oil spilled
C)naturally dispersed
The retailer buys the oil From the tankers and sells it to General
consumers Knowledp
a) tanker brings oil
b) oil bought by retailer
c) oil bought by consumer
Durine the Gulf War, oil was released into the Gulf in
attempt to destroy Kuwait's oil supply
a) war occurred
b) oil released
C) oil couidn't disperse

Another attempt in winning the War was buming Kuwait's

oil wells
a) war occurs
b) oil weils bumed
Saddarn Hussein is the leader of Iraq who bumed Kuwait's
oil wells and poured oil in the Gulf
a) became leader
b) started war
C) put oil in Gulf
d) burned oil wells
There are 3 different cleaning methods used to clean Prestructural
spilled oil
a) oil spills
b)clean up methods
c)oil gets clean
There are 2 methods of preventing oil spills: avoidance Relational-2
and ship improvements
a) avoid oil slicks
b) upgrade ships
The govemment organizes and puts clan up into action. Extended
They pay for it through the tax money Relational-3
a) taxes collected
b) clean up organized
c)put clean up into action
Natural dispersion is where the natural conditions disperse Relational-2
the oil
a) oil spills
b) waves disperse water
Collection when the people collect al1 the oil and can reuse Prestrumral
it
a) oil spills
b) collect oil
c) reuse oil
Chemical dispersion is when chernicals are used to Relational-2
disperse oil
a) oil spilis
b) chernicals are used
A double hull protects the inner huk even if the outer hull Relational-2
is penetrated
a) outer hull is damageci
b) ship keeps running
 Avoidance is when the ship avoids possible dangers
a) ship gays away from danger
Hydrostatic controls, oil doesn't leak even if hull is
damaged
a) hull gets damaged
b) oil doesn't spi11
The tankers keep their distance fiom other tankers Unistructural
a) the tankers keep their distance fiom other tankers
Tankers stay at least 1O miles away fiom shore so oil
slicks are avoided except when docking
a) tankers stay 10 miles away from shore

Table D6

ethod 2 Scom of Suhject 3

Key Concept Number-Link Score

Oil Spills 3-Link
Tankers 5-Link
Clean UP 4-Link
Subject 4

m e D4 Subject 4

Table D7

ethod 1 and Method 3 Scores for Subject 4

1 Link 1 Link Description 1 Method 1 Method 3

I

Tanker Company has a lot to do with oil spills Extended 6a

a) tanker CO. have a lot of &dom and they can take very Relationai-t
effective preventive measures if they choose to do so
b) they are held responsible for spills so they must cough
up the money to clean up the spiil
2 Circumstances before accident are key factors to why the Extended 5
oil spiu happened UriistructuraI
a) as in the case of the Braer the aorm was the main cause
. of the accident
 - - - - - - -- - - - -

Before cleaning up, the circurnstances of the situation Extended

mua be considered to help choose the appropriate Relational-2
cleaning met hod
a) if the circumstances are that the oil will disperse by
itself naturally then not much cleaning is required
b) if as in the text, the spi11 occurs at a salmon farm, then
chernical sprays should not be used
Circumstances before spi11 ar important as they affect the Extended
circumstances after spill Relational-2
a) if the crew are poorly trained and steers the tanker into
some rocks near a fishing village, then the clean up ,which
is affected by the circumstance, must be swift and harmless
to environment
b) if like the water, the oil is spilled into the Persian Gulf
then the situation of the spill will be different fiom if, say,
the oil was spilled in the Pacific, and thus the cleaning
methods witl be different
Thought process is vital before choosing a cleaning Extended
method Relational-3
a) the circumstances of the wreck, the money available,
must al1 be considered before choosing the bea cleaning
method
.- - - - .. . .

