Learning with Media
Robert B. Kozma
University of Michigan
Abstract
This
article describes learning with media as a complementary process within which
representations are constructed and procedures performed, sometimes by the
learner and sometimes by the medium. It reviews research on learning with
books, television, computers, and multimedia environments. These media are
distinguished by cognitively relevant characteristics of their technologies,
symbol systems, and processing capabilities. Studies are examined that
illustrate how these characteristics, and instructional designs that employ
them, interact with learner and task characteristics to influence the structure
of mental representations and cognitive processes. Of specific interest is the
effect of media characteristics on the structure, formation, and modification of
mental models. Implications for research and practice are discussed
Do media influence learning? The research reviewed
in this article suggests that capabilities of a particular medium, in
conjunction with methods that take advantage of these, interact with and
influence the ways learners represent and process information, and may result
in more or different learning when one medium is used compared to another, for
certain learners and tasks.
This paper is in response to a challenge by Clark
(1983) that,"...researchers refrain from producing additional studies
exploring the relationship between media and learning unless a novel theory is
suggested." (p. 457) He extended this challenge after reviewing the
existing comparative research on media and concluding that, "...media do
not influence learning under any conditions." Rather, "...media are
mere vehicles that deliver instruction but do not influence student achievement
any more than the truck that delivers our groceries causes changes in our
nutrition." (p. 445) The theoretical framework supported by the current
review presents an image of the learner actively collaborating with the medium
to construct knowledge. It stands in vivid contrast to an image in which
learning occurs as the result of instruction being "delivered" by
some (or any) medium. The framework is meant to provide the novel approach
required by Clark before research on media and learning can progress.
In this theoretical framework learning is viewed as
an active, constructive process whereby the learner strategically manages the
available cognitive resources to create new knowledge by extracting information
from the environment and integrating it with
information
already stored in memory. This process is constrained by such cognitive factors
as the duration and amount of information in short-term memory, the task -
relevant information that is available in long-term memory, how this
information is structured, the procedures that are activated to operate on it,
and so on. Consequently, the process is sensitive to characteristics of the
external environment, such as the availability of specific information at a
given moment, the duration of that availability, the way in which it is
structured, the ease with which it can be searched, and so on.
The subdomain of the external environment examined
in this paper is "mediated information," not only that which is
intentionally educational (such as a computer-based lesson) but other
information embedded in books, television programs, etc. Not directly addressed
by this review is information embedded in what are sometimes called
"authentic situations" (Brown, Collins, and Duguid, 1989), though the
thesis developed in this paper complements learning in such situations. Nor
does the article examine the larger social environment within which mediated
interactions occur (Perkins, 1985). Ultimately, it may be these contexts, and
the ways media are integrated into them, that have the greatest impact on how
people think and learn. While these broader contexts will be referenced from
time to time, the primary focus of this paper is finer grained: specific
episodes within which a learner interacts with mediated information to influence
learning.
In support of the thesis stated above, this article
will provide a definition of media and use it to examine the theoretical and
research literature on learning from books, television, computers, and
multimedia environments. Each section will examine how the complementary
construction of representations, and operations performed on them, is
influenced by characteristics of the medium, designs that take advantage of
these, and the characteristics of learners and tasks. The intent is to demonstrate
the relative cognitive effects of learning with different media, particularly
effects related to the structure, formation, and modification of mental models.
Media Defined
Media can be defined by their technology, their
symbol systems, and their processing capabilities. The most obvious
characteristic of a medium is its technology, the mechanical and electronic
aspects that determine its function and to some extent its shape and other
physical features. These are the characteristics that are commonly used to
classify a medium as a "television," a "radio," and so on.
The cognitive effects of these characteristics, if any, are usually indirect.
Characteristics such as size, shape, and weight makes it more likely that a
student will learn with a book while on a bus but not a computer, though of
course this is changing as computers get smaller, lighter, and cheaper. On the
other hand, some cognitive effects of technology are more direct. For example,
the size and resolution of many computer screens is such that reading text may
be more difficult than it is with books (Haas, 1989).
However, the primary
effect of a medium's technology is to enable and constrain its other two
capabilities: the symbol systems it can employ and the processes that can be
performed with it. For example, a computer with a graphics board or a speech
synthesis board can use different symbols in its presentations than those
without. Computers with enough memory to run LISP and expert systems can
process information in different
2
ways
than those without. Symbol systems and processing capabilities have a number of
implications for learning.
Salomon (1974, 1979) describes the relationship
between a medium's symbol systems and mental representations. Symbol systems
are "modes of appearance" (Goodman, 1976), or sets of elements (such
as words, picture components, etc.) that are interrelated within each system by
syntax and are used in specifiable ways in relation to fields of reference
(such that words and sentences in a text may represent people, objects, and
activities and be structured in a way that forms a story) . A medium can be
described and perhaps distinguished from others by its capabilities to employ
certain symbol systems. Thus, television can be thought of as a medium that is
capable of employing representational (i.e., pictorial) and audio- linguistic
symbol systems (among others). Such characterizations can also be used to
specify a certain overlap or equivalence of media. Thus video and motion film
can be thought of as equivalent in this regard, while they can be distinguished
from radio which can employ only a subset of these symbol systems.
Salomon (1974, 1979) suggests that these
characteristics should be used to define, distinguish, and analyze media since
they are relevant to the way learners represent and process information from a
medium. He contends that certain symbol systems may be better at representing
certain tasks and that information presented in different symbol systems may be
represented differently in memory and may require different mental skills to
process. The research reviewed here supports and elaborates on this contention.
For example, studies will be examined that illustrate how symbol systems
characteristic of certain media can connect mental representations to the real
world in a way that learners with little prior knowledge have trouble doing on
their own, without the representation of information in these symbol systems.
But, as will be demonstrated, symbol systems alone
are not sufficient to describe a medium and its cognitive effects. Information
is not only represented in memory, it is processed. Media can also be described
and distinguished by characteristic capabilities that can be used to process or
operate on the available symbol systems. Thus, information can be searched or
its pace of progression changed with video disc in a way that is not possible
with broadcast video. Including processing attributes in the definition of
media can create useful distinctions between videodisc and broadcast video,
even though they both have access to the same symbol systems. Computers are of
course especially distinguished by their extensive processing capabilities,
rather than by their access to a particularly unique set of symbol systems.
The
processing capabilities of a medium can complement those of the learner; they
may facilitate operations the learner is capable of performing or perform those
that the learner cannot. As Salomon (1988) points out, if such processes are
explicit and fall within what Vygotsky (1978) calls the "zone of proximal
development," the learner may come to incorporate them into his or her own
repertoire of cognitive processes. This review will examine research which
illustrates how the processing capabilities of certain media can modify and
refine the dynamic properties of learners' mental models.
3
However, it is important to remember that while a
medium can be defined and distinguished by a characteristic cluster, or
profile, of symbol systems and processing capabilities, some of these
capabilities may not be used in a particular learning episode (Salomon and
Clark, 1977). For example, a particular video presentation may use few or no
representational symbols (e.g., a "talking head" presentation). Or, a
viewer may allow a video disc presentation to play straight through and not use
the available search capabilities. In these cases a "virtual medium"
is created that consists of the profile of symbol systems and processing
capabilities that were actually used during the session: a television becomes,
in effect, a radio; a video disc player becomes broadcast television. It is
only the capabilities of the virtual medium that can be expected to have an
effect on learning processes and outcomes.
Whether or not a
medium's capabilities make a difference in learning depends on how they
correspond to the particular learning situation--the tasks and learners
involved-- and the way the medium's capabilities are used by the instructional
design. Tasks vary in their situational characteristics and the demands they
place on the learner to create mental representations of certain information
and operate on that information in certain ways. Learners vary in their
processing capabilities, the information and procedures that they have stored
in long-term memory, their motivations and purposes for learning, and their
metacognitive knowledge of when and how to use these procedures and
information.
Many learners, perhaps most, can and frequently do
supply useful representations and operations for themselves from the information
externally available, regardless of medium used. On the other hand, learners
will benefit most from the use of a particular medium with certain capabilities
(as compared to the use of a medium without these), if the capabilities are
employed by the instructional method to provide certain representations or
perform or model certain cognitive operations that are salient to the task and
situation, and which the learners can not or do not perform or provide for
themselves. These representations and operations, in turn, influence problem
solving and the ability to generate and use representations in subsequently
encountered situations. This view of learning with media as a continuous,
reciprocal interaction between person and situation--between learner and
mediated information--is compatible with evolving aptitude-treatment
interaction theory (Snow, 1989).
Learning with
Books
The most common medium encountered in school
learning is the book. As a medium, books can be characterized by the symbol
systems they can employ: text and pictures. The following sections of the
review will examine the cognitive processes used in processing text and text
along with pictures. They will discuss how a distinctive characteristic of this
technology --its stability--influences the processing of these symbol systems
to construct knowledge representations and how these, in turn, are influenced
by individual differences of learners, primarily differences in their prior
domain-knowledge. The summary will describe how these processes and structures
can be supported by by the author when designing a book.
The reading processes and the stability
of the printed page. The primary symbol system used in
books consists of orthographic symbols which, in Western culture, are words
composed of phonemic graphemes, horizontally arrayed from left to right. That
4
this
arrangement is stable distinguishes text in books from other technologies that
use the same symbol system, for example the marquee on Times Square. This
stability also has important implications for how learners process information
from books. Specifically, the stability of text aids in constructing a meaning
of the text.
Learning with text involves the construction of two
interconnected mental representations: a textbase and a situation model
(Kintsch, 1988, 1989). The textbase is a mental representation derived directly
from the text, both at the level of micro- and macrostructure; it is a
propositional representation of the meaning of the text. While progressing
through the text, the reader assembles the propositions and integrates them
with ones previously constructed. As memory limits are reached, the most recent
and most frequently encountered propositions are retained in short- term memory
and held together by repetition or the embedding of arguments (Kintsch and van
Dijk, 1978). The reader generalizes from these local propositions to form
macropropositions, or summary-like statements that represent the gist of the
text. Integrating the information from the text in this way increases the
likelihood that it will survive in short-term memory and be fixed in long-term
memory.
The situation model is a mental representation of
the situation described by the text (Kintsch, 1988, 1989). While the textbase
is propositional, the situation model can be constructed from propositions or
spatial information. The situation model is connected to and constructed from
information in the text base and from knowledge structures evoked from long-
term memory by information appearing early in the text or that activated by the
reader's purpose. These structures, called variously schemata (Anderson, Spiro,
& Anderson, 1978), frames (Minsky, 1975) and scripts (Schank & Abelson,
1977), can be characterized as a framework with a set of labelled slots in
which values are inserted for particular situations. These structures serve two
related purposes: they provide a "scaffold" upon which the situation
model is constructed from the textbase, and they provide default values so that
the reader can make inferences about the local situation that were not
explicitly mentioned in the text. Learning from text involves the integration
of these representations into the comprehender's knowledge system by updating
the schemata currently in long-term memory or by constructing a new schema for
an unfamiliar situation.
But, what does any of this have to do with media?
How does this symbol system influence mental representations and cognitive
processes in distinctive ways? And why would learning processes and outcomes be
any different for books, which store orthographic symbols in a fixed, stable
way, and another medium, say audiotape or lecture, which may convey the same
linguistic information but in a different symbol system and in a transient way
(i.e., speech)?
In many situations for fluent readers, reading
progresses along the text in a forward direction at a regular rate and the
information could just as well be presented in another, more transient medium.
But on occasion, reading processes interact with prior knowledge and skill in a
way that relies heavily on the stability of text to aid comprehension and
learning.
