Cognitive Efficiency: Toward a Revised Theory

of Media
[ ] Tom C o b b

Within and beyond the field of educational [] The a r g u m e n t about whether " m e d i a influ-
technology, Richard Clark's writings are ence learning" has been with us for more than a
widely believed to have shown that any num- decade, and judging from the ETR&D sl~ecial
ber of media are equally capable of delivering media issue of 1994, the discussion has started to
any instruction, so that media choices are go r o u n d in circles (Clark, 1994 a,b; Kozma,
about cost and efficiency but not about cogni- 1994; Ullmer, 1994). Typically, Richard Clark's
tion and learning (Clark, 1983; 1994b). How- p r o - m e d i a critics p r o v i d e one more stunning
ever, ~cit is accepted that one type of efficiency instance of a strong m e d i a contribution to learn-
is cognitive efficiency, then it follows that ing, and Clark counters one more time that, if
media choices are often about cognition and y o u look carefully, the contribution was caused
learning and can profit from an understanding b y the instructional m e t h o d e m p l o y e d , a
of cognitive processes. Media and learning can method that could have been realized throUgh
then be reconnected in limited ways that do any n u m b e r of media.
not compromise any of Clark's main points, Educational technology as a field n o w seems
and doing this will bring media work into line in a m o o d to move b e y o n d this issue, to
with current research into the role of surface acknowledge that media are here to stay in any
information codes. case, and d r o p the learning issue w i t h o u t resolv-
ing it, or lose it in a soft-focus vision without
separable causes or effects--in which m e d i a and
other variables "interact synergistically," or
w o r d s to that effect. However, the issue can be
resolved in a more principled m a n n e r with one
minor adjustment to Clark's position. If a recur-
ring concept in his discourse, "efficiency," is
e x p a n d e d to include "cognitive efficiency," then
media choices become connected with learning,
in some circumstances. Such an expansion is
motivated by recent developments in cognitive
research, as will be s h o w n in the second p a r t of
this paper.
W h e n a debate can neither die nor be
resolved, there is a case for suspecting the
a p p a r e n t debating point is not the real one. I
argue here that the real resistance to Clark's
position is to his subtext, always i m p l i e d and
occasionally stated, that m e d i a d e v e l o p m e n t is
technician-level work, unconnected with the
interesting questions about learning. This idea
has been intuitively resisted b y m e d i a develop-
ers, w h o think that at least some of their design

ETR&D,Vol. 45, No. 4, 1997,pp. 21-35 ISSN1042-1629 21

22 ETR&D,Vo145,No, 4

decisions, even when only about efficient ways the rapid expansion of communications technol-
to deliver instruction, can benefit from an under- ogy, there is a need for educators to be skeptical
standing of h o w people think and learn. It will of inflated media claims; to notice when expen-
be argued that this intuition is correct. sive media are promoted where cheap would
do; to center instructional designs on the learner
rather than the medium; to track learning effect
BACKGROUND AND PROBLEM to instructional cause at the lowest level of anal-
ysis possible (medium attribute rather than
Clark's famous proposal that media have no medium per se, method rather than medium,
effect on learning "any more than a truck has on message rather than method). On these points,
the quality of the goods it brings to market" is educational technology has grown up under
probably the most widely quoted and debated Clark's strong discipline.
in educational technology. If anyone needs But Clark's writings m a y have had another
reminding, Clark told us that "there are no sort of influence too. By downgrading the
learning benefits to be gained from employing importance of instructional media so thoroughly
different media in instruction" (1983, p. 450), and apparently so irrefutably, and making out-
whether media are conceived psychologically comes-based media research seem impossible to
(as symbol systems), technologically (books, do correctly, they probably helped widen the
television, computers), or attributively (the divide between learning research and media
zoom shots of television, the interactivity of development unnecessarily. There is little doubt
computers). Why? Because the many studies that such a divide exists. It is a c o m m o n observa-
from the early days of educational technology tion that the ongoing development of novel
that appeared to show a learning difference media, particularly involving computers, is pro-
brought about by an instructional medium had ceeding on a commercial basis without much
actually confounded medium and method, or input from learning research (e.g., Dick, 1991).