Thought process is vital before choosing a cleaning Extended

method Relational-3
a) the circumstances of the wreck the money available,
mua al1 be considered before choosing the best cleaning
method
Money is a big factor to tanker companies General
a) the main thing tanker CO. are interested is % or profits. Knowiedge
So if they regard taking preventive measures as a Ioss in
profits, they will not take the meanires
Money is a big factor to tanker companies General
a) the main thing tanker CO.are interested is $ or profits. Knowledge
So if they regard takins preventive measures as a Ioss in
profits, they wiU not take the measures
 Preventive mesure affect the circumstances before the
wreck
a) having a double hull will greatly benefit a tanker and
prevent it from leaking oil upon impact, it might even
prevent a crash
b) if a tanker keeps away fiom "danger zones" then there
is little risk of hitting something
Thought process is vital before choosing a cleaning Extended
met hod Relational-2
a) the circumstances of the wreck, the rnoney available,
mua al1 be considered before choosing the best cleaning
method
Thought process is vital before choosing a cleaning Extended
method Relational-:!
a) the circumstances of the wreck, the money available,
m u a al1 be considered before choosing the bea cleaning
method

Table D8

Method 2 Scom of Sumct 4

1~ e Concept
y 1 Number-Link Score 1
Tankers Concept was not used
Clean Up Concept was not used
Subject 5

Fieure D5 Subject 5

Table 159

Method 1 and Method 3 Scores for Subiect 5

Link Link Description Method 1 Method 3

1 Tankers cause spills Generd -
a) we need to prevent oil spills Knowledge
b) if we don't, Our environment is at risk

I
.- - - - - .-

Oil spills need to be cleaned up Prestructural 1

a) if not our environment is at risk
b) if a lot of fish and animals die, we pay more for what's
lefi
C) we al1 becorne vegetarians
You need money to do stuff 1 Extended
a) tankers always have a lot of money 1 Relational-'
b) al1 of their money should 30 towards cleaning up
The tankers give al1 their money to clean up leaving none Extended
for them Relational-3
a) they don't transport things anymore 1
b) it costs more to transport anything 1
The clean up is s u p e ~ s e dby a govemment 1 Extended
a) they are the people in charge Unistructurd
The govemment makes stricter regdations regarding ship Extended
safety Relational-3
a) hopefùlly, that means there will be no more spills 1
b) the tankers won't go bankrupt (see #4). and al1 of the 1
animais wiII Iive I
Govemments are al1 in debt and don't have enough money Extended
a) would you spent money on hospitals or oil spills Unistructurd
-- - -- - - - -

The government m u a think of a solution. the three Extended

options are Relational-3
a)natildispersion
b)chemical dispersion
C) collection
-- -- - - - - -

Chemicd dispersion doesn't work

a)cannot be used on viscous oils.. change characteristics
quickly .. the texî says in a polite way that this hardly ever
works
b) oil changes in cold water within a few hours. I
challenge you to think of a govemment that could get to
the scene. with papenuork done, chernicals in hand, within
a few hours. That just won't happen
Collection doesn't work Extended
a) if you look back you'll see that the guide doesn't Relational-2
mention cost. This is probably really expensive.
Governments don't have money (see # 7)
Natural dispersion works Extended
a) Our planet repairs itself incredibly well- look at the data Relational-3
for the Exxon spi11
b) it's cheap! !
Table Dl0

Nethod 2 Scores o f Subiect 5

1 ~ e concept
y 1 Nurnber-Link Score 1
1 Tankers 1 3-Link 1
Subject 6

Figure D6 Subject 6

Table D l 1

Method 1 and Method 3 Scores for S u b b t 6

Link 1 Link Description 1 Method 1 1 Method 3

I There are several causes to oil spills
a) there are several causes to oil spills
-
7

3
I Tanker accidents are a cause
a) tanker accidents account for 12%of oil in the ocean

Politics are a cause

Unistructural

Unistructural
3

3
a) politics are a cause A
 Vessel operation is a cause Unistructural
a) vesse1 operation accounts for 33%
Revenge is a politicai tactic Unistructural
a) revenge is a political tactic
Safety is a large concem
a) the Oil Pollution Act
A clean up is required afler an oil spill Unistructural
a) a clean up is required after an oil spill
- - - - - - - - - --- ----