5
In the obvious case, the
effort required of poor readers to decode the text draws on cognitive resources
that would otherwise be used for comprehension, thus increasing the risk of
comprehension or learning failure (LaBerge and Samuels, 1974). But even fluent
readers may have difficulty with longer or novel words, such as technical terms
in an unfamiliar domain. In both of these cases, readers will use the stability
of text to recover from comprehension failure. When encountering difficulty, a
reader will slow their rate, making more or longer eye fixations (Just and Carpenter,
1987) or they may regress their eyes, going back to review a word as an aid to
retrieving a meaning for it from memory (Bayle, 1942). Alternatively, a reader
may retrieve several meanings for a word and may make longer or additional
fixations, or may regress over a phrase, a clause, or even a sentence to
determine which is appropriate for a given context (Just and Carpenter, 1987;
Bayle, 1942). Such difficulties might arise from unusual syntactic structures
(e.g., "The thief stood before the blackrobed judge entered the
courtroom.") or difficulties in interpreting combinations of words to
construct local propositions. Readers will slow their rate for a passage on a
difficult or novel topic (Buswell, 1937), when encountering information within a
passage that is particularly important to the meaning of the text (Shebilske
and Fisher, 1983), or when they must integrate less well organized sentences
into macropropositions (Shebilske and Reid, 1979).
All of these are examples of how readers use the
stability of the symbol system in books to slow their rate of progression or
even to regress over text in a way that would seem difficult or impossible to
do with audiotape's ever-advancing presentation of information. However, this
distinction is likely to be crucial only in certain situations. For example,
readers in the Shebilske and Reid study (1979) reduced their rate from 302
words per minute to 286. While statistically significant, this difference may not
have practical significance with regard to media use since the typical
audiotape presentation rate of 110-120 words per minute would seem to be slow
enough to accommodate these comprehension difficulties. Even the apparent
inability to regress over speech might be accommodated by the two second
duration of information in acoustic memory (Baddeley, 1981) that would allow a
listener to "recover" the three or four most recently spoken words
and achieve the same affect as regression over text. The clearest advantage to
the use of the stability of text to aid comprehension is when the reader must
regress over segments of information larger than a phrase.
Perhaps more important than the use of the stability
of text to recover from local comprehension failure in novel or difficult
situations is its use in conjunction with highly developed reading skills (such
as those described by Brown, 1980) and elaborated memory structures to
strategically process large amounts of text within very familiar domains. This
is most dramatically illustrated in a study by Bazerman (1985), who interviewed
seven professional physicists and observed them reading professional material
in their field. These readers read very selectively, making decisions based on
highly developed schemata that extended beyond extensive knowledge of accepted
facts and theories in the field to include knowledge about the current state of
the discipline and projections of its future development, as well as personal
knowledge and judgments about the work of colleagues. Readers used this domain
knowledge to serve their reading purposes. Most often their interests were to
find information that might contribute to their immediate research goals or to
expand their background knowledge of the field, and they made their selections
based on these purposes.
6
Bringing
schemata and purposes to bear, these subjects would typically read by scanning
rapidly over tables of contents, using certain words to trigger their attention
and question a particular title more actively. If a particular term attracted
their attention, they would look at other words in the title with the result
that about two-thirds of the titles more closely examined were subsequently
rejected based on this additional information. If even more information was
needed to make further selections, they would turn to the abstract.
Having identified an article of interest, they would
read parts of it selectively and non-sequentially, jumping back and forth,
perhaps reading conclusions then introductions, perhaps scanning figures, and
reading those sections more carefully that fit their purpose. If an article did
not readily fit with their comprehension schemata, the readers would weigh the
cost of working through the difficulty against the potential gain relative to
their purposes. If they chose to read through a difficult article or portion,
they would occasionally pause at length to work through the implications of
what had been read, or read it through several times. They might also look up
background material in reference works and textbooks.
The studies above show
the range of ways that readers take advantage of the stable structure of text
to aid comprehension. In the Bazerman (1985) study, strategic readers with
considerable domain knowledge would sometimes progress through the text at a
rapid rate, using a single word to skip a vast amount of information. Other
times, they would slow considerably, moving back and forth within a text and
across texts, to add to their understanding of the field. In other studies
(Bayle, 1942; Shebilske and Reid, 1979), readers encountering difficulties with
unfamiliar words, syntactic structures, or ideas used the stability of the
printed page to slow their rate and regress over passages. None of these
processing strategies are available with the transient, linguistic information
presented in audio tape or lectures.
Multiple symbol systems: Learning with
text and pictures . Orthographic symbols are, of course,
not the only ones available to books. Pictures and diagrams are used in books
from primers to college textbooks to technical manuals. But, how do readers use
pictures? What is the cognitive effect of pictures in combination with text?
And, how does the stability of these symbols, as presented in books, influence
this process compared to another medium, say television, which presents
linguistic and pictorial symbols in a transient way? This last, comparative
issue will be directly addressed in the subsequent section on learning with
television. The following section examines the cognitive effects of pictures
and text.
There is a large body of comparative research on
learning from text with and without pictures. Almost all of the studies examine
only the impact on cued recall and use traditional experimental designs, of the
type criticized by Clark (1983). However, there is consensus among the reviews
of this research that pictures have positive effects under certain conditions.
Pressley (1977), Schallert (1980), and Levie and Lentz (1982) generally concur
that the use of pictures with text increases recall, particularly for poor
readers, if the pictures illustrate information central to the text, and when
they represent new content that is important to the overall message, or when
they depict structural
7
relationships
mentioned in the text. The problem with this type of research is that it does
not reveal the mechanism by which pictures and text influence the learning
process.
The four studies below examine processes of
comprehension and learning with text and pictures. In brief, it appears that
the use of both symbol systems facilitates the construction of the textbase and
the mapping of it onto the mental model of the situation. This is particularly
facilitative for learners who have little prior knowledge of the domain.
A study by Rusted and Coltheart (1979) examined the
way good and poor fourth grade readers used pictured text to learn about
physical features, behavior, and habitat of unfamiliar animals. Including
pictures of animals in their environments along with the text resulted in
better retention by both good and poor readers, over the use of text alone. It
facilitated retention of all information by good readers, but only pictured
information (i.e., recall of physical features) for poor-readers. Observations
of good readers showed that they spent time initially looking at the pictures
and rarely looked at them once they started reading. Poor readers, on the other
hand, frequently moved back and forth between text and pictures. While the
process data are not detailed enough to be definitive, they suggest that good
readers may be using the pictures to evoke an "animal schema" that
guided their reading and aided comprehension. On the other hand, poor readers
frequently moved back and forth between text and pictures, maybe to facilitate
the decoding of particular words and perhaps to aid in building a mental model
of these unfamiliar animals and their habitats.
Stone and Glock (1981) obtained similar findings,
using more precise measures, when they examined the reading of second and third
year college students. Subjects used either text without pictures, pictures
without text, or pictured text to learn how to assemble a toy push- cart. The
text-only group made significantly more assembly errors, particularly errors of
orientation, while the text and picture group was most accurate in their
constructions, making only 18% of the errors of the text-only group. Eye-tracking
data indicated two patterns of picture use. Readers would typically spend the
first few seconds examining the picture. Subsequently they would look from text
to picture as they progressed through the passage, spending an average of more
than eighty per cent of their time looking at text rather than pictures. As in
the Rusted and Coltheart study, the data suggest that readers initially use the
pictures to evoke a schema that serves as a preliminary mental model of the
situation. Subsequently, it seems that the text carries the primary semantic
message, while the pictures are used to map this information on to this
preliminary mental model, elaborating on the components of the push cart and
their relative arrangement.
The usefulness of pictures seems to interact with
specialized ability or domain knowledge. In a study by Hegarty and Just (1989)
college students were tested on mechanical ability and assigned to either a
"short text" or a "long text" describing a pulley system.
The short text merely named the components of the system and described how it
operated. The long text also elaborated on the arrangement and structure of the
components in the system. All texts were accompanied by a schematic diagram of
the pulley system. Precise eye-fixations measured the number and duration of
movements back-and-forth between particular words in the text and specific
locations in the diagram.
8
There
was a non -significant interaction such that low ability students spent more
time than high ability students looking at the diagram when it accompanied the
longer text which described the relationship among the components of the
system. The high ability students spent more time examining the diagram with
the shorter text. The results suggest that people low in mechanical ability
have difficulty forming mental models of mechanical systems from text and use
diagrams to help them construct this representation. People with high
mechanical ability seem to construct this model from prior knowledge and
information from text, without need to refer to the picture. On the other hand,
these high ability students are better able to encode new information from a
diagram when the text does not describe all the information relevant to
understanding a mechanical system.
In a study by Kunz, et
al. (1989), university students majoring in either geography or social science
read passages that contained concepts and rules on meteorology. They received
text either with or without two types of supplements: 1) representational pictures
depicting spatial arrangement, appearance, and configuration of clouds, and 2)
a tree diagram that provided an overview of the main concepts, constituting the
macrostructure of the text. Students were divided on prior domain knowledge.
For students with a higher prior knowledge, the examination of representational
pictures did not correlate with post-task comprehension and the use of the tree
diagram correlated negatively with performance. In contrast, subjects with low
prior knowledge did better if they both inspected the representational picture
very often and spent some time examining the tree diagram. These data suggest
that students with little prior knowledge benefit most from the pictures and
the tree diagram. On the other hand, students with sufficient prior domain
knowledge can rely instead on their own, well-developed mental models to aid
comprehension. Indeed, the tree diagram used in this study may have conflicted
with the idiosyncratic way these students have domain knowledge structured and
it may have actually interfered with their comprehension.
These studies may also explain the conclusion of
Pressley (1977) in his review of studies of text and imaging. He found that
learners who do not receive pictures but are instead instructed to generate
images during the processing of story prose, recall as much as those who
receive pictures, and more than those who neither receive pictures or are
instructed to generate images. However, there were developmental differences
such that children of eight years and older could gainfully generate and use
images during text processing, while those under the age of six appeared unable
to generate useful images in response to text, even when directed to do so. In
these studies, age may be a surrogate measure for accumulated world knowledge
that allows older children to generate mental models which supplement the text
and aid comprehension and recall. Younger children may not have sufficient
world knowledge to generate such mental models and they benefit most from
pictures to aid this process.
Greeno (1989) elaborates on the "situation
model" in a way that can be useful in analyzing the relationship between
text, pictures, cognitive structures, and processes. Greeno proposes a
theoretical framework that defines knowledge as a relationship between an
individual and a social or physical situation, rather than as a property of an
individual only. It extends the information processing paradigm, which focuses
primarily on internal structures and process, to include structures and
processes external to the
9
learner.
This relativistic notion of knowledge depends heavily on a model of the
situation and has considerable implications for learning with media.
In this framework, objects and events organized in
relation to human activities (such as hitting a ball or buying and selling
merchandise), as well as related abstractions (such as "force" or
"profit margin"), are expressed within our culture in various
symbolic notations and structures (verbal descriptions, diagrams, graphs etc.).
Mental representations, or mental models, are derived from the symbolic
structures and
correspond
to real world situations --objects, events, and their abstractions.1
These mental models consist of symbolic objects, or mental entities, that may
have properties associated with the symbol systems from which they were derived
(e.g., arrows representing force vectors), as well as properties of objects in
situations that the symbolic structures represent (e.g., balls moving through
space and time along certain trajectories). Greeno contends that people can
reason in this mental space to solve problems by operating on these symbolic
objects in ways that correspond to operations in real situations.