medium and novelty, or medium and amount of The reasons for this are no doubt many, but it is
effort invested in instructional design. This reas- easy to imagine that several rounds of advice in
sessment led Clark (1983) to the strong conclu- the past decade from a senior figure in educa-
sion that media research should come to an end tional technology would have discouraged more
until someone could come up with a novel the- than a few doctoral candidates from pursuing a
ory to relaunch it on a more useful basis. Until career in learning and media. With such a divide
then, there was little about media to interest in place, the logic of Clark's analysis becomes
aspiring educational technologists: circular and self-fulfilling: media do not affect
learning, so few learning specialists are attracted
the most essential challenge to young researchers. :. is to media work, and then instructional media are
to go beyond the media enthusiasms that brought produced commercially without input from
many of us to this field. Our ambitions far exceed the learning research, and are indeed largely equiv-
narrow efficiency questions that are available in the alent to each other and peripheral to learning.
media area. (1984, p. 241)
In the particular field of media research and
development that I know about, computer-
The "narrow efficiency questions" are the assisted language learning (CALL), practical
media choices that remain after accepting that effects of media-equivalence theory are in evi-
any m e d i u m can in principle deliver any
dence. Publications in this area cite Clark's
instruction, for example between a computer views regularly. For example, the following by
tutorial and a stand-up teacher as the more prac- Chappelle (1996) summarizes the brief history of
tical w a y to implement a branching mode of CALL research, placing Clark at a major turning
instruction in a particular setting, once branch- point:
ing has already been chosen on the basis of
method.
Influenced by research in educational technology,
Out in the field, Clark's work has had two early CALL researchers typically attempted to
sorts of influence. One has been salutary: with demonstrate CALL's effectiveness by using quasi-
TOWARD A REVISEDTHEORYOF MEDIA 2.3

experimental research designs; this research typically WHAT SOME NOVEL THEORIESOF
compared cognitive and affective outcomes of learners MEDIA MIGHT NOT BE
who participated in computer-based instruction with
those who participated in regular classrooms. . . .
[However,] the research design focusing on outcomes A typical rebuttal of Clark's position consists of
• . . [has been] criticized by educational-technology the description of an outstanding learning event
researchers such as Clark (1985; 1994) who for over a that apparently took place only with the aid of a
decade has argued that the concept of investigating
particular instructional medium. For example,
"computer effects" is conceptually flawed (p. 138).
Petkovich and Tennyson (1984, p. 235) expect to
surprise Clark with an account of a flight simu-
In the w a k e of Clark's major papers, leading
lator for i m p a r t i n g landing skills that fihds a
CALL theorists like Jamieson and Chappelle strong benefit for a certain w a y of presenting
(1988) and Dunkel (1991), counseled researchers "dynamic skeletal airport c u e s " - - s i n c e "it is dif-
to move b e y o n d the quasi-experimental, out- ficult to imagine w h a t other m e d i a attributes
comes-and-comparisons model and seek new could be used to present the same information."
p a r a d i g m s to explore roles for computers in lan- But Clark (1984, p. 240) is not surprised b y this:
guage learning.
Conrad (1996) looks at where the quest for a What I find curious about this criticism is that pilots
postcomparative p a r a d i g m in CALL had led. learned to land planes before there were computer dis-
Since the 1980s, he observes, there has indeed plays of dynamic skeletal cues. In fact, blind pilots
been a "shift in research f o c u s . . , a w a y from the have successfully landed planes. My point is that these
media attributes.., must be unique contributors to
traditional single-focus media comparison."
learning if they are to be considered necessary for
A w a y from that, but t o w a r d what? In a r a n d o m learning to take place.
sample of six language-acquisition journals
between 1992 and 1995, Conrad found that 6.2%
of the articles had a CALL focus, but less than In other words, Petkovich and Tennyson
20% of them (1.2 percent of the total) involved have missed the point. With the "unique" stipu-
an empirical study. In other words, CALL soft- lation, clearly no particular m e d i u m is or ever
ware was described and discussed in five times will be "necessary" for any particular learning to
more articles than it was tested: "publications take place, and Clark's version of m e d i a is unas-
focus on the presentation of software or guide- sailable.
lines for implementation without having p u t Another line of rebuttal has been to look at
any of the implied pedagogical assumptions to Clark's methodology, or rather the m e t h o d o l o g y
an empirical test" (p. 172). For many researchers, of the original comparison studies on which
giving u p on empirical comparisons apparently Clark's meta-criticisms are based. Romizowski
meant giving up on empirical research alto- (1988) attributes the no-difference findings in the
gether. Behind this, I a m suggesting, is an uncrit- 1970s studies to an external-internal validity
ical acceptance outside our field that educational problem at the heart of the comparisons p a r a -
technologists have unequivocally established digm. In a comparison s t u d y it is necessary to
that "there are no learning benefits to be gained compare comparables, while any n o r m a l m e d i a
from employing different media in instruction," selection process w o u l d focus on differences:
and that outcomes-based media research is full
of unresolved conceptual problems and so best Naturally you compare two media on a topic where
avoided. both have a reasonable chance of success. No one
would set up a comparison of a printed book and a
Whether or not similar patterns can be found
tape recording for a course on bird-song recognition:
in other subject areas involving media, in educa- one medium is obviously inappropriate. So you
tional technology itself m a n y seem to have felt choose an experimental topic which does not seem to
some d o w n s i d e to Clark's position, judging favor either medium particularly, and are then sur-
from the n u m b e r w h o have attempted to pick prised when no significant differences are found in the
experimental results (p. 60).