The environment is affected by oil spills Unistructural

a) the environment is afTected by oil spills
Clean up costs $ Unistructural
a) clean up costs !§
Nature is destroyed
a) nature is dearoyed
There were many dead animals
a) there were many dead animals
Activists get mad Unistructural
a) activists get mad

Tabte Dl0

Method 2 Scores of Subiect 6

1 Number-Link Score I
1 Tankers
 Appendix E

The Metacognitive Interviews

This appendix reports the six subjects answers to the interview questions. It also

provides a score for each subject. On occasion the inte~ewerhad to prompt or guide

subjects to elicit an appropnate answer. The interviewer's actions are presented in italics.

Table E l

etacognitive Interview: Subiect 1

i
h s w e r . Making s w ihat you'vc w v c d aU the -
1 Question 1. What is ihc h d a t pan about w o q ~mapping 18r o u ' ?
thet would bc. applicable. You linow you
problem from many angles and makr sure that you'vc covc~rxtt h ~ m
to SC* the

Queition 2. What would hclp !ou bcxomc a b ~ n w

conc;cpt mappd?
Ansver. Not sure. M a ~ b do
e an entire conceTt map as iui txunplt. in class.
Qucr;tion 3. Flow do conccpts 3t thc top of thc w n q t map d i l f ~liom
~ thosi: at thc bonorna?

1 î.hmk from top to bottom the go h m mon: technical to more .social.

@~mtion4. b'hich conspb in the mûp are ihmost important*?
Ansaver. - h ebase concepts (a1 the top) are most important

Question 5. Supposr: you don7 bave timr to m a t e r wmplztr m n c q t map. Which c;onccp~sdo >ou incIdr fo
convey ihe most iat'ormation'?
--- - - -

'ou n d to includr ttic: OINS in the midrllc. The! ;in- sort of much mon:vitritxi than the on* at the top so the!
w n v q more intonnation. T&e oil sprlls ( p l a d at top af mrrp) it dmn't m w q much information, A lot o h a l l e r
c o n q i s c m bs p l a d bmeath s d d y pmdures ( p l a c d m third lewl of map) 'io it i'i casier to w n v q more
intonnation
Question 6. &fore !ou start to mDatea map what plans do you makd?

I
1
AIWVCX. 1 didn't m d t ' ~ycciîïcplans for this map. 1 Sort oh-bcgan it and Ici thr i d a s tlow C i fom. 1 t ' thrt's
probûbly a vcq good way to do this. 1 might try 10 corne up with a List of d i t r m t wnwpîs h t I mi@t wmt to
imludr: in the map-if 1 was rushed for the.
 Question 7. Which of the following do !ou thinli wouid bc sasitx io rnap and why'?
A passage wiîb picm
A pa~iiilgcihat's eliq to undastand
A pawgc 'au have w d khn:

A passage that wc'w rcad kforc. II would br: casier to nnismbcr morc details and the idcas would br: mon: flowing
so w ' d bc able IO rnap thm more quickl? and in greattr detail
- -

Question 8. Why do you rtwad the test beforc !ou I

d sornc: w n c q t s togcth~r'?
1 bad to m b a dthe i a t IO get artriin spwiîic:façts. Pmxmtagc?i and numbcr relatai Hçts

Quc-stion 9. What do 'ou do if you Mme to a liak that !ou Jon't und~rstruid'?

AIISWLT.I t h i d I'd stiil indude the 1h.k but gioss OVLT rhc answtr by providing a d-ription of what I did undcrstand
- - -- - -- - - - - - - - -- - - - - -- -

Question 10. Which parts d'the test do 'ou omit fmm ihe i ; o n q t map?