However
too often in school learning, these mental objects and operations have little
correspondence to real world objects, events, and their abstractions and
instead map only onto the symbolic domains from which they were derived. The
research above suggests that for some learners, the use of pictures, in
addition to text, may provide information needed to map mental representations
derived from the text onto mental representations of the real world, perhaps
because pictorial symbol systems share more properties with this domain.
Summary and implications.
We now have a picture of learning with books that illustrates the relationship
between human information processes and the characteristic stability and symbol
systems of the medium. Readers move along a line of text constructing a
representation of the the textbase. They build a mental model of the situation
described with information from the textbase and schemata activated in
long-term memory. They slow down to comprehend difficult or important points,
stop or regress to retrieve the meaning of an unfamiliar word or a confusing
clause or sentence. They may also use their knowledge of the domain along with
highly developed strategies to read very selectively in service of a particular
purpose they bring to the task. They use titles and abstracts to skip sections
or entire articles, or to focus in on sections of interest. They read
summaries, then overviews, reread portions, and move back and forth between
texts.
If
a picture is available, they may refer to it to supplement the text. An initial
look at the picture will evoke domain knowledge, for those that have it. In a
less familiar domain, readers will move back and forth frequently between text
and picture to clarify the meaning of a word or construct or elaborate on a
model of the situation. All of these strategies and their resulting mental
representations are influenced by the knowledge and purpose the reader brings
to the task, the symbol systems and stability of code that characterizes the
book.
An author can use these capabilities in a way that
complements the learner's skills and deficiencies. Text authors can use the
stability of text and pictures in books and
10
knowledge
of comprehension processes to design structures within their texts that support
and facilitate learning. Such structures may include titles (Bransford &
Johnson, 1973), post-questions (Wixson, 1984), explicitly stated behavioral
objectives (Mayer, 1984), cohesive text elements (Halliday and Hasan, 1976),
signals (Meyer, 1975; Mayer 1984), and so on. For example, in the Bransford and
Johnson (1973) study, one group of students had considerable difficulty
comprehending a paragraph even though it was linguistically simple and
contained no difficult words, constructions, or complex concepts. A second
group was presented the same paragraph, but this time the paragraph was
preceded by a title. In this second condition, the subjects rated the paragraph
as more comprehensible and they recalled it better. Presumably, the title
evoked an appropriate schema that allowed the readers to supply information not
explicit in the paragraph but was important for its comprehension. Other text
strategies might evoke different reading processes, such as conducting backward
reviews to facilitate retention (Wixson, 1984), or focus attention on certain
types of information, or build internal connections among concepts in the text
(Mayer, 1984). Such devices designed into the text can reduce the need for the
regressive strategies observed in the Bayle (1942) and Shebilske and Reid
(1979) studies, and support the purpose and schema-driven strategies evident in
the Bazerman study (1985), at least for students with sufficient prior
knowledge.
An understanding of the cognitive function of
pictures can also inform instructional practice. This can provide text authors
with information that can be applied heuristically to identify situations where
pictures would be useful, as well as how they might be usefully designed to
accommodate particular learners and tasks (Winn, 1989). Such guidelines may
suggest the positioning of pictures in the text, the degree of realism, the use
of arrows and other highlighting mechanisms, and so on. For example, the
research above suggests that for knowledgeable readers, pictures should be
placed early in the text if they are used at all. On the other hand, a less
knowledgeable readership would benefit from interspersed pictures, juxtaposed
with their corresponding text. Winn (1989) reviews research that suggests that
the use of arrows to highlight critical attributes of objects can facilitate
subsequent identification, while the inclusion of details in an illustration
can actually interfere with the learning of an object's structure or function.
Learning with
Television
Television differs in several ways from books that
may affect cognitive structures and processes. As with books, television can
employ pictures, diagrams, and other representational symbol systems but in TV
these symbols are transient and able to depict motion. While linguistic
information in television can be orthographic, more often it is oral and, as
with audiotape and radio, transient. Because in television, linguistic and
pictorial symbol systems are transient and because they are presented
simultaneously, it is possible that viewers process this information in a very
different way than the back-and- forth serial processing of linguistic and
representational information in books. It is also possible that the symbol
systems used and their transient nature affects the mental representations
created with television.
Television's window of cognitive
engagement. While popular notions of TV viewing portray
children as staring zoombie-like at the screen, reality is much different.
11
When
alternative activities are available, children generally look at and away from
the TV between one and two hundred times an hour (Anderson & Field, 1983).
Visual attention increases from very low levels during infancy to a maximum
during the late elementary school years, declining somewhat during adulthood
(Anderson, et al., 1986). Although the median look duration is usually only
several seconds, extended episodes as long as a minute are not rare. Looks as
long as ten minutes are exceptional. This discontinuous, periodic attention to
a medium whose information streams by ceaselessly has important implications
for comprehension and learning.
Research indicates that visual attention is
influenced by several factors. One set of factors, termed "formal features"
by Huston and Wright (1983), includes the use of different types of voices
(e.g., children, adult male, female), laughing, sound effects, music of
different types, animation, cuts, zooms, pans, etc. While children's
moment-to-moment visual attention may wander from the set, evidence suggests
that they continually monitor the presentation at a superficial level, such
that their visual attention is recaptured by certain audio cues. Features that
are associated with the onset of visual attention are women's and children's
voices, laughter, peculiar voices, sound effects, auditory changes, and visual
movement (Anderson, et al., 1979). Features associated with continued viewing
are special visual effects, pans, and high physical activity. The offset of visual
attention among children frequently corresponds to the use of men's voices,
long zooms, and inactivity.
While this image of visual attention seems bottom-up
and data-driven, other evidence suggests that these formal features come to be
seen by children as corresponding to the presentation of more or less
meaningful content and it is this second factor, the meaningfulness or
comprehensibility of the presentation, that guides visual attention. For
example, Anderson, et al. (1981) found that visual attention to segments of Sesame
St. was greater for normal segments than for the same visual presentation
for which comprehensibility was experimentally reduced by using backward
speech or a foreign language. Anderson and Lorch (1983) hypothesize that
through extensive viewing experience, children come to acquire knowledge about
the associations between the typical use of various formal features and the
likelihood that the corresponding content will be meaningful and interesting.
For example, men's voices may be perceived as generally corresponding to
adult-oriented content which is less comprehensible and less interesting to
children, and thus male voices do not recruit their visual attention.
Huston
and Wright contend that this comprehensibility influences attention in an
inverted-U relationship, such that content which is very simple or very
difficult to comprehend maintains attention less well than content in an
intermediate range of difficulty. This creates a window of cognitive
engagement, one that is perhaps different for each viewer. Yet within this
window, Huston and Wright (1983) conclude that visual attention is necessary
though not sufficient for comprehension; the depth of comprehension varies.
Salomon (1983) introduces the construct of
"amount of invested mental effort", or AIME, to account for the
difference between what is viewed and the depth of comprehension. AIME
distinguishes the "deep," effortful, nonautomatic elaboration of
encountered material from the "mindless" or "shallow"
processing of information that
12
results
in less learning. AIME is in turn influenced by several factors: One is the
attitudes people have about the amount of effort required to process a medium's
messages; the other is the purpose that people bring to the task.
Salomon (1984) found that a sample of sixth graders
rated TV as an "easier" medium from which to learn than books. When
assigned to view comparable stories from television or print, the effort spent
on learning reported by the reading group was significantly greater than that
reported by the group that viewed the television program. While both groups
scored the same on a test of factual recognition, the print group scored higher
on a test of inferences based on the story.
Krendl and Watkins (1983) exposed fifth-grade
children to a 15 minute educational television program. They manipulated the
purpose of viewing by telling half of the students to watch it for
entertainment purposes; the other half were told that it was an educational
program and they should watch it in order to answer questions. While recall of
the storyline was the same for both groups (i.e., number of recalled actions,
facts, scenes, etc.), the group instructed to view the program for educational
purposes responded to the content with a deeper level of understanding; that
is, they reported more story and character elements and included more
inferential statements about the meaning of the show.
These studies suggest that the perceptions students
have about a medium and the purposes they have for viewing influenced the
amount of effort that they put into the processing of the message and,
consequently, the depth of their understanding of the story. The following
sections elaborate on the cognitive mechanisms involved in effortful learning
with television, and examine the interaction of these processes with the
characteristics of the medium. Three issues related to the processing of
televised information are examined: the relationship between simultaneously
presented auditory and visual information, the processing pace of transient
information, and the use of such transient presentations to inform the
transformation functions of mental models. For the first of these issues, there
is now a considerable amount of cognitive research available; However, there
remains little research on the other two issues.
The simultaneous processing of two
symbol systems. An important attribute of video is its
use of both auditory and visual symbol systems. Within the window of cognitive
engagement, how do these symbol systems work, independently and together, to
influence comprehension and learning with television? Can either symbol system
convey the meaning of a presentation? Does the presentation of both at the same
time inhibit or facilitate learning?
Baggett (1979) found that either pictorial or
linguistic symbol systems alone can carry semantic information, such as a story
line. In this study, college students were presented with either a dialogless
movie, The Red Balloon, or an experimentally derived, structurally
equivalent audio version. They wrote summaries of episodes within the story
either immediately after the presentation or after a week delay. An analysis of
the summaries by trained raters found that those written immediately after
viewing the dialogless movie were structurally equivalent to those written
immediately after listening to the story. While subjects could construct a
semantic macrostructure (i.e., summary)
13
from
either medium, information obtained visually was more memorable. Summaries
written a week after viewing the movie were judged to be more complete than
those written a week after listening to the audio version.
While meaning can be
conveyed by either symbol system, Baggett (1989) concludes that information
presented visually and linguistically are represented differently in memory.
She contends that visual representations contain more information and are
"bushier". Whereas the statement "red leaf" contains only
the name of an object and a modifier, a mental representation of a "red
leaf" obtained from a picture carries with it information about size,
color, shape, etc. Also, the visual representation has more "pegs"
which can be used to associate it with information already in long-term memory.
These additional associations also make it more memorable.
But, it is a significant attribute of
video that the auditory and visual symbol systems are presented simultaneously.
How does a viewer process information from both of these sources? Two basic
hypotheses exist. One possibility is that the simultaneous presentation of
audio and visual information competes for limited cognitive resources and this
competition actually reduces comprehension. Another possibility is that
information presented with these two symbol systems may work together in some
way to increase comprehension.
A number of studies have compared a video program
with its decomposed audio and visual presentations to determine the role of
these two sources of information, individually and together (Meringoff, 1982;
Nugent, 1982; Baggett and Ehrenfeucht, 1982, 1983; Beagles-Roos and Gatt, 1983;
Gibbons, et al., 1986; Hayes and Kelly, 1984; Hayes, Kelly, and Mandel, 1986
Meringoff, 1982; Pezdek and Hartman, 1983; Pezdek, Lehrer, and Simon, 1984; and
Pezdek and Stevens, 1984) . In none of these studies did the combination of
audio and visual information result in lower recall than recall from either
source alone. In most of these studies, the combined use of visual and auditory
symbol systems resulted in more recall than visual-only and audio-only
presentations. This compels the rejection of the hypothesis that simultaneous
presentation of audio and visual information necessarily compete for cognitive
resources at the expense of comprehension.
Several of these studies used multiple measures of
recall to trace the symbol system source of different kinds of knowledge. In a
1982 study, Meringoff asked 9 and 10 year old children to draw and talk about
their imagery and to make and substantiate inferences about a story, The
Fisherman and His Wife. Compared to those who heard the story, the children
who saw the video drew more details and their pictures were more accurate.
Children in the audio groups based their inferences about details on previous
knowledge and personal experiences (more like those of children in the control
group unexposed to the story) and they were frequently in error relative to the
verbal descriptions. Beagles-Roos and Gat (1983) compared animated and audio-
tape presentations of two stories to groups of 1st and 4th grade children.