holes in it since 1983. However, it is not clear
that Clark's critics have really come u p with the
"novel theory of media" he called for. This is clearly correct; a comparison b e t w e e n
24 ETR&D,Vo145,No. 4

a printed book and a tape recording for learning unlike Kozma makes it clear he sees little likeli-
bird-song recognition would probably yield a hood that the quantitative and qualitative halves
significant difference. can ever be integrated. So media researchers
Even so, Clark's main point is not threatened. should resign themselves to looking for "two
Some amount of learning can take place with the kinds of truth."
printed book, maybe more in the case of a bird
song expert, so the tape recording could not be
described as the unique medium for the job. The Focus on Form:
main choices are ultimately about method, for Logic, Concepts, a n d Analogies
example whether to impart the information
through definition and description (in a book) or
The proposal to split media research into two
exemplification (in a recording). Following that,
kinds assumes that Clark's position is empiri-
the media choices are only about efficiency and
cally unassailable. Since the position cannot be
cost of delivery.
attacked frontally, anyone who persists in think-
Other methodology rebuttals come from ing that media could have an interesting role in
Kozma (1991; 1994) and Ullmer (1994). Kozma learning can only worry its flanks with qualita-
argues that in the years since Clark's work, both tive methods and special kinds of truth. How-
media and theories of learning and instruction ever, it is not necessary to go beyond the normal
have changed so much that new research meth- or empirical kinds of truth to find weak spots in
odologies are required to describe and guide the position. It is only necessary to focus on
them. Clark's analysis depends on a noninterac- implications that do not strictly follow from the
tive model of learning (a load of instruction is finding. The finding is that with confounds
packaged for delivery to a learner, who pas- removed, there are no media effects on what is
sively receives it, etc.), while in the present era of learned--but how something is learned is not
constructivism and distributed cognition, learn- dealt with. Unresolved how questions lurk in the
ing has been redefined as a highly interactive set logic, concepts and analogies Clark uses to
of events shared between a learner and various extend his finding into guidelines for media
h u m a n / n o n h u m a n agents, tools, and media in researchers.
differing proportions, dynamics, and synergies
First logic: Clark argues that in any media
through time. In this scenario, isolated variables
choice, it is instructional method that is the
like method, medium, or even learner make lit-
active causal variable. Methods are necessary
tle sense.
and unique, but media are not, because any
The appropriate research model for this new number of media can realize an instructional
learning, Kozma argues, is Salomon's (1991) method. But as Shrock (1994) has noticed, this
"systemic" model, in which quantitative and argument can be turned around. What instruc-
qualitative approaches, both incomplete in tional method is unique or necessary? All teach-
themselves, are integrated. However, Kozma's ers know that any content can be delivered by a
main interest is clearly in the qualitative part of variety of instructional methods, and indeed
the integrated model. The quantitative part perform informal method-comparisons research
receives little mention, apart from a vague plan on an hourly basis. For instance, new concepts
to enlist "smallest space analysis" (1994, p. 15) to can be learned through definitions or examples,
chart the course of highly interactive learning explicitly or incidentally, and so on--learning
episodes, but with no details about the exact theory has provided no final settlement on this
hypotheses to be tested. In the meantime, or any other method question. The choice
Kozma's examples of outstanding new media depends on what is k n o w n about past methods
benefits are just the familiar old benefits that in relation to what is known about present learn-
confuse medium and method, as Clark (1994 ers. Methods may be more unique, so to speak,
a,b) tirelessly points out. than media, in the sense of less numerous or var-
Ullmer (1994) joins Kozma in proposing that ious, but they are far from absolutely unique.
