ArticlePDF Available

Memory, Knowledge, and E-Learning

Memory, Knowledge, and E-Learning
Dietrich Albert1, Cord Hockemeyer1, and Toshiaki Mori2
1 University of Graz, Austria
2 Hiroshima University, Japan
The aims of this chapter are to stimulate ecologically
valid research in psychology of memory and of
knowledge, as well as stimulating the future
development of e-learning. Already Herrmann
Ebbinghaus, the founder of modern memory
psychology, called attention to apply his results to
educational practice; especially his results to
distributed practice and retention. An early textbook
about memory psychology written by Offner (1913):
Das Gedächtnis (The Memory) also focused on
the pedagogical applications of the scientific results.
Nevertheless, more than one 100 years later,
Ebbinghaus' and his successors' results on massed and
distributed practice still have not been applied in
schools. This is more or less the general picture of
applied memory psychology in education even after
Neisser (1978) called it scandalous“ and demanded a
more ecological memory research.
Harry P. Bahrick (2000) recently pointed out the
following reasons for this ignorance. Traditional
memory research had little impact on education, since
it was focused on the short term retention of episodic
content, a domain that has only little relevance to the
concerns of educators. Furthermore, educators have
emphasized the immediate achievements of students,
paying little attention to the effects of their instruction
on long-term retention of content.
Bjork (1994) pointed out that making training too
quick and easy has the additional disadvantage of
giving trainees an unrealistically inflated
metacognitive self-assessment of knowledge and of
the likelihood of long-term retention of knowledge and
skills. Overestimations of competence can have
disastrous consequences in particular work
environments, for example air traffic control, police or
military operations, or nuclear power plants.
Christina and Bjork (1991, p. 47) concluded that
"the effectiveness of a training program should be
measured not by the speed of acquisition of a task
during training or by the level of performance reached
at the end of training, but, rather, by a learner's
performance in the posttraining tasks and real-world
settings that are the target of training."
The results of Bahrick and others can be applied
for making long term predictions of retention, for
maintaining the knowledge and - if necessary - for
refreshing at due date, consequently obtaining
permanent storage of the memory contents. As this
example already demonstrates, not only the models of
memory but also the empirical results themselves can
be used for improving e-learning systems
We would like to add: In the past, the influence of
psychology on the practical teaching processes was
indirect and complicated. Influencing the teacher
training and the teachers behaviour in the classroom
involves many stages. Even an ecologically valid
memory research probably would not have been
successfully transformed into educational practice of
classroom teaching. However, by developing
e-learning systems, memory psychologists can
influence learning and teaching more directly by
improving the teaching systems. It may be much easier
to shape the 'behaviour' of the e-learning systems than
that of the teachers.
Indeed, there exist at least three successful
examples: The multi-talented Richard Atkinson and
Patrick Suppes co-founded a very successful company
(Computer Curriculum Corporation; now integrated
into Pearson Education Technologies) for
computer-based learning and teaching more than 30
years ago. John R. Andersons ACT model has
successfully been applied for elearning applications.
Recently, Jean-Claude Falmagne and his group created
the fully automated, interactive, adaptive, and
integrated mathtutor ALEKS (Assessment and
LEarning in Knowledge Spaces) - available only on
the Internet - and founded the ALEKS corporation.
Conversely, using ecologically valid memory
research for improving the e-learning systems, these
systems can be used to improve the results of
ecologically valid memory research. E-learning
Systems may be used as a kind of world wide
learning- and memory-research laboratory.
The aim of teaching is, of course, to create
knowledge and competence. Thus, in addition to
psychology of memory, teaching must also be based
on the psychology of knowledge (see, e.g., Albert,
1994; Albert & Lukas, 1999). As a result, the
psychology of memory may profit methodologically
from the psychology of knowledge.
Thus, three sub-disciplines have to be combined,
namely psychology of memory, psychology of
knowledge, and psychology of e-learning. With
“Offensive content reduction,” the search engine
FAST SEARCH found on Monday March 4, 2002,
at 23:00 for memory” 7,975,787 pages, for
knowledge 14,512,585 pages, and for “e-learning
OR eLearning 620,650 pages in the Internet. For
Memory AND knowledge AND (e-learning OR
eLearning),” 4,136 entries were found. Even if we take
into account, that all these terms are also used in
technological research and development, it does not
seem to be necessary to search also other data bases,
e.g. for psychological literature, for demonstrating
how huge each of the three sub disciplines are. Thus,
we have to focus in this chapter on the intersections of
memory, knowledge, and e-learning.
Former and Recent Contributions of
Psychology to E-learning
Technician's E-Learning Definition
Current definition or aims of e-learning by technicians
are: having access to electronically based learning
resources anywhere, at anytime, for anyone.
Nowadays the most advanced technology for
e-learning is the World Wide Web, WWW, Web or
Internet. Thus our aim is Web-based or Internet-based
e-learning, however directed by Psychology and
A Psychological Structure of E-Learning
Depending on the taken point of view (e.g. computer
and network technology, artificial intelligence, service,
educational, and psychological point of view), an
e-learning system consists of a special set of
components (see, e.g., Albert & Mori, 2001). In
Psychology, usually four components of a computer
based learning and teaching system are included:
Knowledge Base
The Knowledge Base contains the structured
expert knowledge about the knowledge domain.
Student Model
The Student Model represents the hypothetical
knowledge state and other attributes of the student.
It may capture e.g. the student‘s knowledge,
misconceptions and general skills. It is the basis
for individualised pedagogical intervention. It
has to be adapted to the learning progress.