These researchers found that the explicit story content was learned equally
well by both treatment groups. The visual groups recalled more details from the
story, did better at a picture sequencing task, and based their inferences on
depicted actions. The audio groups more frequently
14
retold
the stories using expressive language, and based their inferences on verbal
sources and prior knowledge.
While people can
construct a mental representation of the semantic meaning of a story from
either audio or visual information alone, it appears that when presented
together each source provides additional, complementary information that
retains some of the characteristics of the symbol system of origin. Children
recall sounds and expressive language from the audio track and visual details
from the visual track. It also appears that the "bushier" nature of
representations derived from the visual symbol systems are better for building
mental models of the situation than are representations based on
audio-linguistic information. Students listening to an audiotape are more
likely to get information for this model from memory. While audio may be
sufficient for those knowledgeable of a domain, visual symbol systems supply
important situational information for those less knowledgeable.
These results parallel
those for text and pictures. However, the processing of text appears to be
driven by the construction of a representation of the linguistic information.
Comprehension of video appears to be driven by the processing of visual
information.
This
is apparent from a study by Baggett (1984), who varied the temporal order of
audio and visual information within a video presentation on the names and
functions of pieces of an assembly kit. In this study, the narration was
presented in synchrony, or 21, 14, or 7 seconds ahead of or behind the visual
presentation. College students performed best on immediate and 7 day delayed
tests of recall in the synchrony and the 7 second visual-then-audio group. The
worst performance was by groups with the audio presented first. This suggests
that in a video presentation, the visual symbol system serves as the primary
source of information and the audio symbol system is used to elaborate it.
The processing of transient information.
Another important characteristic of television is that the information it
presents can be, and usually is, transient. Comprehension is affected by the
pace of this presentation and by its continuity. Wright, et al. (1984) used
sixteen, 15 minute long children's television programs that varied in pace and
continuity. Pace was defined by these researchers as the rate of scene and
character change. Low-continuity programs were those whose successive scenes
were independent and unconnected (i.e., magazine formats). High-continuity
programs where those with connected scenes (i.e., stories). These programs were
shown to groups of elementary school children whose recall was measured using
seriation tasks of still pictures from the shows. The children who viewed slow
paced and on high-continuity programs performed better on these tasks. The
effect was additive for younger children.
While surprisingly little research has been done on
the effect of pace on comprehension, this is a potentially crucial variable
which distinguishes the process of learning with television, and other
transient media, from learning with stable media, such as text. Wright, et al.
(1984) defined pace as a characteristic of the presentation - the amount of
information presented per unit time (i.e., scene and character changes).
But from a cognitive perspective, the critical consideration is cognitive pace,
the amount of information processed per unit time. From this
perspective, the hypothetical unit of information is the "chunk." The
chunk is a semi-elastic unit whose size depends on the familiarity and
meaningfulness of the information (Miller, 1956; Simon, 1974). A single
15
word
may be a chunk in the following list of words: Lincoln, calculus, criminal,
address, differential, lawyer, Gettysburg. As rearranged into Lincoln,
Gettysburg, address, criminal, lawyer, differential, calculus; the chunk might
be larger than a word (e.g., "Lincoln's Gettysburg address"), but
only if this phrase had some meaning in long-term memory. Simon (1974) examines
the results of several experiments to conclude that the capacity of short-term
memory is five to seven chunks. He also concludes that it takes between five to
ten seconds to fixate each chunk in long-term memory. Thus, while the amount of
time it takes to process information is relatively constant (i.e., one chunk
per 5 to 10 seconds), the number of words processed per unit time depends on
the size of the chunk. This, in turn, is dependent on relevant prior knowledge
in long-term memory.
With books, the reader creates chunks of variable
word size to affect a reading pace (i.e., words per unit time) that
accommodates the cognitive requirements of comprehension. With television, the
pace of presentation (i.e., words or visual elements per unit of time) is not
sensitive to the cognitive constraints of the learner; it progresses whether or
not comprehension is achieved. The television viewer may be familiar enough
with the information to process it at the pace presented, even if it is
"fast." That is, the viewer's chunks may be large enough so that the
cognitive pace of processing words and ideas keeps abreast with the pace at
which the they are presented. Even if attention waivers and information is
missed, knowledge of a familiar domain can be used to fill in the gaps by
supplying information from long-term memory. If the viewer has little domain
knowledge, the chunk size will be smaller and the cognitive pace will drop,
perhaps below the pace at which ideas are presented. Also, there is less
information from long-term memory to compensate for the information that might
be missed. Because the information is transient, the viewer can not regress
over it to refresh short-term memory. This situation may result in cascading
comprehension failure mentioned by Anderson and Collins (1988). However, for
lack of research, these contentions remain speculative and empirical work in
this area is needed.
The discussion above concentrates on the potential
problems created by the transient nature of video information. On the other
hand, this transience may have some advantages in the development of dynamic
mental models. As mentioned, Greeno (1989) contends that people use mental
models to reason through the solution of problems. This is possible because a
mental model is considered to be composed of a connected, "runnable"
set of objects, or mental entities (Williams, Hollan, and Stevens, 1983). Each
of these has an associated representation of its state, a set of parameters, a
set of procedures which modify its parameters, and a set of relationships that
connect it with other objects. The model is "run" by means of
propagating a change of state in one object over time to those of connected
objects, using the associated procedures and relationships to modify their
parameters. Thus the representation is transformed from the current state to
some future state. This information is used to make inferences and solve problems
(Holland, et al., 1986).
For example, mental models in physics typically
include entities that correspond to physical objects that are encountered in
the situation, such as "blocks", "springs",
"pulleys", etc. (Larkin, 1983). People operate on the mental entities
as they would in real time and make inferences about "what would happen to
them next" in order to solve physics problems.
16
Holland, et al. (1986)
contend that learning a representation of the transition function is the
critical goal in the construction of a mental model. The prospect exists that
the transient, time-based character of video information could be used to
inform the dynamic properties of mental models, such as those in physics. The
observation of objects moving along paths, for example, could provide learners
with information needed to make estimates of changes in state. This information
would not be available with static information, such as that in text. Whereas
learners familiar with the domain might be able to supply such dynamic
information from memory or use their prior knowledge to infer dynamic
properties from static pictures, those novice to a domain may not be able to
supply such constructions and might benefit from the dynamic character of
televised information. However, as will be discussed in the subsequent section
on learning with computers, this information may not be sufficient to overcome
misconceptions that novices frequently bring to tasks, such as those involving
the motion of objects (Clement, 1983; di Sessa, 1982; McCloskey, 1983). Again,
for lack of research in this area, these contentions remain speculative.
Summary and implications.
This research paints a picture of television viewers who monitor a presentation
at a low level of engagement, their moment to moment visual attention
periodically attracted by salient audio cues, and maintained by the
meaningfulness of the material. This creates a window of cognitive engagement.
Within this window their processing is sometimes effortless, resulting in the
construction of shallow, unelaborated representations of the information
presented. However, when viewing with a purpose, people will attend more thoughtfully,
constructing more detailed, elaborated representations and drawing more
inferences from them.
The visual component of the presentation is
particularly memorable and the representations constructed with it are
especially good for carrying information about situations. The auditory symbol
systems carry information about sounds and expressive language and help in
interpreting the visual information. Auditory symbol systems alone draw
primarily on prior knowledge for a construction of the situation model and this
may be problematic for those with little prior knowledge.
Viewers
use their prior domain knowledge to process information at the pace presented
and supplement information that they may have missed. The transient information
in the presentation may be useful in building the dynamic properties of mental
models, so that inferences can be made about the phenomena they represent.
However, if the topic is unfamiliar, little information exists in long -term
memory to supplement viewing, the pace of the presentation may exceed their
capacity to process it, and comprehension failure may result.
This knowledge can be used by instructional
designers to make media-related decisions. For example, people who are very
knowledgeable about a particular domain can process information at a much
faster rate and more strategically with text than they can with audiotape or
video, suggesting that text would suffice for these learners. However, people
who are novice to a domain are likely to benefit from the ability to slow the
rate of information processing, regress over text, and move back and forth
between text and pictures, as they are presented in books. These same people
are more likely to fail at comprehending some portion of a video presentation
because their pace of
17
processing
information may fall below the pace at which it is presented. For learners
moderately familiar with a topic, television's symbol systems can supply
complementary information, particularly useful in constructing a situation
model, and its pace will accommodate comprehension. In such productions, the
linguistic information should be presented simultaneous to or just following
the visual information.
Learning with
Computers
So far, media have been described and distinguished
from each other by their characteristic symbol systems. Some media are more
usefully distinguished by what they can do with information - that is,
their capability to process symbols. This is particularly the case for
computers, the prototypic information processor. For example, computers can
juxtapose, or transform, information in one symbol system to that in another
(Dickson, 1985). A learner can type in printed text and a computer with a voice
synthesizer can transform it into speech. The computer can take equations and
numerical values or analog signals and transform them into graphs. Research is
reviewed below which shows how the computer can be used to aid students in
constructing links between symbolic domains, such as graphs, and the real world
phenomena they represent. The research shows that it is the transformation
capabilities of the computer, rather than its symbol systems, that are crucial
in this regard.
The computer is also capable of
"proceduralizing" information. That is, it can operate on symbols
according to specified rules, such that a graphic object on the screen can move
according to the laws of physics, for example. Research is reviewed below which
illustrates the role that this capability can play in aiding learners to
elaborate their mental models and correct their misconceptions with the use of
microworlds.
Connecting the real world to symbols
with MBL. An important part of school learning is acquiring
an understanding of the relationship between various symbol systems and the
real world they represent. Yet, students are frequently unable to connect their
symbolic learning in school to real world situations (Resnick, 1987). The
transformational capabilities of the computer can be used to make this
connection.
Graphs provide an example of this. Mokros and Tinker
(1987) found frequent errors among seventh and eighth grade students in the
interpretation of graphs. Two patterns were identified. First, there was a
strong graph-as-picture confusion. Half of the students drawing a graph of a
bicyclist's speed uphill, downhill, and on level stretches drew graphs
representing the hills and valleys rather than speed. In a less striking
pattern, 75% of the students responded incorrectly when asked to specify
maximum warming or cooling on a graph. About half of these selected the highest
(or lowest) point on the graph as that showing the most rapid change.
Mokros and Tinker went on to use a microcomputer-
based lab (MBL) with 125 seventh and eighth graders for three months. MBL
involves the use of various sensors (temperature probes, microphone, motion
sensors, etc.) connected to the computer to collect analog data. The computer
transforms these data and displays them in real time on the screen as a graph.
In a typical unit, the user can turn a heater on for a fixed period, thereby
delivering a fixed quantity of thermal energy to a liquid. Using temperature
18
probes
interfaced to the computer, the increase and decrease of temperature is
instantaneously graphed over time. Mokros and Tinker found a significant
increase from pre to posttests on the interpretation of graphs (from m=8.3 to
m=10.8 on a 16 item test). Of particular importance was the fact that students
made the greatest gains on items sensitive to the graph-as-picture error.
In a similar study,
Brasell (1987) used MBL with high school Physics student. One group of student
spent a class session collecting and observing MBL data in real time (the
Standard-MBL group) . A second group used the MBL equipment to collect data but
it was displayed after a 20 second delay. One control group plotted data with
pencil and paper and another control group engaged in testing only. Brasell
found that the posttest scores from the Standard- MBL treatment were
significantly higher than scores from all other treatments. The analysis
indicated that real-time transformation of data (i.e., the difference between
Standard-MBL and Delayed-MBL) accounted for nearly 90% of the improvement
relative to the control. Brasell suggests that unsuccessful students lack
appropriate techniques for referring to previous events or experience and they
fail to make explicit links between physical events and the graphed data, even
when they are displayed after only a 20 second delay. The transformation
capabilities of the computer made the connection between symbols and the real
world immediate and direct.