Question 1 1. Wliilc you rcad the test whrt do you do to help !ou rcrn~mbcrit'?
-

h w r . WeU tvficn 1rcad an~thing1 0 to summarim the ideas in rn? hcad. Sort ot'mdic a mental point form list
Question 12. If '.ou w m çreating a map in prqamtion for a test. which of tbc following would help y u the most and
w hy '?
Savine the concepts over and ovcr
Read thc rnap as many timcs ris possiblc
D k u s the map with someonc CL% to mitkc sure you undcrsimd it
h s w r . Discussing the rnap with somconr sLw hcausc with rtpt.iition !ou s o m c i i m ~90~ b l d aHer ü certain ?oint.
If you d k u ~ it,
s ncw idem md wordings will wmc up iio you'll ~ m t m b c itr in man! din.~xcntfoms

Question 13. If you are creating a rnap fmm a techcal passage. whiçh would help !ou the most and tvh!'?
Sounding out the wods !OU doa't kaow
Skipping thc p a s you don't unrlcmtmd
Writing the passage out in Four o\m ~ o r û beet'ore
s startïng the map

~ S W L T .Writing the passage out in my o w wor& wouId be most hclptiil bexause th~mit would help to gain a kiad
of undmtmding for it-

Table E2

Interview Scores: Subiect 1

~Metaco~nitive
r

-rotal 20
Su bject 2
Table E3

M e t m ~ n i t i v eInterview: Subiect 2

1 Question 1 . What is the hardesi part about concept mopping for p u ' ?

1 Question 3. How do concepts ar hc:top of

-- -- -- -- - -
the concep map ditl'" fom h s c ût thc bottom?
- - - -

Anwtr. Sm not s m . What do !ou meau'?Ilèfiyoitr twp \tus dwec l e w b how tb tftr Irwfs tI~,@r?Bcfnre and
it will lead to clcm up
aftc~-iheoil spi11 happend bcLorc and ~~cntually

Question 4. Which concqts in t h map an:thc most important'?

I hwer.
subjat 2
'fie intmiowr was distractcd by the conhion tfiat aro'lc:in question 3. 'mis question was not askd of

AIISWLT. Tdm. oil spills and clcm up of cou= as t h ' s rcitUy wh~rr:!ou start Faur inl'onnatiou tirun. Stut u-ith
oil spiils and aprcad out h m &a(. You start with tank~xsluid spread out h m that. same [or clem up
I

Question 6. Bcfore !ou start to creatc a map what plans do !ou m;rkcb?
- - - - - - - - -

ti U~IIIPI
sure 1 have cnough room to Hnte it al1 down. ~ ~ c ' ohud ~ncmhow worrid llrur
chunge? 1 g u s for thr: onc's I know havc a lot ut*points1 will @vc: h~xn morc m m .L i k clmn up 'ou h o u it has a
lot of points so 1 kind of mtcrwl it so tht~e'dbr: mon: room for it.

Question 7. Wbicb of ibc foUo\ving do !ou tbink would bc: casier to rnap ;mi tvhy'?
A passage with picnires
A passage thst's e a to~undastand
A p ~ s a g >ou
e have =ad bcforc

h l ~ a . A passage that's as? to undmtiind or one I'vr: read kfore. 1 pess with wods I'm not sure of I muld
organize the wonis, comment on idas etc..
Question 8. Wh! do o u ratad the text &tore you Link Mme concepts to~ethrr'?

Ansver. Weil ?ou çm't d y juït reûd it once and then go on to creatc the map. You have to go back and make sure
you have al1 the pointc.

Question Y. What do o u do if ?ou corne to a that o u don't undastand'?