Teaching Model
The Teaching Model decides about the
pedagogical interventions taking into account the
knowledge base and the student model.
Interactive Human Computer Interfaces
The Interactive Human Computer Interfaces are
needed for presenting information to the student
and receiving information from the student.
Cognitive psychology contributes to the development
of e-learning systems by improving these four
components by applying its theoretical and empirical
Former and Recent Contributions
Programmed Learning
Using the computer for teaching aims based on
psychological findings, started with Programmed
Learning which is based on Skinnerian principles of
operant conditioning. The basic idea was to break
the information into small simple pieces presented to
the student sequentially and giving immediate
feedback ("reinforcement") after each learning step -
depending on the student's answer. Thus, programmed
learning is addressed to the systems teaching model. It
is the first of the psychological models realizing the
principle of an adaptive, personalized testing and
teaching. However programmed teaching machines
were limited in their ability to adapt to individual
differences among students and to provide a
stimulating, responsive environment for students.
Programmed learning was popular in the sixties.
However, is was finally not successful. The main
reason is of theoretical character. In traditional operant
conditioning, contingencies between already available
and easily accessible behaviour and the contingent
behavioural outcomes have to be learned. In classroom
learning for instance, new and difficult cognitive
contents and operations have to be acquired and to be
applied. However, some aspects of programmed
learning are still used nowadays in computer based
training of the drill type.
Mathematical Models of Rote Learning
Mathematical models of rote learning assume that the
learning progress is described either by incremental
growth or by stepwise transitions between learning
states. By individualised parameters for learning tasks
difficulty and learners ability, the knowledge domain
and the students are represented. The models have
been developed in the fifties and sixties of last century
based on Estes’ Stimulus Sampling Theory or Bush
and Mosteller’s Linear Operator approach. ‘Computer
Assisted/Aided Instruction‘ (CAI) applications starting
with language acquisition have been initiated by R.
Atkinson, and P. Suppes. Especially the latter is one of
the pioneers and leading figures in educational
computer applications on the basis of psychology. He
is still active in this field, e.g. by contributing to
Stanford University's e-learning program for gifted
Cognitive Modelling
The Cognitive Modelling approach has the aim to
simulate the cognitive processes of humans by
computers, e.g. the cognitive processes involved in
storing information, solving problems, meta cognitions,
detecting inconsistencies, etc. By socalled Intelligent
Tutorial Systems (ITS) or Intelligent Computer Aided
Instruction (ICAI) these models have been used as the
student model component in e-learning systems with
strong impact on the teaching model component of the
system. Since the eighties of last century, learning to
program a computer language, to solve complex
physical problems and so on have been modelled.
Among others, the working groups around John R.
Anderson, Hans Spada, and Karl F. Wender should be
mentioned. Several of these approaches until now are
more or less only of academic value, while interesting
results have been reached, many of them still remain
theoretical and difficult to apply in the real world of
broader field applications. However, in 1998 the
Pittsburgh Advance Cognitive Tutoring (PACT) Center
with the company Carnegie Learning was founded by
a group around John R. Anderson. Their algebra tutor
and other courses, originally developed at Carnegie
Mellon University based on Andersons ACT-theory,
have been very successful applied in the field.
Knowledge Space Theory (KST)
Knowledge Space Theory is a mathematical
psychological framework (based on order and lattice
theory) for representing knowledge structures. The
theory is using dependencies between the problems
and other learning objects in a knowledge domain in
order to structure the assessment process and the
teaching process for adaptive, personalized teaching,
like a private teacher. Knowledge Space Theory was
founded by Doignon and Falmagne in 1985 and has
been enormously extended and applied to knowledge
assessment and web-based teaching since that time
(see, e.g., Albert & Lukas, 1999; Doignon & Falmagne,
1999). Key concepts of knowledge space theory are
surmise or prerequisite relationships between test
items or learning objects, knowledge states and the
knowledge space, and learning paths. E-learning
applications based on Knowledge Space Theory are
the systems ALEKS, and RATH, which is a prototype.
Some basic concepts of Knowledge Space Theory will
be introduced in more detail below.
Current Aims of Applying Psychology
to E-learning
Psychology should have a strong impact on e-learning
systems (see Albert, Hockemeyer, & Wesiak, 2002), as
it is, e.g., demonstrated by Clark and Mayer (2002).
Here we would like to focus on only a few aspects.
Psychological Expertise in Methodology
Psychologists expertise in computer- and
internet-based presentation, assessment, data
collection & experimentation will improve the
e-learning systems and it‘s usability. The most
advanced technology in e-learning is the internet. Also
from psychological points of view internet-based
e-learning systems are preferable. Why is cognitive
psychology a strong partner in developing
internet-based e-learning systems? The main
methodological reasons are:
Applying empirical findings to structuring content,
creating student models and teaching models.
Theory-based models of learning, reasoning,
problem solving, knowledge retrieval,
remembering and retention which have been
proven to be empirically valid.
Test theories for assessment and diagnosis of
performances, knowledge, and acquisition
Methodology for creating and presenting learning
Clearly defined learning objects and strict control
on learning objects.
Clearly defined and specified (classes of) actions
of the teaching system.
Methodology for recording the students' data,
their traces of learning behaviour.
Methodology for collecting, analysing and
interpreting behavioural data, which are not only
answers to multiple choice questionnaires or
solutions of problems, but also latencies, eye
tracking data, movements e.g. of the computer
mouse, video recordings, as well as
psycho-physiological data.