media researchers adopt Salomon's model, but And where there is no dear difference between
TOWARD A REVISEDTHEORY OF MEDIA 25

methods, or it is impractical to determine what between them can clearly be made only with ref-
the difference is, then the basis of method deci- erence to variables "connected with learning,"
sions is cost and efficiency, just as it is for media specifically with learners' prior skills and
decisions. knowledge--ability to read music, and prior
It is not clear that Clark would object to the knowledge of bird songs. For most learners, the
analysis so far, because it still gives priority to recording would be the obvious choice; for
methods over media• Or he might, since it learners who could read music and knew m u c h
reduces method decisions to cost and efficiency, about bird songs, the book might be more effi-
the same yardstick reserved for media. Actually, cient. Such indeed is little more than c o m m o n
the latter is more likely, because throughout sense--the c o m m o n sense we have somehow
Clark's writings on media, cost and efficiency lost sight of by confining ourselves to the terms
are presented as lower-order concerns with little of Clark's argument.
bearing on learning. For example, here is more Undeniably, learning from either medium is
advice for newcomers to educational technology logically possible--the beginner could learn to
(1984, p. 240): read music, and so forth--so it is simply effi-
ciency rather than cognition per se that makes
• . graduate students who are enthusiastic about
. up the difference. But surely "efficiency" has
media should limit their research questions to delivery been framed more narrowly than it needs to be,
issues (e.g. cost, efficiency, equity, and access). While I
and could be usefully broadened to include a
personally think that issues in that area are less engag-
ing than those connected with learning, delivery is a cru- space for "cognitive efficiency" as distinct from
cial aspect of any instructional technology. [Emphasis the economic or logistic kind. Such a conceptual
added.] broadening would readmit to the discussion
many important features of otherwise equiva-
By implication, media efficiency research, lent media, such as one medium's being more or
though officially "crucial," is not "connected less effortful than another, more or less likely to
with learning." Nor has Clark weakened this succeed with a particular learner, or interacting
position in the meantime, but rather the oppo- more or less usefully with a particular prior-
site: "there is no cognitive learning theory that I knowledge set.
have encountered where media, media attri- Cognitive efficiency would have varying
butes or any symbol system are included as vari- degrees of relevance to media decisions. Take
ables that are related to learning" (1994b, p. 7). the role of color (which can be considered a
If efficiency is not related to learning, does m e d i u m when representing arbitrarily related
this mean that learning bird songs with a tape information) in learning to distinguish two
recording might be easier and faster than learn- objects: if the task is learning to recognize your
ing it with a book, but this would have "nothing new car in a crowded parking lot, then the
to do with learning?" Even if the cognitive effort importance of color is not very great, since the
were 50 or 100 times greater with the book? The learning will depend not on color alone but also
idea flies in the face of c o m m o n sense. However, shape and size, and will normally take place
as already discussed, books vs. recording is under conditions of high error tolerance. But if
really a method difference, description vs. exem- one were charged with designing the world's
plification, and the superiority of the recording first traffic lights, the role of color in learning
is really the superiority of exemplification in this migh t be more important.
particular case. Logically, there is no doubt that drivers could
But other method-medium configurations for learn to associate stop-wait-go with any three
the scenario can be imagined. Suppose the book colors, or even shades of colors--say, three
contained not text, but the song encoded as writ- shades of blue. But for the first few months with
ten music. Then, the choice between the book three shades of blue, the accident rate would be
and the recording is not between two methods steeper than the learning curve. Given the rods
but between two media, since in either case the and cones of the h u m a n visual system, color
method is exemplification. Further, the choice associations are learned faster initially and
26 ETR&D, Vo145, No, 4

accessed faster ever after if the colors are distinct trucks that deliver food to a market; in 1994 by
(red and green are "processed independently" the equivalent forms that a medicine might
in the neural system, according to Marr, 1982, p. take---"tablets, liquid suspension, suppositories,
258). Compared to blue in three shades, red, yel- or injections" (1994a, p. 26)--and yet remain the
low, and green are a cognitively efficient learn- same medicine. The ingestion theme has been
ing medium, leading to faster learning, fewer maintained, but medicine replaces food as the
errors, and in this case fewer injuries and fatali- ingested substance. The pharmaceutical analogy
ties. seems intended to communicate the notion of
Not that the historical choice of red, yellow, equivalence more clearly, or perhaps more
and green was necessarily the uniquely best freshly (given the mileage on the truck), but in
choice (color-blind people confuse red and fact it introduces several uncontrolled novelties.
green), just a better choice than some others. The new analogy includes the consumer-
After all, the "unique" stipulation, so long a red patient-learner, w h o was formerly outside the
flag to media specialists, m a y actually have been picture. It focuses on the point of consumption,
more of a red herring. Uniqueness and necessity rather than a remote point in the chain of deliv-
play relatively minor roles in social science ery. It raises the issue that wrong medicine is
research, especially education. Take instruc- normally more serious than wrong food. Most
tional method: instructional designers choose interestingly, it elevates the media specialist
between methods all the time without being from truck driver to physician.