A n m m WcîL I'd look through the k'it and try to h d some intomation in it. if not I might l a v e it oul that w e
could end it thm.
 Question 10. Whch parts of the tcxt do 'ou omit h m thc concept map?
I

hswt'r. 1 omittd things that bad to do with s p i î i c açcidcrits Ur:E n o n VaIda. I didn't msntion anythmg spccifc
about that but gavc aii o v e n i w or lilic il s u m r n q of al1 the test

I1
- - - - -

We11.i took it in wtioas 1 startd out with tankm and th- I read al1 ihr: siutrabout tankers and thtm ail spills. Bid
FOU nrup each section individuul&:~First I w d the cmtk test then put ail spills, tankers and deon up dowm thm
~ clcm up.
rcad aU hc. links about oil spiiis thcri a11 tûe links aboui t d c thm
Qucstion 12. If o u 1vtn: creating o rnap in pnplrntioa 10r r test. which of the Collowving would help ?ou the rnost and
w hf?
1
Saping I ~ LconçcpLs
' OVLT and OVLT
R a d the rnap as man): timw as possible
Discuss Lhe rnap with somwnc e1.w IO milJic sure you undtxstand ii
tbwm. I'm not sure I'd do any of t h . 1 guc?is I'J go over it and rniikc: sun: i'd u a d e r s t d what I'd writtcm down
Qucstion 13. Ifyou are mcating a mrp tiom a tcchnical passage. which would hcip you the most and why?
Sounding out the words ?ou don't know
Skipping the parts !ou don't und~mtand
Writùig the p a w g e out in vour own woh .ifore st;trting hr:map
I

h w e r . Writing it out. that wa? !ou can gct a bcttcr p p ofccxtain conwpts irnd samctimcs Xyou cm't gct iato
Four homework ît muld m d c it casier.

Table E l

Metacomitive Interview Scores: Subiect 2

Subject 3

Table ES

Metawenitive Interview: Subject 3

Question 1. Whot is the hardwt part about concept mapping for >ou?
Anwcr. Actually I found it mcult to s e whi~bway the m w s wat. But da 1 got uwAl to it 1 hund it oasim
Question 2. Whst would hdp !ou b w m e a bettrr wawpt mappm'?
Answcr. Isupposr: ifyou wcn: io pvc an esmplc of which way thc arrows writ
Quation 3. F iow do wnwpts at the top of the concept rnap ditkr from thow at the bottom'?
Ansver. At the bottom th+ gct mucb more dctail~dbaausc ai thc top t h q an.kind of basic gczieral concept-smd
th+ get more detailai as you go d o m
-- - -- - - - - -

Qucstioo 4. Which conwpt.. in the map an: the most important'?

hswer. 1 supposi: the ones in ihr: middlc arc the most important. -In- have somewhat deiaikd inîbrmation but t h q
don't have tw detailed Ultonnation. The ont%at the bonom are just one topic and the onrz; at the top arc on a wide
h ~ m ofd topics. Onts on h e middk an:somswhat to the point but t h e do cover r fcw topics. 'fie? get the mrp
going and clarifj more
Question 5. Suppose you don't have timr to cxate a complete concept map. Wbich conwpis do !ou inçludc to
çonvey the most intonnation'?
Amwer. 1 would use esampics from al1 thnx Icveis so FOU cm c l f i thr: most important parts. 'fhat is I would usc:
çkm up but 1 wouldn't go al1 the i v q to the bottom. I mi@ include g o v e m a t but thex rirm't as important as
ofhm
- -- - - .-. -

Question 6. Bet'orc you start to m a t e a map what plw do FOU make?

.4ns~w.lot do\vn the main points 1 wanttd io IL% in the map

Question 7. Which of ihc: foliowing do !ou thdi would be casier to niap and wb'?
A pasirigc with pic-
A passage that's e a to~undastand
A pi~~~sageyou have r a d bsfori:
Aaswer- A p-p you'vl: rcd brfon: would br casier ba-ausc. if you rad wmrthing more thm once thro you get
more out of it.
Question 8. Why do you reread the test before you link some concepts togethm'?
- -

h 7 v e r . Som~ t have more important Ihings in it. Too much to

of it may be âiîlicult to undastand or some ~ f irnay
mmk.Therefore I mead to cIiui@it for m y e K

Quedion 9. What do o u do if !ou corne to a link that -ou don't undastand?