Expertise in computer-based and Web-based
experimental methodology for assessing and
shaping behaviour and the processes and
structures behind the behaviour no other
discipline offers this.
World wide collection of data and information for
testing the psychological and educational models
or theories of teaching and improving or refining
the e-learning systems.
Individualisation and Adaptivity
The aim is to make the Internet-based systems
behaving like a private teacher. A private teacher has
only one student at a time and can adapt his/her
teaching behaviour to the individual student, to the
student‘s learning state, his abilities, needs and so on,
using his knowledge about this student. However a
private teacher is also using his/her knowledge about
the theories, results, methods, procedures and
techniques of learning and cognitive science, including
the psychology of memory, cognition, and knowledge.
Already in the 60s Patrick Suppes focused on
using computers to individualize instruction“. By far
the most important topic in current e-learning
Research & Development (R&D) is adaptivity.
Adaptive hypermedia systems bridge the gap between
technology driven tutoring systems the risks: mental
overload, de-motivation - and student driven learning
environments the risk: being lost in the hyperspace.
We distinguish between the directions and objectives
of adaptivity, the objects of adaptivity, and the level
of individualisation of adaptivity (Albert & Mori,
2001; Riley et al., 2001, Section 2.3).
Directions and Objectives of Adaptivity
Directions and Objectives are the adaptivity to the
requirements of different learning cultures, the
adaptivity to the teacher‘s/student‘s aims and goals,
the adaptivity to the student‘s properties (e.g. his or
her (pre-) knowledge, preferences in Human Computer
Interaction (HCI), communication style and needs,
cognitive and learning style, and cultural background).
Learning itself is an adaptive process, the
e-learning system has to support this process by being
adaptive itself. Supporting the student by an adaptive
E-Learning-System means: (a) The intended
individual learning process has to be optimised
according to specified criteria. (b) „Learning to
Learn“, i.e. control processes and meta cognitive
processes may be also a learning aim. (c) Other
learning processes and demands to the student have to
be minimized to overcome of limited cognitive
processing capacity and time constraints.
Objects of Adaptivity
The question of what is adapting in an e-learning
system has a strong relationship with the objectives of
adaptation. Certainly, from a psychological point of
view the four components of the e-learning system
Knowledge Base, Student Model, Teaching Model,
and the Interactive Human Computer Interface have to
be made adaptive by the documents‘ contents, their
kind of presentation, documents‘ granularity, courses
contents and structure, navigation by dynamically
generated learning paths, and navigation support
Level of Individualisation of Adaptivity
The level of individualisation of adaptivity may differ
depending on the learning culture, the status of
technology and the status of research. In some cultures
a high amount of standardization is required. Group
level adaptivity is common for subgroups or minorities
in a culture, e.g. for people with special needs.
Individual level is usually aimed in western cultures
with the model of private teaching in behind.
However, in many cases the actual status of
research allows only group level adaptivity, e.g. based
on an empirical effect which has been found valid for
a special group of persons.
Some of these objectives of adaptivity have
already been realized in existing e-learning systems.
We will introduce two of them which are based on
Knowledge Space Theory.
E-learning and Knowledge Space
Knowledge Space Theory (KST) was originally
developed for the efficient, adaptive assessment of
knowledge (Doignon & Falmagne, 1985). Meanwhile,
however, the focus of applying KST has moved to
e-Learning. We will first introduce some basic
concepts of KST and, afterwards, look at some
e-Learning systems based on it.
Basic Concepts in Knowledge Space Theory
Within the scope of this chapter only some basics of
Knowledge Space Theory are presented. Here we want
to mention that the theory is not merely a psycho-
metric model but that it is linked to cognitive
psychology, including memory psychology. For a
more detailed and more technical description, the
reader is referred to Doignon and Falmagne (1999)
and Albert and Lukas (1999); a comprehensive list of
literature can be found at
Example Problems
Without going into details, obviously a structure is
inherent in the problems presented in Table 1. For
instance, the solution process of problem (a) is
included in the solution process of problem (d). Thus,
(d) is more difficult than (a), and mastering (a) is a
prerequisite for mastering (d).
Table 1: Example problems for the knowledge
domain “Elementary Calculus” (Lukas & Albert,
1999, p. 5)
Prerequisite Relation
For another set of problems u, v, w, and x a
prerequisite relation (surmise relation) is illustrated by
the graph in Figure 1 (a). For instance, problem w is
more difficult than problem x. Problem x is a
prerequisite for problem w. (From solving problem w
solving problem x can be surmised.) Problems v and w,
also w and x are independent.
Knowledge Space
For each prerequisite relation (surmise relation), there
exists exactly one knowledge space (a theorem of
Birkhoff), which is the set of knowledge states, i.e. the
set of expected answer patterns corresponding to the
respective prerequisite relation. For the prerequisite
relation exemplified in Figure 1 (a), the corresponding
knowledge space is illustrated in Figure 1 (b).
Doignon and Falmagne‘s Generalisation of Basic
Knowledge Space Theory
Until now we took into account for each item only one
set of prerequisite items. The illustrating graph is an
AND-graph. However in practical applications items
are often solved by different kinds of prerequisite
knowledge. In these cases the illustrating graph is an
AND-OR-Graph. A proof by Doignon and Falmagne
shows that also in these cases a one-to-one
correspondence exists between surmise/prerequisite
{v,x} {w,x}
{v} {x}
Figure 1: Example of a surmise relation and its
corresponding knowledge space
structures and knowledge spaces.