Building mental models with microworlds.
Experts in a domain are distinguished from novices, in part, by the nature of
their mental models and how they use them to solve problems. The processing
capabilities of the computer can help novices build and refine mental models so
that they are more like those of experts.
In physics, a series of studies (Chi, Feltovich, and
Glaser, 1981; Hegarty, Just, and Morrison, 1988; Larkin, 1983; Larkin, et al.,
1980) has established that experts have extensive domain knowledge organized
into large, meaningful schemata, or chunks, which are structured around the
principles or laws of physics. These schemata not only contain information
about the laws of physics, but information on how and under which conditions
they apply. That is, they contain both declarative and procedural knowledge.
When encountering a text book physics problem,
experts use the objects (e.g., springs, blocks, pulleys, etc.) and features
mentioned in the problem statement to cue the retrieval of one or more relevant
schemata (e.g., "force-mass" or "work-energy"). They
construct a mental model, which contains both information that has been
explicitly provided by the situation, as well as information supplied from
memory. These mental models include mental entities that correspond to physical
objects that are mentioned in the problem, such as "blocks,"
"pulleys," etc. (Larkin, 1983; Larkin, et al., 1980), as well as
entities which correspond to the formal constructs of physics that have no
direct, concrete referent in the real world, such things as "force",
"vectors", "friction", and "velocity". The
relationships among these entities correspond to the laws of physics. Experts
reason with this model to test the appropriateness of potential quantitative
solutions. It is only after this qualitative analysis is complete that the
expert will use an equation to derive a quantitative solution to the problem.
Novices represent and use information in this domain
in a very different way. Not only do they have less knowledge about physics
than do experts, but their knowledge
19
is organized quite differently. For some novices,
their physics-related knowledge is composed of a set of fragments or
"phenomenological primitives", that are not connected by formal
relationships but based on real world objects and actions. They evoke these
fragments to construct a representation of a particular problem (di Sessa,
1988). Other novices may have coherent and consistent, though erroneous
"theories", or misconceptions of the phenomenon (Clement, 1983;
McCloskey, 1983). These may represent procedural relationships that are
contrary to established laws of physics, such as "an object remains in
motion only as long as it is in contact with a mover" or "an object
should always move in the direction that it is kicked."
Confronted by a text book problem, novices will use
the same surface cues as experts to evoke this information from memory.
However, unlike those of experts, the mental models that novices construct with
this information are composed primarily of entities that correspond to the
familiar, visible objects mentioned in the problem statement (Larkin, 1983).
These representations do not contain entities that represent formal physical
constructs, such as "force" or "friction". Nor do they
contain information on physical laws and principles, or this information is
inaccurate or incomplete. Thus, the models are insufficient to determine a
solution or the solution that is specified is incorrect.
How do people modify such incomplete and inaccurate
mental models to form more accurate, expert-like models? First of all, this
process is not automatic. Indeed, such misconceptions can be held into
adulthood and after taking courses in the domain (McCloskey, 1983) . Rather,
modification of a mental model is triggered by certain conditions, such as the
failure of a model to adequately predict or account for phenomena when it is
used to achieve some desired goal (Holland, et al., 1986). In such cases, a person
can drop the current mental model in favor of another, maintain the model but
lower confidence in its ability to reliably predict, or modify the model. The
latter is most often the goal in school learning. One way a model is modified
is by elaborating its situational components. These are the criteria used to
evoke and select the appropriate model in response to a particular problem.
Another way to modify a model is by changing the transformation rules
associated with the situation. Which of these various changes ultimately occurs
depends on the accumulated previous success with the model (a model which has
been used successfully many times is more likely to be modified rather than
replaced), the perceptual elements of the situation that might allow for differentiation
(the existence of salient perceptual elements will be used to refine the
selection criteria so that it is used in a somewhat different set of
situations), and the future success of alternate models and rules when they
compete to explain subsequent situations (modifications in the model that
successfully predict subsequent situations are more likely to be retained).
Expertise is developed through a series of such differentiations and
elaborations as a result of extensive experience within a domain-- both
successful and unsuccessful.
Now, how might the processing capabilities of
computers be used by novices to aid them in building more expert-like models?
First, an important attribute of the computer is its ability to symbolically
represent entities in ways that might inform mental models. Not only can they
graphically represent concrete objects, such as carts and springs, but also
formal, abstract entities (such as the forces, velocities, etc.,), entities
that
20
novices do not normally include in their models.
Second, the computer has the important capability of being able to
"proceduralize" the relationships among these symbols. While abstract
concepts can be represented in text by symbolic expressions, such as f=ma
, or denoted in diagrams by arrows, Greeno (1989) points out that such symbols
do not "behave" like forces and accelerations. With computer models,
arrows and other symbols can behave in ways that are like the behavior of
forces, velocities, and other abstract concepts. For example, a velocity arrow
can become longer or shorter, depending on the direction of acceleration.
Furthermore, learners can manipulate these symbols and observe the consequences,
successful or otherwise, of their decisions. By using their mental models to
manipulate these entities governed by the laws of physics, novices may become
aware of the inadequacies and inaccuracies of the models. Through a series of
such experiences, they can progressively move from initial fragmented,
inconsistent, and inaccurate understanding to more elaborate, integrated, and
accurate mental models of the phenomena.
This is illustrated in several studies by White
(1984, 1989), who examined students as they learned principles of Newtonian
dynamics within computer-based microworlds. She extended the work of di Sessa
(1982), who created a computer-based Logo environment, called Dynaturtle,
in which the task was to hit a target through a series of directional
"kicks" imparted to the turtle. di Sessa observed that physics-
naive, elementary school students in his study commonly operated with an
"Aristotelian" model of force and motion, expressed as, "if you
impart a force on a moving object, then it will go in the direction last
pushed." This Aristotelian notion of force can be contrasted with the
Newtonian principle that the motion of an object is the vectorial sum of the forces
that have acted on it. An Aristotelian strategy universally used by these
students was to wait until the moving turtle was at the same height as the
target and give it a 90° kick directly toward the target. The result in this
Newtonian environment would be a
compromised
motion of 45° that would miss the target.2
White (1984) analyzed the correct, Newtonian
strategy, decomposing it into component principles (i.e., the scalar sum of
forces, the vectorial sum of forces, etc.), and created a series of games that
progressively incorporated these component strategies. Each game instantiated
both observable objects (e.g., a space ship) and formal physical objects (e.g.,
a force, represented by a key press). These objects were governed by one of the
component Newtonian principles (e.g., combining two forces to increase speed in
one direction). The series led up to the target game used by di Sessa. White
found that the group of high school physics students who used these games for
less than an hour, not only used the Newtonian strategies in the target game,
but they showed significant improvement on transfer verbal force and motion
problems. They also performed significantly better on these problems than did a
control group of students who attended a physics class but were unexposed to
the games.
White and Frederiksen (1987) present a paradigm for
the development of a progression of computer models that support conceptual
change. The progression leads the learner from advanced to simple models,
increasing in their number of rules, qualifiers, and constraints taken into
account, and in the range of problems they accommodate. The models allow
students to make predictions, explain system function
21
and
purpose, solve problems and receive feedback and explanations. Each is designed
to build upon and facilitate transformation from the previous model.
White (1989) applied this progressive paradigm to
develop a two month curriculum in Newtonian mechanics. This version contained
significant improvements in the design. Additional formal constructs from
physics were represented by dynamic symbols. For example, an history of the
object's speed was represented by a "wake", and the vectorial
components of forces acting on the object were represented by a
"datacross". As the learner applied more force to the object he or
she saw not only the resulting effect on the object as it moved, but a dynamic
decomposition of the force into its orthogonal vectors (i.e., the datacross)
and a dynamic representation of the change in velocity (i.e. its wake). The
students were also provided with additional structure, such as a set of
possible "laws" to test within the microworld, and set of real word
transfer problems. Additional forces, such as friction and gravity, could be
introduced into the system. Two classes of sixth graders were assigned to this
curriculum for forty-five minutes a day, instead of their regular science
course. At the end of the period, the groups using the microworld scored
significantly better on a range of real world transfer problems than did two
classes of sixth graders attending the regular science class. They also scored
significantly better on these items than did four classes of high school
physics students, including two classes that had just spent two and a half
months studying Newtonian mechanics.
Summary.
The studies above examined the processing capabilities of the computer and
showed how they can influence the mental representations and cognitive
processes of learners. The transformation capabilities of the computer
connected the symbolic expressions of graphs to the real world phenomena they
represent. Computers also have the capability of creating dynamic, symbolic representations
of non-concrete, formal constructs that are frequently missing in the mental
models of novices. More importantly, they are able to proceduralize the
relationships between these objects. Learners can manipulate these
representations within computer microworlds to work out differences between
their incomplete, inaccurate mental models and the formal principles
represented in the system.
White's research (1984, 1989) shows that novice
learners within these environments benefit from structured experiences of
progressive complexity which help them build and elaborate their mental models.
Research by Brasell (1987) and others suggests that such symbolic-operational
environments would be particularly powerful if directly connected to real time
phenomena. These could help learners connect their more elaborated models to
the real world experiences which they can explain.
Learning with
Multimedia
This final section is the most speculative. Little
research (particularly process research) has been done on learning with
multimedia environments, in part because most efforts in the field are focused
on development, and in part because the field is still evolving. However,
multimedia present the prospect that the various advantages of the individual
media described above can be brought together in a single instructional
environment and strategically used to facilitate learning.
22
The term multimedia has
been around for several decades (Brown, Lewis, and Harclerod, 1973). Until
recently, the term has meant the use of several media devices, sometimes in a
coordinated fashion, such as synchronized slides and audiotape, perhaps
supplemented by video. However, advances in technology have combined these
media so that information previously delivered by several devices are now
seemlessly integrated into one. The computer plays a central role in this
environment. It coordinates the use of various symbols systems--presenting
text, then in another window presenting visuals. It also processes information
it receives, collaborating with the learner to make subsequent selections and
decisions.
The following sections
review work on two, somewhat different but soon to be integrated approaches to
multimedia environments: interactive videodisc environments and hypermedia
environments. The literature reviewed reports on developments within these
fields, speculates on the cognitive impact of these environments, and raises
issues that must be addressed in future research.
Connecting mental models to the real
world with interactive video. Interactive video
integrates computer and video technologies in a way that allows both video and
computer-generated information to be displayed together. In some
implementations this information is displayed on the same screen and can be
overlayed. So for example, the video could present a view of a boulder rolling
down a hill in one window on the screen. The computer could generate force
vectors and overlay them on the moving object. In another window, a graph could
be generated that plotted velocity or acceleration over time. Alternatively,
the student may be given a workspace within which she or he could compute
acceleration or velocity.
The Cognition and Technology Group at Vanderbilt
University (1990; Sherwood, Kinzer, Bransford, & Franks, 1987; Sherwood,
Kinzer, Hasselbring, & Bransford, 1987) has developed a series of
interactive video-based, complex problem spaces (or "macrocontexts")
that are anchored in realistic goals, activities, and situations. These
macrocontexts provide semantically rich environments in which students and
teachers can collaboratively explore concepts and principles in science,
history, mathematics, and literature, and use these multiple perspectives to
solve realistic problems. The Group contends that videodisc provides a more
veridical representation of events than text, and that its dynamic, visual and
spatial characteristics allow students to more easily form rich mental models
of the problem situation.