sure they have found the uniquely best one for a A truck driver might accept that it hardly
given task, just the best from among the alterna- matters which truck a food is delivered in (to a
tives they can think of in the time available. So if market), but no physician will accept that it
method takes precedence over medium, then hardly matters which form a medicine is deliv-
w h y should medium be held to a stricter stan- ered in (to a body). Tablet and suspension, sup-
dard than method? There may be no unique pository and injection--each has a different w a y
medium for any job, but this does not mean that of getting into a body, and interacts differently
one medium is not better than another, or that with different types and conditions of bodies.
determining which is better is not an empirical Knowing about this is a large part of a
question. physician's expertise. True, the medicine is the
Efficiency-based empirical media research same whatever the delivery, and efficiency is the
would tackle such questions as these: only consideration in choosing--but an effi-
• H o w m a n y hours are needed to learn a bird ciency that can mean wellness or illness, life or
song with a recording vs. with sheet music? death. In other words, Clark's images of equiva-
• H o w m a n y accidents take place over the lence are far from equivalent.
learning process with various colors of traffic There is no need to decide whether the new
lights? analogy or the old one is correct, or even which
is better. Each captures a possible relationship
This research could proceed on a trial-and-error
between learning and medium: sometimes
basis, setting up real situations, awaiting out-
media choices are as remote from learning as
comes, and tallying the results in box scores. Or
trucks are from consumers, sometimes as inti-
it could proceed with models of situations and
mate with it as forms of medicine are to the ill
predictive hypotheses instead. This in turn
and wounded. A comprehensive approach to
would introduce a role for theories to constrain
media will acknowledge the conditional appro-
hypotheses--and for media designers w h o
priateness of both images. What is needed is an
understood the theories.
expansion rather than a replacement.
Clark's recent writings suggest that he him-
self sees the need for some changes in emphasis.
Consider the evolution between 1983 and 1994 Models for Media Research
in the analogies (media) he uses to deliver the
idea of media equivalence. In 1983, the idea was The medical analogy also has a bearing on the
represented by the equivalent function of the kind of research that is appropriate to instruc-
TOWARDA REVISEDTHEORYOF MEDIA 27

tional media. Clark (1985) argued that outcomes ies which provisionally links media to learning.
and comparisons research should cease until The rationale for this inclusion is that while dif-
c o n f o u n d s - - m e d i u m with method, medium ferent media m a y not create different cognitive
with message--could be eliminated. The prob- products, such as concepts, schemas, and mental
lem is, they never can be, not only in education models (frankly, the jury is still out on this ques-
but in many social and applied sciences includ- tion), they clearly do create different cognitive
ing medicine. processes at different levels of efficiency (with
"Message" in education is roughly "medi- regard to speed, ease, effectiveness). In other
cine" in the health sciences. While we educators words, the form in which information is pre-
debilitate ourselves worrying about how to sep- sented can determine h o w it is prdcessed in a
arate method and message from medium, medi- mind, and hence how it can be learned.
cal theorists accept that a medicine must enter a
b o d y through some means of delivery and that
there is no neutral delivery that does not interact MEDIA DOWNGRADE AND
COGNITIVE THEORY
with the b o d y to some degree. Medical research
proceeds in the face of this problem, mainly by
building up a taxonomy of interaction effects, Two important ideas of the early cognitive era
because its brief is to cure the ill, not to close the seemed to support a downgrade of the import-
hospital until clean variables are available. This ance of media in instruction. The first was that
is in the nature of an applied science. symbol systems are caused by but do not cause
cognition. The second was that all interesting
N o r does medical research retreat from
cognitive representations and computations
empiricism as a response to indeterminacy, as is
exist within individual minds. These two ideas
typically proposed in education, tn medicine,
have now undergone extensive revision.
the efficiencies of candidate delivery systems are
compared empirically with regard to outcome,
in full knowledge that unique or final causes
ASSUMPTION 1:
m a y not be forthcoming, immediately or ever. SYMBOL SYSTEMS ARE NONCAUSAL
All social science research proceeds in the face of
one ultimate unknowable, the relative contribu-
The importance of the stimulus in behaviorist
tions of nature and nurture in human affairs, but
learning theory is well known. Since an instruc-
still finds ways to proceed on a mainly empirical
tional medium can be seen as a collection of
basis. It is hard n o w to imagine w h y education
stimuli organized for maximal associative learn-
should have been modeled on a philosophical
ing, it was probably inevitable that the cognitive
program (positivism?) or a basic science (phys-
attack on stimulus-response theory would entail
ics?) rather than an applied science (medicine).