E-Learning Systems based on Knowledge
Space Theory
Adaptive Learning Environment based on Knowledge
Spaces (ALEKS)
ALEKS is based on Doignon and Falmagne‘s
generalisation of their Knowledge Space Theory.
ALEKS has been developed at the University of
California by a team around J.-Cl. Falmagne with the
support of a multi-million dollar grant from the US
National Science Foundation. Falmagne is also the
founder and chairman of ALEKS Corporation
ALEKS contains the complete K12 mathematics
contents. The structure of the objects was obtained in a
twostep procedure: First, experts were queried about
prerequisite relationships and, afterwards, the preli-
minary structure was refined based on student data
(Cosyn & Thiéry, 2000).
The system itself is not a classical eLearning
system insofar as it is centred around test items. After
adaptively assessing a new learner’s knowledge state,
appropriate new items are suggested as training items.
Whenever the learner has shown a certain performance
on such a new item, it is included in their knowledge
state which also leads to a change in the set of
appropriate items
Relational Adaptive Tutoring Hypertext (RATH)
A teaching-centred system based on Knowledge Space
Theory is the internet-based tutorial system RATH
(Hockemeyer, Held, & Albert, 1998; http://wundt. It is a prototype demonstrating adap-
tive presentation of learning objects depending on the
specific (pre-) knowledge of the student. This is done
through linkhiding, i.e. links to learning objects for
which the student does not yet fulfil all prerequisite
are hidden.
Due to its prototypical nature, RATH starts with
the assumption of a complete novice. Whenever a
teaching document is presented to a learner, it is
preliminarily assumed that its content is acquired by
that learner. However, if there are test items within the
prerequisite structure, the learner has to solve them in
order to get beyond that point.
Currently, RATH contains a tiny course on
elementary probability theory structured through
demand analysis and componentwise ordering (see
below). However, system and content were developed
independently, i.e. a different course can be installed
into RATH at any time given an appropriate
specification of the prerequisite relationships between
the objects.
How to obtain Knowledge Spaces
In order to apply a knowledge space to adaptive
assessment, to classroom teaching, or to adaptive
tutoring, to developing or restructuring a curriculum,
and to describing learning objects by computer-
readable metadata, (a) first one has to obtain the
knowledge space and (b) second the knowledge space
has to be investigated to be empirically valid. Here we
focus onto the first question, that is, on how to get the
knowledge space.
There exist three traditional, straight forward
methods to get knowledge spaces: to analyse mass
data of performance tests with respect to prerequisite
relationships, to get answer patterns which may
correspond to knowledge states, and to ask experts
about the surmise and prerequisite relationships. From
these kinds of data the missing other structure, the
knowledge space or the surmise/prerequisite relation,
can be derived easily because of the one-to-one
correspondence between the two structures.
In the sequel, we discuss approaches based on
psychological analysis, taking into account the
cognitive demands, latent, unobservable knowledge,
and the cognitive processes (see Albert & Lukas,
Demand Analysis and Component-wise Ordering
Learning objects can be described by attributes on
several components. The attributes can be ordered
according to difficulty (Figure 2, upper part) based on
analysing the related cognitive demands (Albert &
Held, 1994; Held, 1999). The learning objects can be
located in a structure obtained by a Cartesian product
of the components. Thus, the prerequisite
relationships between the learning objects (Figure 2,
lower part) can be obtained by component-wise
ordering (dominance). This method has been applied
successful for different knowledge domains. A detailed
example is shown below.
Direct Skill and Competence Assignment
By direct skill and competence assignment (elaborated
by Doignon, 1994; Düntsch and Gediga, 1995; and
Korossy, 1993, 1997), each item, problem, or learning
object is mapped onto a subset of skills and
competencies. The advantage of this approach is that
the number of items and objects can increase without
increasing the number of skills and competencies.
Applications have been made, e.g., in the fields of
b b
ab ab ab
ab ab
Figure 2: Component
geometry and algebra. The aim is to assess the
competencies and to learn and teach competencies
instead of performances.
Process Analysis
Process analysis (Schrepp, 1999) uses process models
of cognitive psychology, individualises these models,
e.g. by means of sub-models of an expert production
systems, and derives the knowledge space for a set of
items in the respective domain. Successful
applications have been made in inductive reasoning
tasks, e.g. letter series continuation.
Demands, Componentwise Orders, and
Indirect Skill Assignment in RATH
Demand analysis and componentwise ordering have
been successfully applied for developing the RATH
course based on the work by Held (1999). The
methods used for the course development have been
described in more detail by Albert and Hockemeyer
(2002). Learning objects can be described by attributes
on components. Types of learning objects are obtained
by a Cartesian product. The attributes are ordered
according to difficulty. The attributes' difficulty orders
are obtained by the inclusion principle applied to the
sets of demands assigned to the attributes. The
prerequisite relationship of the learning objects are
obtained by component-wise ordering (dominance).
An Example Problem in Elementary Probability
Theory (used in RATH)
A typical example of a problem used in the RATH
tutor is "An urn contains three red and three blue balls.
Two balls are drawn successively. Drawing is
performed with replacement. The drawn balls are red.
Compute the probability of this event."
Six problems, labelled A, B, C, ..., F, of this kind have
been structured. The resulting prerequisite structure
and its corresponding knowledge space are shown in
Figure 3.
Problem Components and their Attributes
Three components are used to describe the problem
types with two or three attributes for each component
(in parenthesis).