Nationally, a number of interactive videodisc
environments are now in the stages of development and formative evaluation. One
such environment is Palenque (Wilson and Tally, 1989). Palenque
is intended to be an entertainment and educational exploratory environment for
children aged 8-14. With Palenque, the viewer becomes a member of an
archaeological team of scientists and children exploring ancient Maya ruins in
search of the tomb of Pacal, the 12 year old ruler of Palenque during its
heyday.
In an "explore
mode" the viewer can use a joystick to engage in "virtual
travel;" that is, the video uses a subjective camera perspective to allow
the viewer to "see" what he or she would be seeing if actually there,
walking and climbing among the ruins. This is accompanied by a dynamic you-are-here
map. The child can use simulated research
23
tools such as a
camera, compass, and tape recorder. In the "museum mode" the viewer
can browse through a database of relevant information including text, still
photographs, motion video, graphics, and so on. These are organized into theme
"rooms" such "Maya glyphs" and the "tropical rain
forest." In the "game mode," the viewer engages in such
activities as putting back together fragmented glyphs and constructing a jungle
symphony. Formative evaluation is examining the system's user friendliness, the
appeal of the various components, and its comprehensibility.
These systems may be particularly powerful in
representing social situations and tasks, such as interpersonal problem
solving, foreign language learning, or moral decision-making. Situational
information needed to understand and solve these semantically rich problems is
sometimes difficult to represent by computer alone and can be better
represented with video. On the other hand, as mentioned earlier (Salomon,
1983), video information alone can easily be processed in a "mindless"
and "shallow" way, thus reducing the inferences that viewers draw
from it. With interactive video, the computer can be used to help the learner
analyze the rich information present in a video scene and carefully think
through all of the factors that impinge on the problem.
For example, Covey (1990) has created a particularly
compelling moral case study, entitled A Right to Die? The Case of Dax Cowart.
In this environment, students are faced with the real-life dilemma of a young
man who, having just returned from the war in Vietnam, is involved in a flaming
accident in which he is burned over sixty percent of his body and loses his
sight. In addition, as part of this burn therapy he must be subjected to daily,
painful antiseptic washings. He demands to have the treatments discontinued and
be allowed to die. On the other hand, if the treatments are continued he can be
rehabilitated to a functional but disabled life. The student is confronted with
an important moral decision: should the treatments be discontinued?
The goal of the program is not to "teach"
or argue the student toward a specific position, but provide the viewer with a
"moral sensorium" within which to explore these issues. Covey
contends that to understand the moral position of another one must do more than
"walk in his shoes," one must "live in his skin." With this
program, which is based on a true case and filmed with the actual people
involved, the student can see the patient's treatments and in effect "talk"
to the patient, the patient's mother, and the doctors, a nurse, and a lawyer.
The student is guided through a consideration of the issues of pain and
suffering, competence and autonomy, quality of life, and the role of health
professionals. Whichever decision the student makes, she or he is presented
with contrary information intended to push them toward a deeper understanding
of their position.
Cross-media
research on the Dax case study is currently underway to examine the impact of
video alone, text alone, and interactive video on the representation and
processing of this information, and on the moral reasoning of learners. Also
being examined is the interaction between these media and students' prior
knowledge, experience, and opinions. Of particular interest will be the social and
interpersonal cues embedded the video information and how these are moderated
by computer-generated text and guidance to affect the learners' construction of
a model of the situation.
24
Stevens (1989) shows how these cues can be built
into a system and used in problem solving. In this system, a subjective camera
view is used to put the learner at the head of a conference table in the role
of team leader. The task before the team of programmers is to review and
critique program code generated by various members of the team. Critiques can,
of course, be done in ways that generate defensiveness and otherwise reduce
team productivity, and such incidents are built into the episode as it is
played out. The task of the learner/team leader is to manage the meeting and
interject comments at appropriate times to facilitate group process. The
precise timing and nature of these interjections is left open and up to the
learner. Successful behavior within the system must be responsive to social
information embedded in the presentation. The learner can interrupt the session
at a particular point, use various menus to construct a verbal statement and
give it an affective/emotional loading. The feedback is also contextual; an
expert system knowledge-base is use to present reactions of the team members as
they might be in a real meeting.
Holland,
et al. (1986) indicate that mental models of social worlds are also filled with
misconceptions and stereotypes. Typically, people believe social behavior to be
more predictable at the level of the individual than it is actually. People
tend to explain social behavior in terms of dispositions of actors rather than
the character of the situation confronting the actor. Interactive video
environments, such as the ones above, may help learners build models of social
situations and use them to understand social behavior and solve social
problems.
Navigating through symbolic expressions
with hypermedia. To this point, the paper has spent a
considerable amount of time discussing the relationship between media and the
construction of situation models. Kintsch (1989), however, points out that some
texts, such as literary texts, are studied in their own right. In these cases,
a major component of the task is to understand a text in the context of other
texts and cultural artifacts to which it refers, and within which it was
constructed. This section describes an implementation of multimedia called
"hypermedia" and speculates on its cognitive effects.
Although
hypertext and hypermedia have become common terms only recently, they are ideas
that have also been around for several decades. The terms were coined by Nelson
(1974/1987) in the sixties but his thinking was strongly influenced by the even
earlier work of Bush (1945). As defined by Nelson, hypertext is "nonlinear
text". What it has come to mean in its many emerging implementations is a
set of windows on the computer screen that are linked to information in a data
base (Conklin, 1987). "Hypermedia" is an extension to include a
variety of symbolic expressions beyond texts.
These terse definitions can benefit from an
illustration. Picture a text document displayed in a window on the computer
screen. This document can be searched by various means, including a Boolean key
word search using logical functions such as AND and OR. Imagine that the
document is an English translation of Plato's Republic, and that if
desired the user could display the document in Greek, as well, in another
window on the screen. In the English version, one could select a word and the
computer could identify its corresponding word in the Greek text; this
operation would be reciprocal. There may be other information connected to a
word or passage in the text.
25
For
example, a passage could be connected to a contemporary scholarly article that
comments on it; this article could be retrieved from the data base and be
displayed on the screen. A reference to Homer would allow the user to retrieve
and display the Iliad. Or, a word could be associated with a dictionary
definition, or a diagram, or a sound, or a bit-mapped, high resolution
photograph of an ancient artifact or sculpture or building. The name of a city
or country could be linked to a map of it. The title of a play could be linked
to a video enactment of its dramatization which could be displayed in yet
another window.
Much of the educational development of hypermedia is
occurring in a few universities, such as Project Perseus at Harvard (Crane,
1990), Intermedia at Brown University (Landow, 1989), and Hyperties at the
University of Maryland (Marchionini and Shneiderman, 1988). The domains include
the Greek classics, works of English literature, and technical material.
The potential cognitive effects of such systems
become apparent when one compares their capabilities to the reading behavior of
experts, as described in the previously mentioned Bazerman (1985) study. These
experts would read very selectively, making strategic decisions based on a
particular purpose and on highly developed schemata of their field. They scan
tables of contents and read parts of articles selectively and in a personally
constructed order. Sometimes they progress through the text rapidly, other
times they slow down, moving back and forth within a text and across texts.
This nonlinear reading would certainly appear to be facilitated the richness of
information and the nonlinear structure of hypertext.
The process may also be facilitated by an
implementation of hypertext that is not yet widely used. Most current
implementations of hypertext systems are search-and - browse systems; that is,
the learner is presented with an established database, which has been
structured by an author, and the user is free to navigate through it in
whatever way he or she may want. On the other hand, other systems (for example
Kozma, 1989, Kozma and Van Roekel, 1986) allow learners to add their own
information and construct their own relationships, perhaps symbolically
representing them by graphic, node- and-link structures. Such systems can be
made to correspond to the processes learners use when constructing
interrelationships among concepts. As Salomon (1988) points out, this may
prompt learners not only to think about ideas but to think about how they are
interrelated and structured. More importantly, they provide an explicit model
of information representation that, under certain conditions, learners may come
to use as mental models of their thinking.
Beyond the considerable literature that lauds the
potential for such systems and describes individual projects there is little research
on hypertext to date. Those studies that have been done (for example, Gay,
Trumbull, and Mazur, in press; Marchionini, 1989; and Egan, et al., 1989) focus
on the more rudimentary functions of hypertext (such as search functions) and
relatively simple tasks (e.g., identifying specific information in text),
rather than learning or problem solving. While there are some encouraging
preliminary findings in these studies to indicate that hypertext both calls on
and develops cognitive skills in addition to those used with standard text,
much more research is needed. The Bazerman (1985) study suggests that much of
the reading behavior
26
exhibited by expert physicists is due to
their considerable domain knowledge and skill with the medium. Similar research
is needed on the impact of domain knowledge and skills in hypertext.
Indeed, in a note of caution, Charney (1987)
suggests that some of the very features that make hypertext so appealing, may
make it more difficult to use for certain students. For example, the nonlinear
nature of hypertext requires readers to decide what information to read and in
what order; building such sequences is likely to be particularly difficult for
readers new to a domain. By comparison, the author-determined sequence of
information in text and the use of certain cues to signal structural
relationships may be particularly facilitative of comprehension for novices.
"Getting lost" in hypertext is another potential problem,
particularly for novices who lack the extensive schemata that would allow them
to easily locate new information within that previously encountered. Finally,
lacking domain-based selection criteria, novices may end up reading a great
deal of material that is not relevant to their purpose. While hypertext seems
to hold some promise it also poses some challenges, challenges that warrant
research in this area.
Summary and implications.
Integrated multimedia environments bring together the symbolic and processing
capabilities of various media described above to help learners connect their
knowledge to other domains. Interactive videodisc environments hold the
potential for helping learners build and analyze mental models of problem
situations, particularly social situations. Hypermedia environments are
designed to help the reader build links among texts and other symbolic
expressions and construct meaning based on these relationships. While plausible
rationales have been given for the expected effectiveness of such environments,
these must be tested and in some cases serious questions have been raised.
Nonetheless, instructional designers will find these to be powerful development
environments and they have important implications for practice.
For
example, these environments may dramatically change the nature of the media
decisions made by instructional designers. Until now, the selection of media
has been a macro-level decision. That is, the decision-- should video be used
or is audiotape sufficient?--has been based on various instructional
considerations in balance and it applies to the entire instructional
presentation and to all learners. The desirability of presenting visual
information for one component of the task would have to be balanced against the
increased cost for the entire presentation.
The structure of these
traditional, macro-level decisions has effected the conduct of media research.
The important question for media researchers has been: What is the overall
impact of one medium versus another across learners, and is this impact going
to be sufficient enough to justify the additional production and delivery costs
that might be involved? This is the meta-question that has driven research on
media for the past thirty years and has resulted in little understanding of
learning with media.
On the other hand, media decisions for integrated
multimedia environments will be micro-level decisions. With these environments
it is possible to reconfigure a presentation on the fly in response to the
needs of a particular learner. The moment to moment selection of appropriate
media can respond to specific learner needs and task demands. While
audio-linguistic or even text information may be sufficient for most of
27
the
presentation or for most learners, visual information can easily be presented
to a particular learner, for a particular segment, at a particular moment, and
for a particular purpose.
The macro-level decision still exists; the cost of
such multimedia delivery environments is high, relative to other devices.
However equipment costs are likely to continue to come down and they are, for
the most part, one time costs. Production costs can actually be lower for such
systems. Only selected segments need be videotaped; a single segment can be
produced based on pedagogical grounds without having to incur the costs of
videotaping the entire presentation. Design costs need not go up if the system
is used to make these decisions on the fly so as to avoid the need for
programming all possible branches in advance (Stevens, 1990).