 h s w t r . I might ask for hclp or look it up in a diçtionary.
- - -

Question 10. Whiçh p a s OC the text do FOU omit fiom îhe concept rnap*?
Answtr. Rtxause it wir?in't important LWOU@ or it didn't go with w hat 1 wris tqing to do. Ilow did yorr clrcrtfe whar
wm inrponunt? [ s u p p w the b i g p topics likc gov~rnmcntinvolvtmt~tin clt';ui up. but i woulh't put down thc
iavolv~mcntot'an individual person.
Qut3tion 1 1. mile ?ou r a d the tcxt what do 'ou do to help o u r m e m k it'?

- - - -

Question 12. ICyou wtw crcating a map in prcparation for a tcst. rvhiçh ot'tbc tellowing would hdp 'ou thc most and
w hy0?
Saying the conwpts ovtr and over
Rcad ibr: map as ma! tirnt3 as possible
Discws tht. rnap wiib sornconz ~1st.to rnakc: sun: y u understand it
A n s w r . Probably Ji-wuss the map. I ï o u r d Ihc rnap o u oniy remcmbm what p u r,-rofc but if !ou discuss it !au
wver the whok tcst iastead of ju.t the arts >ou W V L T ~

Qut.stion 13. If 'ou art. creating a map from a technical passage. which would help !au the mast and why'?
.%undilie, out the words !ou dodi know
Skipping the parts !ou don't und~mtmd
Writing the pil.ssiip out in vour own words berore starting the map
Answcr. Probably witing thc passagr: out in my own words. Saunding out the uords wouliil't help mc undcrsiand il.
But if 1 \rote it out in my o w rvords would widtxstand what 1 wrotr md would undmtand it r e d y well

Table E6

Metacognitive Interview Scores: Subiect 3

Question SCOK
1 1
Subject 4

Table E7

Interview: Subiect 4
etaco~nitive
-

1 h s w r r . Rovidc rnorc information. Show mort: csamplrs

Question 3. t fow do concepts at the top of ihc conccpt map diîTt'r fonn thow at the battom'?
r

h s w c r . 'Top morc g~meralkdthm bottom

I Question 5. Supposc: !ou don't have timc to crcritc a cornpkte conccpt map. Which c o n q t s do you uicludr: to
con\- the most Latorrnation'?

Question 6. Bcfore you start to c ~ t a emap whai plans do you midie'?

h u ' c r . Which gcneralizcd ours art. ri@-w ill d k t the outcomw OC more qxciiilizrxi ones. Chm carch& so
dcn't go a i ï topic

Question 7. Which of the tollowing do !ou thid would bt: t.aster to map md why'?
A pwagc. with picturcs
A passage that's cas? to understaud
A passage >ou have read kforr

Question 8. Why do !ou mead the test before !ou Link ,=me wnwpts togethers?

Ansaver. Btxauss: whik: o u rad o u don? h o w what's going to k a p p Iater w-ha !ou =ad. You mi@ forget and it
helps you further understand the f i timds just a triai. You don't pa! much attention !ou just \vant io k n o ~what it's
about.

-- - -- - - - - - - -- -

Answcr. 1 imagine it and tha guw what might h a p p

Question 1O. Which parts of the text do you omit h m the wncq t map'?

1 Question 1 1. While !ou rrad the text what do o u do to help o u m a n b e r if?

 Ansver. Yrah 1 mi@ rralize this page is not important so 1 r a d anotha o n e I ~ t L bnck
r IO the fcxt
Question 12. If you wrc: crcating a map in prtpariition tor a test. whiçh of the Lollowing would help !ou th most and
w hy'?
Saying the w n q t s ovcr and ovtr
Rcad thc map as many thcs a,*i possible
Discus*ithe msp witb somwtlt. e1.w to mnkc sure !ou undtmtand it

Quwtion 13. If you are cmating a map tlom a tcchnical paw7gc. tvhich would hslp !ou the most and why'?
Sounding out the wods !ou don? knaw
Skipping the parts you don't undmtnnd
Wnting the pa.~riilgt.out in your OHII w o d s bdow starting thc rnap

Table ES

iMetacornitive Interview Scores: Subicet 4

Qucstian Swn:
1 3

2 1

3 2
4 2
5 2
6 2
7 O

8 2

9 2
1O 2
11 2
12 2
13 2
Total 23
Subject 5

Table E9

Metacwnitive interview: Subiect 5

- -- - -

Question 1. What is the hanIcst pai? about w n q t mapping h r p u ' ?