(a) Numerical ratio of differently coloured balls
(equal to one, not equal to one)
(b) The way of drawing
(one-off, with replacement, without
(c) Specification of the asked event
(only equally coloured balls are drawn, exactly
m of the drawn balls have the same colour, at least m
of the ...)
Attribute Ordering by Demand and Competence
Two examples for attribute sets (components) ordered
by inclusion of the demand sets which are assigned to
the attributes are presented in Figure 4 for illustration.
On the left side, we obtain a linear order while, on the
right side, we have a partial order due to the
differently structured demand assignments.
Demands of the Problems
The following demands have been used for assigning
to the attributes.
1. Knowledge that, in general, Laplace probabilities
are computed as the ratio between the number of
favourable events and the number of possible events.
2. Ability to determine the number of possible events.
3. Ability to determine the number of favourable
events if one ball is drawn.
4. Ability to determine a favourable event if one ball is
drawn, or if the sample for which the probability has
to be computed consists of equally coloured balls.
5. Knowledge that if an outcome like “exact/at least n
balls are of colour x” is asked for, all possible
sequences of drawing are favourable events.
6. Knowledge that probabilities are added for two
disjoint events A and B.
7. Knowledge that probabilities are multiplied for two
events A and B that are (stochastically) independent.
8. Knowledge that the probability of drawing a ball of
a specific colour is not equal to 0.5 if there are
different numbers of balls of different colours in the
9. Knowledge that drawing without replacement
reduces both, the total number of balls in the urn as
well as the number of balls that have the same colour
as the drawn ball.
10. Knowledge that drawing at least a number of
{d, d, d}
{d, d}
{d, d, d}
Figure 4: Attribute orderings based on
demand analysis
Figure 3: Prerequisite structure (left) and
knowledge space (right) for exercises in the
RATH course
certain balls includes the - not explicitly stated -
results of drawing more balls of the certain kind.
Teaching Units
We took the demands as teaching contents in our
prototype RATH. Thus, they are the tobetaught
competencies which have to be learned for solving the
problems. Each of these 10 demands/competencies is
represented by a teaching unit containing a teaching
lesson (an instruction). Furthermore, most of these
teaching units also contain some examples. These
examples have the respective lesson as a prerequisite
but are no prerequisite of any other learning object
RATH‘s Learning Paths
During the tutoring process, the student's knowledge
state can be updated at any step corresponding to the
lessons learned. At any time, the student has access to
those learning objects which are the next step on some
learning path from the current state to the destination
or goal state (Figure 5).
Key Concepts of RATH
Prerequisite relationships establish possible
knowledge states and learning paths. Depending of the
student‘s actual knowledge state she or he can choose
among next steps for those learning paths he or she
has the necessary prerequisite knowledge for
understanding a lesson or solving an exercise. Other
lessons or exercises are hidden for the student and are
not accessible. The system adapts to the student‘s
changing knowledge. The knowledge of a student is
defined as well by his/her performance and his/her
RATH is an example of an adaptive system
applying results from cognitive psychology. Based on
this, the question arises how such systems might be
improved by the integration of results from memory
Potential of Memory Research for
The human memory and the memory traces are the
basis and the substrate of human learning and
knowledge. Certainly psychology of memory (see ,
e.g., Albert & Stapf, 1996; Tulving & Craik, 2000) has
the potential to improve e-learning systems like, e.g.,
the RATH system. To make this explicit, some
examples will be given.
Optimising Speed of Learning and Duration of
Guiding the student during the acquisition of new
knowledge aiming not only to effective learning but
also to long term retention has been requested already
by Bjork. The models and results of Memory
Psychology referring to massed and distributed
practice for learning and life long retention are the
basis for individual guidance during the acquisition
Optimising by Maintaining and Refreshment
Weeks, months and years after new learning,
signalling the individual student on activities to
maintain knowledge, to avoid forgetting of prior
knowledge, is another aim, as mentioned already by
Bahrick. The models and results in the field of
forgetting, retention and maintenance of knowledge
can be the basis for the system‘s individual guidance.
Optimising Coding and Cognitive
The methods and results for multi-attribute coding,
multiple-modality coding and deeper levels of
processing are the basis for optimal coding and
adequate rehearsal strategies. The different kinds of
cognitive representations, e.g. pictorially, verbal, are
important for cognitive functioning in learning,
problem solving and creative processing. The same
content can be represented mentally in different ways.
Knowledge about KST is a good example, because
different representations are possible: sets, graphs,
vectors, matrices, formulas, orders, lattices, logical
operations are used in representing the concepts of
KST. Furthermore the representations' interrelation-
ships and their relations with the external learning
environment have to be specified. The aim is to
facilitate the generation and interpretation, as well as
the flexible and creative usage of the suitable
representations by e-learning and e-teaching for
problem solving and storage.
Optimising Mental Load
Since G. E. Müller and A. Pilzecker the concept of
narrow consciousness or limited working memory has
been worked out in memory psychology, and many
empirical results demonstrate the importance of this
concept. Thus, also in applied settings, like e-learning,
the model of working memories in combination with
coding and rehearsal strategies may be the basis for
Figure 5: Learning paths in RATH
Letters refer to exercises, numbers refer to teaching
optimising mental load.
Optimising the Conditions for Successful
Retrieval and Usage
For applying KST, it is essential that available
competencies and knowledge are accessible and used.
Thus retrieval problems and decision failures have to
be minimized or coped with. The results on context
effects, fan effect, recall latencies and so on are the
basis for optimising retrieval and usage of available
Recall and Recognition
In e-learning the question of answer-formats is of great
importance because it has been shown that
performance strongly depends on the answer format.