a diminished status for instructional media in
Of course, trucks-to-market theory was not educational theory. One of Richard Clark's early
an application of positivist or physics principles, projects was to break the link between educa-
but rather of some research in early cognitive tional technology and behaviorism-based
psychology that seemed at the time to provide a audiovisualism (see, e.g., Clark & Snow, 1975),
firm, scientific, postbehaviorist basis for instruc- and much of his position on media and its wide
tional theory and practice including a theory of acceptance follows from this.
instructional media. For a representative listing The behaviorists, however, were not the only
of this research, see the reference list to Clark students of h u m a n affairs to believe in a causal
and Salomon (1986). The remainder of this role for various kinds of stimuli in learning. The-
paper will contextualize trucks-to-market theory orists in literature and art history had long held
in this early cognitive research, and then will that information codes such as painting, music,
look at more recent cognitive research to see or particular forms of literacy played causal
h o w the theory could evolve. roles in h u m a n cognition. For example,
To summarize, I am proposing to include McLuhan (1962) believed that the Greeks, by
cognitive efficiency as a variable in media stud- adding vowels to the consonantal script of the
28 ETR&D, Vo145, No. 4

Phoenicians, laid the very basis for Western civ- defined in seven-color languages like English)
ilization: "it is by the alphabet alone that men easier than off-colors (like puce or chartreuse).
have detribalized or individualized themselves In other words, color may be a continuum in
into civilization" (p. 63). physics, but in human physiology the rods and
The clearest and most influential statement of cones of the visual system pick out focal bands
the causal hypothesis came from Whorf (1956), for emphasis, and this is true whether or not the
who argued that different symbol systems (in bands happen to need naming within a particu-
this case languages) create different concepts, lar evolutionary niche. In other words, these
and indeed different mental universes: subjects' cognitive systems were not limited to
the coding system they happened to be using.
We dissect nature in lines laid down by our native lan- Therefore, surface coding systems are only tan-
guages. The categories and types that we isolate from gentially related to underlying cognitive sys-
the world of phenomena we do not find there because tems.
they stare every observer in the face; on the contrary, And if symbolic media do not cause
the world is presented in a kaleidoscopic flux of
impressions which has to be organized by our minds-- thoughts, do they have any role in learning?
and this means largely by the linguistic systems in our Clark and Salomon (1986, p. 470) spelled out the
minds (cited in Pinker, 1994,p. 59). instructional meaning of the early cognitive
studies: "the particular surface-symbolic
However, no empirical evidence was appearance of a message may be relatively less
adduced for these exciting ideas by either Whorf consequential in learning, as it is going to be
or McLuhan, and when tested empirically by handled propositionally anyway during deeper
early cognitive researchers they crumbled. processing." In other words, learning happens
at a propositional or abstract level, and it makes
Take literacy effects, such as vowelled and little difference by which route the message
unvowelled scripts: in a large, fine-grained arrives there, so long as it does somehow (trucks
study in West Africa, Scribner and Cole (1981) to market).
tested empirically the age-old assumption that
The prototype of the idea that surface form is
certain kinds of literacies either caused or
irrelevant to learning is Chomsky's (1975) the-
enabled certain kinds of cognition. After com-
ory of input in language acquisition. Any natu-
paring illiterates and literates in two types of
ral language input, however "degraded," is
writing systems on a large number of cognitive
sufficient to activate a child's internal grammar;
measures, these researchers were forced to con-
and this grammar, when fully formed, will be no
clude that there are no cognitive effects of liter-
different from anyone else's, and "vastly
acy per se--of knowing a particular script, or
underdetermined" by the input. However,
indeed of knowing any script.
drawing educational implications from this line
Or take Whorf's linguistic determinism: of research, particularly in areas outside of lan-
Rosch (1978) put this idea to the test in New guage acquisition, may have been premature on
Guinea, using cross-cultural color conceptual- both theoretical and empirical grounds.
ization as her laboratory. Color was chosen
because the spectrum is a wavelength on a con-
tinuous dimension, with no nonarbitrary divid- Premature Implications
ers between red, orange, yellow, green, and so
forth--roughly the "kaleidoscope" mentioned First, on theoretical grounds: Many early analy-
above---so following Whorf's reasoning, differ- ses of cognitive performance showed the
ent naming systems shotfld create different color importance of abstract knowledge structures
concepts. However, when Rosch tested several (whether grammars, schemas, propositional net-
speakers of a language in Papua-New Guinea works, production systems, or others) in pe r -
who had no words for color except "dark" and forming a variety of information-processing
"light," she found them nonetheless able to learn tasks. For example, novice physicists were
color words easily. Furthermore, they were able shown to form stimulus-bound surface repre-
to learn, name, and remember solid colors (as sentations of physics problems while experts
TOWARD A REVISEDTHEORYOF MEDIA 2~

formed deeper, more abstract representations, to This time, the correct cards were identified by
very different effect (e.g., Chi, Glaser, & Rees, 79% of subjects (a media effect of 600%).