A shift from macro to micro-level design decisions
requires an understanding of the moment- by-moment collaboration between a
particular learner and the medium. They raise a different set of questions for
the media researcher: What is the prior knowledge of a particular learner? How
is this represented and structured and how does the learner operate on it to
solve problems? What is the range among learners of such representations and
operations? What symbol systems can best represent various components of the
task domain? How do these correspond to the way learners represent the task?
What skills do the learners have in processing various symbol systems? How do
they process various symbol systems together? How can the medium process these
in a way that supports the learner?
Many of these questions were addressed in the
research reviewed above and this research can inform micro-level media
decisions. However, that these questions are now asked from within an
integrated, multimedia environment will raise other, more novel questions, ones
not yet addressed in research.
Conclusions
Do media influence learning? While Clark (1983)
contends that media do not influence learning under any condition, the research
reviewed in this article suggests that this position must be modified. While
some students will learn a particular task regardless of delivery device,
others will be able to take advantage of a particular medium's characteristics
to help construct knowledge.
Various aspects of the learning process are
influenced by the cognitively relevant characteristics of media: their
technologies, symbol systems, and processing capabilities. For example, the
serial processing of linguistic and pictorial information in books is very much
influenced by the stability of this technology. Some learners rely on pictures
to help construct a textbase and map it onto a model of the situation; others
can provide this model from information in memory and pictures are not needed
or audio presentations are sufficient. The processing of linguistic and visual
information in television is very much influenced by the simultaneous
presentation of these symbol systems and the information in their codes. Some
learners use these to build rich representations of situations, particularly
their dynamic aspects; others can supply this information from memory, and text
or audio presentations will suffice. The process of learning with computers is
influenced by the ability of the medium to dynamically represent formal
28
constructs and instantiate procedural
relationships under the learner's control. These are used by some learners to
construct, structure, and modify mental models; other students can rely on
prior knowledge and processes and the use of computers is unnecessary.
However, Clark (1983) contends that even if there
are differences in learning outcomes, they are due to the method used, not the
medium. With this distinction, Clark creates an unnecessary schism between
medium and method. Medium and method have a more integral relationship; both
are part of the design. Within a particular design, the medium enables and
constrains the method; the method draws on and instantiates the capabilities of
the medium. While some attributions of effect can be made to medium or method,
there is much shared variance between them and a good design will integrate
them. While in the various studies cited above learning was influenced by the
methods used, it was in part because they took advantage of the medium's
cognitively relevant capabilities to complement the learner's prior knowledge
and cognitive skills. Many of these methods would have been difficult or
impossible to implement in other media.
Finally, while Clark (1983) calls for a moratorium
on media research, this article provides a rationale for additional research on
media. There is a growing understanding of the mechanisms of learning with
media, but a number of questions remain and the cognitive effects of the more
recently developed environments are speculative. Research is needed to extend
this understanding.
This research can itself be facilitated by the use
of media. Computers provide a unique opportunity to examine learning processes
and how these interact with the capabilities of a medium. Particularly useful
is the computer's ability to collect moment-by-moment, time stamped log files
of key presses, typed responses, menu selections, etc. These data, supplemented
by videotapes of students working individually and thinking aloud can be used
to examine the effects of media on learners' mental representations and
cognitive processes (Ericsson & Simon, 1984). Videotapes of several
students working together and talking can provide insights into how cognition
is shared among students and between students and media (Roschelle & Pea,
1990). The integration of computer and video records will allow for powerful
analyses of qualitative data and the sharing of these among researchers. The
examination of the same "raw" qualitative data by psychologists,
anthropologists, and sociologists can bring multiple disciplinary perspectives
to bear on media research, as well as facilitate the linkage of these knowledge
domains that too often go unconnected.
Ultimately, our ability to take advantage of the
power of emerging technologies will depend on the creativity of designers their
ability to exploit the capabilities of the media, and our understanding of the
relationship between these capabilities and learning. A moratorium on media
research would only hurt these prospects.
29
Endnotes
1.
Greeno also points out that at least in
some cases information in the situation may be used directly without the need
to construct and operate on mental models. Pictures can be considered either as
symbolic expressions or as concrete objects in the environment.
Pictures
as situated objects may be a more efficient source for processing certain kinds
of information, quite apart from how that information is represented in memory.
For example, Larkin and Simon (1987) created formal computer models to analyze
the number of processing steps required (i.e., computational efficiency) to
extract information needed to solve problems in mechanics. Two types of data
structures were created as they correspond to the same information presented
linguistically or diagrammatically. They found considerable computational
efficiency in the processing of the diagrammatic structure, such that it was
easier for information to be searched and recognized in a diagram, and thus
inferences based on this information were more easily made by the computer
program. Similarly, Larkin (1989) contends that real objects (such as a coffee
maker), as well as manipulatable diagrams of these (as they might be created in
computer environments), facilitate cognitive tasks (such as making coffee) because
they display the current state (e.g., the filter is currently empty) and thus
reduce the need to retain this information in memory or make it easier to
recover from memory failures. Such objects let important parts of the problem
solving be done by perceptual rather than logical inference.
2.
The
graphic objects used by di Sessa may not be symbolic. That is, the objects may
not be viewed as having a referent in another domain, such as physics. Rather,
the students may learn to operate on them directly in their own right without
taking them to represent concrete objects or physical concepts. The objects
used by White (1989) are specifically designed to symbolically represent these
physical objects and concepts. However, the symbolic nature of objects remains
subjective with reference to the learner, and this factor should be explicitly
addressed in research with symbolic environments.
References
Anderson, D.R. &
Collins, P.A. (1988). The impact on children's education: Television's influence
on cognitive development. Washington: Office of Educational Research and
Improvement.
Anderson, D.R. & Field, D. (1983). Children's
attention to television: Implications for production. In M. Meyer (Ed.), Children
and the formal features of television. Munich: Saur.
Anderson,
D.R. & Lorch, E.P. (1983). Looking at television: Action or reaction? In J.
Bryant and D.R. Anderson (Eds.), Children's Understanding of Television.
New York: Academic Press.
30
Anderson,
D.R., Alwitt, L.F., Lorch, E.P., & Levin, S.R. (1979). Watching children
watch television. In G.A. Hale and M. Lewis (Eds.), Attention and Cognitive
Development. New York: Plenum.
Anderson, D.R., Lorch, E.P., Field, D.E., &
Sanders, J. (1981). The effects of TV program comprehensibility on preschool
children's visual attention. Child Development, 52,
151-157.
Anderson, D.R., Lorch, E., Field, D., Collins, P.,
& Nathan, J. (1986). Television viewing at home: Age trends in visual
attention and time with television. Child Development, 57,
1024-1033.
Anderson, J.R. (1985). Cognitive psychology and
its implications. (2nd Ed.) New York: Freeman.
Anderson, R.C., Spiro, R.J., & Anderson, M.C.
(1978). Schemata as scaffolding for the representation of information in
discourse. American Educational Research Journal, 15,
433-440.
Baddeley, A. (1981). The concept of working memory:
A view of its current state and probable future development. Cognition, 10,
17-23.
Baggett, P. (1979). Structurally equivalent stories
in movie and text and the effect of the medium on recall. Journal of Verbal
Learning and Verbal Behavior, 18, 333-356.
Baggett, P. (1984). Role of temporal overlap of
visual and auditory material in forming dual media associations. Journal of
Educational Psychology, 76, 408-417.
Baggett, P. (1989).
Understanding visual and verbal messages. In H. Mandl and J. Levin (Eds.), Knowledge
acquisition from text and pictures. Amsterdam: North-Holland.
Baggett, P. & Ehrenfeucht, A. (1982).
Information in content equivalent movie and text stories. Discourse
Processes, 5, 73-99.
Baggett, P. &
Ehrenfeucht, A. (1983). Encoding and retaining information in the visuals and
verbals of an educational movie. Educational Communication and Technology
Journal, 31(1), 23-32.
Bayle, E. (1942). The nature and causes of
regressive movements in reading. Journal of Experimental Education,
11, 16-36.
Bazerman, C. (1985). Physicists reading
physics. Written Communication, 2(1), 3-23. Beagles-Roos, J.
& Gat, I. (1983). Specific impact of radio and television on children's
story comprehension. Journal of Educational Psychology, 75(1), 128-137.
Brandsford, J.D. & Johnson, M.K. (1972). Contextual prerequisites for
understanding:
Some
investigations of comprehension and recall. Journal of Verbal Learning and
Verbal Behavior, 11, 717-726.
31
Brasell, H. (1987). The effect of real-time
laboratory graphing on learning graphic representations of distance and
velocity. Journal of Research in Science Teaching, 24(4),
385-395.
Brown, A.L. (1980). Metacognitive development and
reading. In R. Spiro, B. Bruce, and W. Brewer (Eds.) Theoretical issues in
reading comprehension. Hillsdale, NJ: Erlbaum.
Brown, J.S., Collins, A., & Duguid, P. (1989).
Situated cognition and the culture of learning. Educational Researcher, 18(1),
32-42.
Brown, J., Lewis, R., & Harclerod, F. (1973). AV
Instruction: Technology, Media, and Method. (4th Ed.) New York:
McGraw-Hill.
Bush,
V. (1945). As we may think. Atlantic Monthly, July, 101-108. Buswell,
(1937). How adults read. Chicago: University of Chicago Press. Clark, R.
(1983). Reconsidering research on learning from media. Review of
Educational Research , 53 (4),
445-459.
Clement,
J. (1983). A conceptual model discussed by Galileo and used intutitively by
physics students. In D. Genter and A. Stevens (Eds.) Mental models.
Hillsdale, NJ: Erlbaum.
Charney, D. (1987). Comprehending non-linear text:
The role of discourse cues and reading strategies. In Hypertext '87
Proceedings, 109-120.
Chi, M.T., Feltovich, P.J., & Glaser, R. (1981).
Categorization and representation of physics problems by experts and novices. Cognitive
Science, 5, 121-152.
Conklin, J. (1987). Hypertext: An introduction and
survey. IEEE Computer, 20(9), 17-41.
Cognition and Technology Group at Vanderbilt, The
(1990). Anchored instruction and its relationship to situated cognition. Educational
Researcher, 19(6), 2-10.
Covey,
P. (1990). A right to die?: The case of Dax Cowart. Presentation made at
the Annual Conference of the American Educational Research Association, Boston,
MA.
Crane, G.
(1990). Challenging the individual: The tradition of hypermedia databases.
Academic Computing, 6(11), 36-41.
di Sessa, A. (1982). Unlearning Aristotelian
physics: A study of knowledge-based learning. Cognitive Science, 6,
37-75.
di Sessa, A.
(1988). Knowledge in pieces. In G. Forman and P. Pufall (Eds.)
Constructivisim in the computer
age.
Hillsdale, NJ: Erlbaum.
Dickson, W. P.
(1985). Thought-provoking software: Juxtaposing symbol systems.
Educational Researcher, 14(5),
30-38.
32
Egan, D.E.,
Remde, J.R., Landauer, T.K., Lochbaum, C.C., & Gomez, L.M. (1989).
Acquiring
information in books and SuperBooks. Paper presented at the
Annual Meeting of the American Educational Research Association, San
Francisco.
Ericsson, K. & Simon, H. (1984). Protocol
Analysis: Verbal reports as data. Cambridge, MA: MIT Press.
Gay, G., Trumbull, D., & Mazur, J. (in press).
Designing and testing navigational strategies and guidance tools for a
hypermedia program. Journal of Educational Computing Research.