Psychology of memory has elaborated different types
of retrieval methods, like recall and recognition, cued
recall and prompting. The retrieval methods
correspond to some extent to the different answer
formats and may explain performance differences.
Thus, the results and models on retrieval can be used
in e-learning systems in order to optimise the
performance measurement.
Optimising Transfer
The aim of at least higher education is to enable the
student to use his or her knowledge not only for tasks
used in learning and training but rather in other
contexts and environments. The research on general
and specific transfer as well as on context effects are
the basis not only for acquiring specific competencies
and for learning how to apply them, but also for
transferring them.
Levels and Hurdles in Applying
Memory Research
Modern memory psychology has been existing since
more than 100 years and a large amount of wisdom
about human memory has been accumulated since
Ebbinghaus, however not finally integrated. The one
and only one theory of human memory is missing and
many empirical facts are still not integrated by any of
the existing models and theories. Thus different levels
of memory research have to be distinguished in
application. Applications can base on
Empirical Effects and Phenomena explained ad
Partial Models
Global Memory Models
Global Cognition Model
The Theory of the Human Mind
Empirical Effects and Phenomena explained
ad hoc
We may blame ourselves, however as former memory
researchers and current teachers of experimental
psychology we are not able to overview all the
empirical effects and phenomena in memory research.
Experts for special fields in memory psychology may
be helpful for developing e-learning systems. However
are they willing to participate not only in research but
also in development like engineers? (a) Needed are
even more standardised, and computer readable
descriptions of the investigations and their results in
memory research. (b) Availability of the rough data in
a computer database is necessary. (c) An expert system
is necessary for scientific handling of (a) and (b) for
basic research, applied research, and applications like
e-learning. (d) Interfaces between these three
components, the programs for computing or simu-
lating memory models, and elearning systems are
required and should be developed.
Partial Memory Models
Partial models are good candidates for applications in
specified learning situations, for instance a model for
recognition (e.g. Signal Detection Theory based or
Random Walk Model based) or for free recall (e.g.
Linear Death Process Model-based or Associative
Network based) depending on the kind of knowledge
retrieval. However in applications usually different
models have to be used simultaneously, even both
recognition and recall are often involved in the same
situation, e.g. in comprehending an instruction or
solving a problem. Thus, in applied settings the
question is how to combine different partial models or
how to find a global model.
Global Memory Models
Recovering from the decline of behaviourism first
partial and than global models of memory have been
developed. The problem is, that there are several of
them which are excellent candidates for applications,
however with a lot of overlap, e.g. SAM, MINERVA2,
TODAM, REM. Which one should be used for
e-learning applications? Two aims are involved in
applying the models. (a) The relationships between the
models have to be analysed. (b) It should be specified
how to link the memory models with other models of
cognitive functioning, e.g. for problem solving, for
reading and comprehension. Alternatively, we can take
one of the global cognition models for application.
Global Cognition Models
With the availability of powerful computers several
schools developed global models of human cognition
arisen, e.g. the teams around John R. Anderson and
around Dietrich Dörner. Anderson's ACT model has
already stimulated successful applications in
e-learning. Often, however the following problem
arises in applying these approaches of cognitive
modelling. One has to be a member of the respective
school for being informed detailed enough about the
functioning of the computer based model in order to
be able to understand and to apply them.
The Theory of Human Mind
As far as we know, a theory of the human mind which
takes into account the empirical facts of cognitive
psychology does not exist.
The Impact of Knowledge Space
Theory on Memory Research
The impact of Knowledge Space Theory on memory
research methodology will be at least twofold,
regarding data analysis and validation of hypotheses
and models.
Data Analysis (an example)
Correlational analysis are very popular in psychology
and have been performed also in memory research, e.g.
recently by Kahana (2000). He recommends
contingency analyses of memory performance data
using Yule‘s Q-equation. The quasi-order used in KST
allows high or low contingency corresponding to
high or low Q-values, which in KST means
„equivalence“ or independence. A moderate Q-value
however can correspond to a strong dependency of the
surmise or prerequisite type.
Kahana for instance presents the following
contingency table for the number of items, which are
recognised and recalled by the same subjects or not.
Table 2: Contingency table for hypothetical data
(Kahana, 2000, p. 62)
Recognition Test
1 0
Recall 1 9 2
Test 0 9 6
1 = correct; 0 = incorrect
Correlation: A moderate relationship with Q = 0.5
From the viewpoint of Knowledge Space Theory
there exists a strong relationship between recall and
recognition in this case - with only two exceptions
because (1,0) = 2 instead of zero: From recalling an
item, recognising it can be surmised. Recognition and
recall are in a prerequisite relationship. For (1,0) = 0,
instead of two, the relationship would be perfect. This
is not captured by using Q-values.
Validation of Hypotheses/Models
Because of the one-to-one correspondence between
surmise relations and knowledge spaces, confidence
table data as well as the answer pattern of the
individual subjects can be used for validating
hypotheses and models. Very convenient is Schrepp’s
(1999) method to individualise the models. Creating a
variety of sub-models we expect different knowledge
states for the different subjects. These can be
compared with the observed answer patterns for
validating the models. This procedure can also be
applied for latencies. The models are explaining
individual differences and they can be tested taking
individual differences into account, which is seldom in
memory research.