Answur. The data itrrns. Structuriag the links. Thc way ibis part c;onncxts to that p u t md how to writc that out.
Quesiion 2. What woufd help !ou b~xomca k t t e r concept mapprr'?

An'iwcr. Mon. time ovcrnight to thiak about it.

Qumtion 3 . 1Iow da mucepis ai the top of the wnupt map diffcr r o m th= at thc bottom'?

h s w r . 'i'hctop is a broad subjcxt 3s !au kcxp piag d o w p u gct morc details

Qumtion 4. Which concepts in the map an:thc:most important:'

Question 5. Supposc.'ou don't haw tirne to crcatc: a complctc conccpi rnirp. Whiçh wnwpts do !ou uiçludc: to
con\;+ the most Xormation'?

t h i v e r . Thc top b a a w ?ou cm do thc top hzlf without the bonom. Bottom LS just clrborriting more on the top.

Qut?ition 6. Berore you stûri to creatr: a rnap what plans do !ou mdd?

Aus1vsr. llink about it a bit and t

h a d~xidcwhat I would do in tht: situation. Sort of t h i d if 1 n.ally did m a t c an oil
spill what would I do.

Question 7. Which of the foiiowing do you thdi wouId bc casier to rnap and ~ d g ?
A pac;sage with picturcr;
A passage that's e a q to undtrstmd
A passage you ha= r a d before

Answer. A passage I've nad before (i-; I'd bt: mort: cornfortable w i h the subject matter

Question 9. What do you do if ?ou corne IO a LinCi that !ou don't undmtmd?

h 7 v r r . I'd probably skip it and go back to it later. rd sort of fiusricxl on other parts of the maps and then uk
somebcdy
- -

Question 10. Which parts of the text do you omit h m the çonccpt map?

Answcsr. 1 don't think f r e d y omitted anphhg maybe tfie ~pecificpiirts of what b p p m d to s v q t h h g at the
kpinnin@
 Question 12. If 'ou w m mcating a map in prcparatian for a tmt. wbicb of the following would hdp !ou the most and
why'?
Swing the macepis over and o v w
Read thc msp a!s mm! timrs as possiblr:
Discuss ihc map with somwnr: CL* IOmdt. surc !ou und~rstsndit

Qucstioa 13. If 'ou ;in:crcaiing a map h m a t ~ ~ h n i c passage.

al which wouid hctp !ou thc most riad Hh'?
Sounding out thc wods you don7 h o w
Skipping the parts p u don't undcrstmd
Writing thc pi~s~ilgcout in your own uords bcforc starting thc map

\vords. Not surc if î'd \Vtitr: it d o w or W&C it out in rny hcad

ha?itvcr. Writing it out in my o ~ m

Table EIO

etacomitive Interview Scores: Subiect 5

Subjeet 6

Table EL 1

etacopitive Interview: Subject 6

Question I . M a i t*thc hordcxt piut about concept niapping t'or ?ou8?

AUSTVLT.Pmbably widcxstanding what ha5 Io be donc was h c most difiçult part for me. I rcally didn't u n d ~ n t m d
what had to bt: donc in thc kginning. Lsacify hot, it should bc sct up. whiçh May thc m o ~ go.
s what gocy w r c
1 Question 2. M a t would hrlp p u b ~ w r n ar bcttcr wnwpt rnoppm*? 1
Answw. 1 pues5 thc woy it wali esplaind. I don't know bow it would be a s i m !or rnc to understand that. Did 1 do it
C)K so 1 didn't catch on ma~bc:ifvou wuld h d a dift'~nat\vay ofesplaining ir.
ri@? ~'L'.P
1

Questiou 3. 1IOH. do c'onwpts al the [op of tbc concept mop ditlia tom those at ihc bottom'?