1982). But how ,did these abstract structures get This same point was made in many contexts,
into experts' heads in the first place? By what for example by psychologists worl<ing in literacy
steps were surface representations transformed acquisition. Gleitman and Rozin (1973) studied
into abstract? American children with reading problems and
Maybe in the interests of doing one thing at a found that for some reason they could learn to
time, or in reaction to the behaviorists' emphasis read English if it was first recoded as a logogra-
on learning, early cognitive researchers did not phy (like Chinese) or a syllabary (like Korean)
normally deal with learning questions. As Gla- rather than a phoneme-based alphabet. Simi-
ser (1990) pointed out, the early cognitive larly, Tzeng and Hung (1981), after years of
research agenda was performance and did not work with subjects learning radically novel
entail a learning theory--even in the expert-nov- scripts and displaying widely varying levels of
ice comparisons where one might have been difficulty, concluded that different writing sys-
expected--much less an instructional theory. tems impose very different learning demands.
And if a learning theory, and following that an So did scripts, and symbol systems, and
instructional theory, both logically precede a media play a causal role in cognition after all?
media theory, then a cognitive media theory in
the early 1980s was premature.
Second, on empirical grounds: Although
C o g n i t i v e Products vs. Processes
learning was not a priority in early cognitive
research, some of it nonetheless had
Tzeng and Hung's script effect studies led them
implications for learning. The many studies of
to reconsider the wholesale rejection of Whorf's
problem solving were essentially studies of trial-
linguistic determinism. These researchers did
and-error learning, since the task before the sub-
jects was to solve a novel problem; that is, learn not doubt that different scripts eventually
to solve it. In several of the problem-solving mapped to common conceptual entities, but
studies, a clear role was indicated for the import- merely argued that different scripts could none-
ance of surface information and its form. For theless pose different processing demands along
example, Rumelhart (1980) had subjects solve a the way:
problem represented by one of two surface rep-
resentations (media). One group was given four The proposal that reading in different writing systems
may entail different processes, which in turn pose dif-
cards bearing either a letter or a number, for
ferent problems for the beginning reader, in a sense
example F, 8, 7, E, and asked to indicate which argues for a view of linguistic determinism. However,
cards had to be turned over to verify the truth of it differs from the renowned Whorfian hypothesis in
the statement, "If there is a vowel on one side of its particular emphasis on the formation of written lan-
the card there is an odd number on the other." guages (rather than spoken language per se) and on
processing differences (rather than production differ-
The correct cards were identified by 13% of the
ences...) [1981,p. 238; emphasis added.]
subjects. Other subjects were given the same
problem represented in more familiar terms: the
cards were order forms from a furniture store Tzeng and Hung were criticized in the early
with statements such as the following on their 1980s for being unreconstructed Whorfians, but
studies of script effects in text-processing
visible faces:
research now take their product-process distinc-
• Chair $75 tion for granted.
• Approved Working within this framework, psy-
• Table lamp $25 cholinguists and applied linguists have studied
empirically and in detail the differential process-
• Approved BER ing demands of several scripts. One strand in
The statement to be verified by turning cards this research has focused on the processes by
over was, "Purchases over $30 must be signed." which English and Arabic speakers read their
30 ETR&D,Vo145,No. 4

native scripts, particularly with regard to lexical linguistics has shown that cognitive processes
access (decoding word meaning). Arabic and are strongly affected by surface forms of infor-
Roman script are relatively similar (compared to mation, such as different script configurations.