The Impact of E-learning on Memory
Research Methodology
The impact of memory research on e-learning and that
of e-learning on memory research are in reciprocal
relationships: Memory research methodology can be
used for preparing and improving e-learning systems,
and internet-based e-learning can be a very useful
application of the results of memory research (see
above). Elearning systems, on the other hand, can
provide powerful research tools for applied memory
research in an ecological setting using field experi-
ments as well as prediction and control methods.
Applied Field Experiments
Internet experiments are already used in basic memory
research, see e.g.
exponnet.html. From a methodological point of view,
internet experiments have a lot of disadvantages, e.g.
the control of the experimental conditions is reduced,
and the sample of participants is not representative for
a population.
Even by planning the experiments according to
the state of the art, not all of the disadvantages can be
avoided. This is the price for an ecological setting.
Thus, applied internet experiments in the
e-learning setting are meaningful for verifying results
of lab experiments in the field and as a control before
applying the results of basic lab- or web-experiments
for improving the e-learning and e-teaching.
However internet based e-learning experiments
are, in case of applied memory research, based on
carefully designed memory lab experiments. Thus the
convergence of the results of both kind of experiments
are an important criterion for the successful
application of basic research.
Prediction and Control by Application
Results from basic research may be applied directly to
e-learning systems without prior applied field
experimentation. The models and experimental results
of basic memory research will be implemented
directly for improving the e-learning environment. The
convergence between the predicted and the observed
behaviour or performance of the student is the
criterion for successful application of basic research.
In computerbased learning and teaching this method
has a long tradition, see, e.g., the work of Patrick
Suppes in the 70s.
Even in case of discrepancies between predictions
and observations, the e-learning system can be
improved by adjustments of the models or their
parameters and the models can be improved, too! In
case of unexplainable discrepancies, basic research
will be stimulated. Memory psychology has the
chance to verify its models and results in an ecological
e-learning setting through field experiments and
control of predictions.
Concluding Remarks
One of the traditional and most elaborated fields in
cognitive psychology is the psychology of memory.
The substrate of learning is memory. Thus the theories
and results of memory psychology should be
systematically applied to create the e-learning systems
of the future.
On the other hand, besides using ecological
memory research for improving them, the e-learning
systems can be used for improving ecological memory
research, using them as a kind of world-wide learning-
and memory-research laboratory for testing and
validating the models of memory in ecological
The aim of teaching is to create knowledge, of
course. Thus the psychology of knowledge which is
already much closer to e-learning than memory
research has to be linked with the psychology of
memory for improving e-learning systems. Knowledge
Space Theory provides methods for data analysis,
validation of hypothesis, and models, which may
improve basic and applied memory research.
Some hurdles in applying memory research results
and models have to be overcome. Modern information
technology has to be used for storing the descriptions
of investigations and the data in computersearchable
and readable form for applications and model testing.
The relationships between the models in memory
psychology have to be analysed, as well between as
within the partial and the global models.
We would like to conclude this chapter by the
proposition that e-learning and Knowledge Space
Theory offer a lot of chances and challenges for
memory psychologists.
Albert, D. (Ed.) (1994) Knowledge Structures. Berlin,
Heidelberg, New York: Springer-Verlag.
Albert, D. & Held, T. (1994). Establishing knowledge spaces
by systematical problem construction. In D. Albert
(Ed.), Knowledge Structures (pp. 78112). New York:
Springer Verlag.
Albert, D. & Hockemeyer, C. (2002). Applying demand
analysis of a set of test problems for developing an
adaptive course. In Proceedings of the International
Conference on Computers in Education ICCE 2002.
Albert, D., Hockemeyer, C., & Wesiak, G. (2002). Current
trends in elearning based on knowledge space theory
and cognitive psychology. Psychologische Beiträge, 44,
Albert, D. & Lukas, J. (Eds.). (1999). Knowledge Spaces:
Theories, empirical research, applications. Mahwah,
NJ: LawrenceErlbaum Associates.
Albert, D. & Mori, T. (2001). Contributions of cognitive
psychology to the future of e-learning. Bulletin of the
Graduate School of Education, Hiroshima University,
Part I (Learning and Curriculum Development), 50,
Albert, D. & Stapf, K.-H. (Eds.). (1996). Gedächtnis.
Enzyklopädie der Psychologie, Themenbereich C, Serie
II, Band 4. [Memory. Encyclopedia of Psychology].
Göttingen: Hogrefe.
Bahrick, H.P. (2000) LongTerm Maintenance of
Knowledge. In E. Tulving, and F. I. M. Craik (Eds.),
The Oxford Handbook of Memory. (pp. 347362)
Oxford: Oxford University Press.
Bjork, R. A. (1994). Memory and metamemory:
considerations in the training of human beings. In J.
Metcalfe & A. Shimamura (Eds.), Metacognition:
knowing about knowing (pp. 185206): MIT Press.
Christina, R.W., & Bjork, R.A. (1991) Optimizing long-term
retention and transfer. In D. Druckman & R. A. Bjork
(Eds.), In the mind's eye: Enhancing human
performance (pp.2356). Washington, DC: National
Academy Press. (Also available online:; access
July 17, 2003).
Clark, R. C. & Mayor, R. E. (2002). E-Learning and the
Science of Instruction: Proven Guidelines for
Consumers and Designers of Multimedia Learning.
San Francisco: John Wiley and Sons.
Cosyn, E. & Thiéry, N. (2000). A practical procedure to
build a knowledge structure. Journal of Mathematical
Psychology, 44, 383407.