Anwtr. I ~ U L Wthe top one is the one that a f i a t s thc most m a . Acrually uow that 1 look at it 1 g u ~ ~safkty
i s akT~r:tsa
loi of people. 'ïhat LS 1 ptxs the top one is more f w u s d to the subject in i t s d tbun the bottom on&%..Ire yrm
s-.ing rhar rire nne ut rltc hortcrtii uw rtrorr gctrrrcil? Ycah .lrid rlir orle sur rhc [op are n i c m .rpt.ciJk? Ycih
I

Question 1.Which concepts in the rnap arc the most important*?

1 hm~r. Probob. the ones at the top as th~$re thr: most qxciîiï IO the topiç 1
Question 5 . Suppose ?ou don't have the to creatr: a mmpletc: concept mûp. Which c o n ~ ~ pd~
t syou IllçIUdc: 10
convcy thr: most d'ormation'?

1 Qualion 6. klbre -ou start ro mal<: n rnap ahnt plans do p u mnlic.? 1

Ans-va. What 1 thought about right a w q was the m n d Ievel luid thtm I a d d d to that md thra 1 a d d d mon: IcvcLs

Question 7. W c h of the toilowing do !ou thinCi would bt: saster to map and why'?
A passage with picturc5
A passage that's t'a+- to undt'~?itaud
A passage o u have =ad beforc

AIMWLT.Probobly a passage that i've =ad bet'ore Lt'you Cr:me rn opportunity to rcad it thm w h a I'm mapping it
I'm obviously familiar wth it-then to retir=sh if I hat-e it wit me it hclps 1
- --- - - -- -

Probably if somethnlg wasn't CIW to me or it- 1 had forgottcn what the test had said
1 l

Quadion 9. What do ?ou do if 'ou corne to a that !ou don't undmtmd?

1

Answtx- Robably stop and ignore that pari CCbuldo u r+r back lu iltr zerl? Wetl it dqxnds. if it ixiw a th! simal1
topic: t h t wasi't n w he top lmel then i probabh wouldn't

1 Question 10. hi ch pvrs of the tzvt do 'ou omit h m the concept mapy I
 Answtr. 1 cm't givc you )rpwitics5s 1 don't rcmcmbtr thc test ruijrnorc. 1 g u a s anythuig that d-n't .ama!!large a
w n c m as those 1 did chose.

1 Question 1 1. Whilc o u r a d the test wliat do ?-oudo to help )ou wmwnbcr ilo?

Quatioa 12. It'you werc crwting a map in ptwparatian for a tmt. rvhich of the foIlowiag would hclp !ou thc most mci
w h!?
~ ovtr
Saying the concepts o v t and

I h w c r . Disussing it with somwne elsc. Myxiï I cm't rcmcmbtr v e n wsll Espc-iaily if' it's j uït reading or ing
it ta rnyseif If I'm taking about it and heu d i f f a n t ideos 1'11 probably casily i m i a n b c ~it ai ihr tirne of the test

C)uestion 13. If !ou arc cxatuig a rnap h m a twhnical passage. whiçh would help !ou the most md why?
Sounding out the wanls !ou don7 how
Skippmg the parts !ou don't undtr'itaad
Writing the pwsagc out in Four own wonts kforc starting thc rnap

A n s w x . Weil I woulda't do thc tint onc as 1 have bad mcmory. if 1 skippd parts t dih't undmtmd thtm 1 might
not undastand things Lirrthrr on in the passage. So I gurris writing it out in rny own words. Finding d&nitians and
exact mcaning of certain things.

Table El2

.Metaco~nitiveln terview Scores: Subiect 6