Gibbons, J., Anderson, D.R., Smith, R., Field, D.E.,
& Rischer, C. (1986). Young children's recall and reconstruction of audio
and audiovisual narratives. Child Development, 57,
1014-1023.
Goodman, N.
(1976). Languages of art. Indianapolis, IN: Bobbs-Merrill.
Greeno, J. (1989). Situations, mental models, and
generative knowledge. In D. Khahr and K. Kotovsky (Eds.) Complex Information
Processing. Hillsdale, NJ: Erlbaum.
Haas, C. (1989). "Seeing it on the screen isn't
really seeing it": Computer writers' reading problems. In G. Hawisher and
C. Selfe (Eds.), Critical perspectives on computers and composition
instruction. New York: Teachers College Press.
Halliday, M.A.
& Hasan, R. (1976). Cohesion in English. London: Longman.
Hayes,
D.S. & Kelly, S.B. (1984). Young Children's processing of television:
Modality differences in the retention of temporal relations. Journal of
Experimental Child Psychology, 38, 505-514.
Hayes,
D.S.; Kelly, S.B.; & Mandel M. (1986). Media differences in children's
story synopses: Radio and television contrasted. Journal of Educational
Psychology, 78(5), 341-346.
Hegarty, M. & Just, M.A. (1989). Understanding
machines from text and diagrams. In H. Mandl and J. Levin (Eds.), Knowledge
acquisition from text and pictures. Amsterdam: North-Holland.
Hegarty, M., Just, M.A., & Morrison, I.R.
(1988). Mental models of mechanical systems: Individual differences in
qualitative and quantitative reasoning. Cognitive Psychology,
20, 191-236.
Holland, J., Holyoak, K., Nisbett, R., &
Thagard, P. (1986). Induction: Processes of inference, learning, and
fiscovery. Cambridge, MA: MIT Press.
Huston,
A. & Wright, J. (1983). Children's processing of television: The
informative functions of formal features. In J. Bryant and D.R. Anderson (Eds.)
Children's understanding of television. New York: Academic Press.
33
Just, M.A. & Carpenter, P.A. (1987). The
psychology of reading and language comprehension. Newton, MA: Allyn
Bacon.
Kintsch, W. (1988). The role of knowledge in
discourse comprehension: A construction-integration model. Psychological
Review, 95(2), 163-82.
Kintsch, W. (1989). Learning from text. In L.B.
Resnick (Ed.), Knowing and learning: Essays in honor of Robert Glaser.
Hillsdale, NJ: Erlbaum.
Kintsch, W. & van Dijk, T. (1978). Toward a
model of text comprehension and production. Psychological Review, 85(5),
363-394.
Kozma, R.B. (1989). Principles underlying
Learning Tool. Paper presented at the Annual Meeting of the American
Educational Research Association, San Francisco.
Kozma,
R.B. & Van Roekel, J. (1986). Learning tool. Ann Arbor, MI:
Arborworks. Krendl, K.A. & Watkins, B. (1983). Understanding television: An
exploratory inquiry
into
the reconstruction of narrative content. Educational Communications and Technology
Journal, 31(4), 201-212.
Kuntz, G.C., Drewniak, U., & Schott, F. (1989). On-line
and off-line assessment of self-regulation in learning from instructional text
and picture. Paper
presented at the Annual Meeting of the American Educational Research
Association, San Francisco.
LaBerge, D. & Samuels, S.J. (1974). Toward a
theory of automatic information processing in reading. Cognitive Psychology,
6, 293-323.
Landow, G.P.
(1989). Course assignments using hypertext: The example of Intermedia.
Journal of Research on Computing
in Education,
21(3), 349-365.
Larkin, J.H. (1983). The role of problem
representation in physics. In D. Genter and A. Stevens (Eds.) Mental Models.
Hillsdale, NJ: Erlbaum.
Larkin, J.H. (1989). Display-based problem solving.
In D. Khahr and K. Kotovsky (Eds.) Complex Information Processing.
Hillsdale, NJ: Erlbaum.
Larkin, J.H., McDermott, J., Simon, D., & Simon,
H. (1980). Expert and novice performance in solving physics problems. Science,
80(4450), 1335-1342.
Larkin, J.H. & Simon, H. (1987). Why a diagram
is (sometimes) worth ten thousand words. Cognitive Science, 11,
65-100.
Levie, W. &
Lentz, R. (1982). Effects of text illustrations: A review of research.
Educational Communication and Technology
Journal, 30 (4), 195-232. Marchionini, G.
(1989). Information seeking in electronic encyclopedias. Paper
presented
at the Annual Meeting of the American Educational Research Association, San
Francisco.
34
Marchionini, G. & Schneiderman, B. (1988).
Finding facts vs. browsing knowledge in hypertext systems. IEEE Computer,
21(1), 70-80.
Mayer,
R. (1984). Aids to prose comprehension. Educational Psychologist, 19,
30-42. McCloskey, M. (1983). Naive theories of motion. In D. Genter and A.
Stevens (Eds.)
Mental models. Hillsdale, NJ:
Erlbaum.
Meringoff, L. (1982). What pictures can and can't
do for children's story understanding. Paper presented at the Annual
Meeting of the American Educational Research Association, New York City.
Meyer, B.J.F. (1975). The organization of prose
and its effects on memory. Amsterdam: North-Holland.
Meyer, B.J.F. (1985). Signaling the structure of
text. In D. Jonassen (Ed.), The technology of text. Englewood
Cliffs, NJ: Educational Technology Publications.
Miller, G.A. (1956). The magical number seven plus
or minus two: Some limits of our capacity for processing information. Psychological
Review, 63, 81-97.
Minsky, M.A.
(1975). A framework for representing knowledge. In P.H. Winston (Ed.),
The psychology of computer vision. New York:
McGraw-Hill.
Mokros,
J. & Tinker, R. (1987). The impact of microcomputer-based labs on
children's ability to interpret graphs. Journal of Research in Science
Teaching, 24(4), 369-383.
Nelson, T.
(1987/1974). Computer lib. Redmond, WA: Tempus Books.
Nugent, G.C. (1982). Pictures, audio, and print:
Symbolic representation and effect on learning. Educational Communication
and Technology Journal, 30(3), 163-174.
Perkins, D.N. (1985). The fingertip effect: How
information-processing technology shapes thinking. Educational Researcher,
14(7), 11-17.
Pezdek, K. & Hartman, E.F. (1983). Children's
television viewing: Attention and comprehension of auditory versus visual
information. Child Development, 54, 1015-1023.
Pezdek, K., Lehrer, A., & Simon, S. (1984). The
relationship between reading and cognitive processing of television and radio. Child
Development, 55, 2072-2082.
Pezdek, K. & Stevens, E. (1984). Children's memory
for auditory and visual information on television. Developmental Psychology,
20, 212-218.
Pressley, M. (1977). Imagery and children's
learning: Putting the pictures in developmental perspective. Review of
Educational Research, 47 (4), 585-622.
Resnick, L.
(1988). Learning in school and out. Educational Researcher. 16(9),
13-20.
35
Roschelle, J. &
Pea, R. (1990). Situated learning: Computer-based multimedia analysis of
learning activities. Paper presented at the Annual Conference of the
American Educational Research Association, Boston, MA.
Rusted, R. & Coltheart, V. (1979). The effect of
pictures on the retention of novel words and prose passages. Journal of
Experimental Child Psychology, 28, 516-524.
Salomon, G. (1974). What is learned and how it is
taught: The interaction between media, message, task, and learner. In D. Olson
(Ed.), Media and symbols: The forms of expression, communication, and
education. Chicago: University of Chicago Press.
Salomon, G. (1979). Interaction of media,
cognition, and learning. San Francisco: Jossey-Bass.
Salomon, G. (1983). The differential investment of
mental effort in learning from different sources. Educational Psychologist,
18(1), 42-50.
Salomon, G. (1984). Television is "easy"
and print is "tough"; The differential investment of mental effort in
learning as a function of perceptions and attributions. Journal of
Educational Psychology, 76(4), 647-658.
Salomon, G. (1988). AI in reverse: Computer tools
that turn cognitive. Journal of Educational Computing Research,
4(2), 123-134.
Salomon, G. & Clark, R. (1977). Reexamining the
methdology of research on media and technology in education. Review of
Educational Research, 47(1), 99-120.
Schallert, D.L. (1980). The role of illustrations in
reading. In R. Spiro, B. Bruce, & W. Brewer (Eds.) Theoretical issues in
reading comprehension. Hillsdale, NJ: Erlbaum.
Schank, R.C. & Abelson, R.P. (1977). Scripts,
plans, goals, and Understanding. Hillsdale, NJ: Erlbaum.
Shebilske,
W.L. and Fisher, D.F. (1983). Eye movements and context effects during reading
of extended discourse. In K. Rayner (Ed.), Eye movements in reading. New
York: Academic Press.
Shebilske, W.L. & Reid, L.S. (1979). Reading eye
movements, macro-structure and comprehension. In P.A. Kolers, M.E. Wrolstad,
and H. Bouma (Eds.), Processing of Visible Language (Vol. 1). New
York: Plenum.
Sherwood, R., Kinzer, C., Bransford, J., &
Franks, J. (1987). Some benefits of creating macro-contexts for science
instruction: Initial findings. Journal of Research in Science
Teaching, 24(5), 417-435.
Sherwood, R., Kinzer, C., Hasselbring, T., &
Bransford, J. (1987). Macro-contexts for learning: Initial findings and issues.
Applied Cognitive Psychology, 1, 93-108.
36
Simon, H.A.
(1974). How big is a chunk? Science, 183, 482-88.
Snow, R. (1989). Aptitude-Treatment interaction as a
framework for reseearch on individual differences in learning. In P. Ackerman,
R. Sternberg, and R. Glaser (Eds.), Learning and individual differences.
New York: Freeman.
Stevens, S.
(1989). Intelligent interactive video simulation of a code inspection.
Communications of the ACM, 32(7), 832-43.
Stone, D. & Glock, M. (1981). How do young
adults read directions with and without pictures? Journal of Educational
Psychology, 73(3), 419-426.
Vygotsky,
L.S. (1978). Mind in society. Cambridge, MA: Harvard University Press.
White, B. (1984). Designing computer games to help physics students understand
Newton's laws of motion. Cognition
and Instruction, 1(1), 69-108.
White, B. (1989). A microworld-based approach to
science education (Technical Report). Newton, MA: Bolt Beranek &
Newman.
White, B. and Frederiksen, J. (1987). Causal
model progressions as a foundation for intelligent learning environments.
(Technical Report). Newton, MA: Bolt Beranek & Newman.
Williams, M.D., Hollan, J.D., & Stevens, A.L.
(1983). Human reasoning about a simple physical system. In D. Genter and A.
Stevens (Eds.) Mental models. Hillsdale, NJ: Erlbaum.
Wilson, K. (1990). The "Plenque" Project:
Formative evaluation in the design and development of an optical disc
prototype. In B. Flagg (Ed.), Formative Evaluation for Educational
Technologies. Hillsdale, NJ: Erlbaum.
Winn, W. (1989). The design and use of instructional
graphics In H. Mandl and J. Levin (Eds.), Knowledge acquisition from text
and pictures. Amsterdam: North-Holland.
Wixson, K. (1984). Level of importance of
postquestions and children's learning from text. American Educational
Research Jounral, 21(2), 419-433.
Wright,
J.C., Huston, A.C., Ross, R.P., Calvert, S.L., Tolandelli, D., Weeks, L.A.,
Raeissi, P., & Potts, R. (1984). Pace and continuity of television
programs: Effects on children's attention and comprehension. Developmental
Psychology, 20(4), 653-666.
37
Tidak ada komentar:
Posting Komentar