Doignon, J.-P. (1994). Knowledge spaces and skill
assignments. In G. H. Fischer & D. Laming (Eds.),
Contributions to Mathematical Psychology,
Psychometrics, and Methodology (pp. 111121). New
York : SpringerVerlag.
Doignon, J.P. & Falmagne, J.Cl. (1985). Spaces for the
assessment of knowledge. International Journal of
Man-Machine Studies, 23, 175196
Doignon, J.P. & Falmagne, J.Cl. (1999). Knowledge
Spaces. Berlin, Heidelberg, New York: Springer.
Düntsch, I. & Gediga, G. (1995). Skills and knowledge
structures. British Journal of Mathematical and
Statistical Psychology, 48, 9–27.
Held, T. (1999). An integrated approach for constructing,
coding, and structuring a body of word problems. In D.
Albert & J. Lukas (Eds.), Knowledge Spaces: Theories,
Empirical Research, Applications (pp. 67102).
Mahwah, NJ : Lawrence Erlbaum Associates.
Hockemeyer, C., Held, T., & Albert, D. (1998). RATH a
relational adaptive tutoring hypertext
WWWenvironment based on knowledge space theory.
In C. Alvegård (Ed.), CALISCE`98: Proceedings of the
Fourth International Conference on Computer Aided
Learning in Science and Engineering (pp. 417423).
Göteborg, Sweden: Chalmers University of
Kahana, M. J. (2000) Contingency analysis of memory. In E.
Tulving, and F. I. M. Craik (Eds.), The Oxford
Handbook of Memory. (pp. 59–72) Oxford: Oxford
University Press.
Korossy, K. (1993). Modellierung von Wissen als Kompetenz
und Performanz [Modeling knowledge as competence
and performance]. Inauguraldissertation, Fakultät für
Sozial-- und Verhaltenswissenschaften, Universität
Korossy, K. (1997). Extending the theory of knowledge
spaces: A competenceperformance approach.
Zeitschrift für Psychologie, 205, 53–82.
Lukas, J. & Albert, D. (1999). Knowledge structures: what
they are and how they can be used in cognitive
psychology, test theory, and the design of learning
environments. In D. Albert and J. Lukas, (Eds.),
Knowledge Spaces: Theories, Empirical Research,
Applications (pp. 3–12). Mahwah, NJ: Lawrence
Erlbaum Associates.
Neisser, U. (1978). Memory, What are the important
questions. In M. M. Grunenberg, P. E. Morris & R. E.
Sykes (Eds.), Practical aspects of memory (pp. 324).
London: Academic Press.
Offner, Max (1913). Das Gedächtnis. Die Ergebnisse der
experimentellen Psychologie und ihre Anwendung in
Unterricht u. Erziehung [Memory. Results of
experimental psychology and its applications in
Teaching and Education]. Dritte, vermehrte u. teilweise
umgearbeitete Aufl. Berlin: Reuther & Reichard.
Riley, K., Pavlidou, T., Wade, V., Conlan, O., Akeroyd, J.,
Sandford, P., Albert, D., Hockemeyer, C., Lefrere, P.,
Grant, J., Romano, L., Da Bormida, G.., Trabucchi, R.,
Cardinali, F., and Kay, R. (2001). EASEL D03
Requirements Specification, version 1.4. Online
available at
RequirementsSpecification.doc; access July 31, 2003.
Schrepp, M. (1999). An empirical test of a process model for
letter series completion problems. In D. Albert and J.
Lukas, (Eds.), Knowledge Spaces: Theories, Empirical
Research, Applications (pp. 133154). Mahwah, NJ:
Lawrence Erlbaum Associates.
Tulving, E. and Craik, F. I. M. (Eds.) (2000). The Oxford
Handbook of Memory. Oxford: Oxford University
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
examine 2 . . . contributors to nonoptimal training: (1) the learner's own misreading of his or her progress and current state of knowledge during training, and (2) nonoptimal relationships between the conditions of training and the conditions that can be expected to prevail in the posttraining real-world environment / [explore memory and metamemory considerations in training] (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Procedures which are to test a subject’s knowledge concerning a specific domain obviously require (in addition to other prerequisites) a set of problems.
Suppose that Q is a set of problems and S is a set of skills. A skill function assigns to each problem q i.e. to each element of Q — those sets of skills which are minimally sufficient to solve q; a problem function assigns to each set X of skills the set of problems which can be solved with these skills (a knowledge state). We explore the natural properties of such functions and show that these concepts are basically the same. Furthermore, we show that for every family K of subsets of Q which includes the empty set and Q, there are a set S of (abstract) skills and a problem function whose range is just K. We also give a bound for the number of skills needed to generate a specific set of knowledge states, and discuss various ways to supply a set of knowledge states with an underlying skill theory. Finally, a procedure is described to determine a skill function using coverings in partial orders which is applied to set A of the Coloured Progressive Matrices test (Raven, 1965).
The coherence principle is important because it is commonly violated, is straightforward to apply, and can have a strong impact on learning. This chapter summarizes the empirical evidence for excluding rather than including extraneous information in the form of background sound, added text, and added graphics. What is new in the chapter is some updating of the growing research base, but the main conclusion remains the same: Adding interesting but unnecessary materials to e-learning can harm the learning process. The chapter explores the merits of adding extra sounds, pictures, and words that are intended to make multimedia environments more interesting to the learner. The authors recommend avoiding adding extraneous sounds or music to instructional presentations, especially in situations in which the learner is likely to experience heavy cognitive processing demands. Much of the research reported in the chapter deals with short lessons delivered in a controlled lab environment. e-Learning