Chinese), with the main difference that vowels And mainstream cognitive research is now
are not normally written in Arabic, a seemingly explicitly dealing with learning processes
small coding difference, but one that causes (Anderson, 1995). Following these lines of devel-
some large processing differences. opment, the ground for a postbehaviorist learn-
Koda (1988) showed that Arabic script facili- ing theory may soon be cleared, and following
tates meaning recovery via a mainly phono- that an instructional theory, and eventually even
logical route, English via a mainly visual route. a media theory. Whatever shape these theories
Randall and Meara (1988) showed that Arabic take, they are unlikely to cast instructional
readers fixate on centers of words, English read- media in the role of trucks to market.
ers on a series of points over lengths of words. In the meantime, there is no cognitive theory
Abu Rabia and Segal (1995) showed that while of media. There are merely guidelines from cog-
skilled lexical access in English is context free, in nitive research for media design and develop-
Arabic it is characterized by reliance on context. ment, to be discussed below.
None of this research, of course, shows that Ara-
bic and English speakers are living in different ASSUMPTION 2:
conceptual universes, just that their writing sys- COGNITION IS IN THE HEAD
tems create handling differences, at levels that
can be called "cognitive" since they involve pro- A second assumption from early cognitive
cessing and not merely intake of information. research that contributed to the downgrade of
instructional media was that all cognitive repre-
Neurological may even be the appropriate sentations and processes, or all the interesting
word in some cases, as shown in Sasanuma's ones, have their locus in individual heads. This
(1975) studies of Japanese dyslexics. Japanese assumption is no longer universally accepted.
uses two writing systems, Kanji (Chinese char- Cognition is now widely seen as being more typ-
acters) and Kana (sound-based syllables), and ically "distributed" than individual, in other
Sasanuma showed that Japanese dyslexics could words shared either between two or more
lose or recover these two systems independently humans, or between humans and various exter-
of one another, suggesting they are processed at nal symbol systems that store and even process
different brain locations. This research led our information and hence are able to do some
Coulmas (1989) to conclude "it is clear that the of our cognitive work for us. The idea of cogni-
differences b e t w e e n . . , writing systems are not tive work being shared between people and
just superficial differences of coding, but relate external representations could be expected to
to neuropsychological differences concerning yield some interesting approaches to instruc-
the storage and processing of written language tional media.
units" (p. 135). Indeed, few if any researchers
Such is the current enthusiasm for distrib-
doing empirical work in this area any longer
uted cognition that it takes an effort to remem-
regard script differences as "superficial differ-
ber the idea, or at least its articulation, is
ences of coding."
relatively novel. How did cognitive research
In summary, it was never shown that symbol ever proceed without an architecture of distribu-
systems, stimuli, and media played no role in tion? According to Zhang and Norman (1994),
cognition and learning. Some early cognitive early cognitive researchers handled distributed
research appeared to downgrade the import- cognition in one of two ways: they either
ance of surface information codes, but this ignored activity that was cognitive but not indi-
research did not distinguish between cognitive vidual, or else miscategorized as individual
product and process, and in any case did not activities that were actually distributed between
deal explicitly with learning. However, both individuals or between individuals and sym-
inadequacies are now being addressed. bolic media. Zhang and Norman discuss these
Research in specialized areas like psycho- mechanisms in the context of a classic program
TOWARD A REVISEDTHEORYOF MEDIA 31_

of early research, the Tower of Hanoi studies of notations map to the same underlying entities
problem solving (e.g., Hayes & Simon, 1977). and relations. Nonetheless, in Arabic notation,
These studies were seen at the time as dealing the problem "6 x 100 = 600" is represented so
mainly with feats of individual cognition, but in clearly that the answer springs from the state-
fact their tasks incorporated uncontrolled pro- ment of the question, while "VI times C = DC"
portions of internal and external information involves a computation two or three steps while
storage and processing. holding information in memory. In other words,
Briefly, the Tower of Hanoi puzzle involves Arabic notation is cognitively efficient for multi~
moving three disks of different sizes from one plication because it does some of the cognitive
peg to another, from a starting configuration on work involved.
the first peg to a terminal configuration on the However, it is not impossible to multiply
third (say, big-medium-small to small-medium- with Roman numerals, so no unique or neces-
big). The disks were moved to and fro several sary efficiencies are claimed for Arabic. Indeed,
times to reach the target configuration. The efficiency can be measured only against an
object was to discover patterns of human prob- objective--usually short-term efficiency of
lem solving within a limited, well structured, learning vs. long-term efficiency of use. For
and totally defined task for which the entire example, simple addition in Roman notation is
"problem space" of possible moves was known. easy to learn, involving little more than counting
Rules were imposed to vary task difficulty in a natural symbols (I + II = III), while in Arabic,
controlled manner, for example stipulating that addition cannot even begin until numeric sets
only one disk could be moved at a time. Follow- (three objects) have been recoded as arbitrary
ing the equivalence-of-media assumption, it was symbols (3), in other words until much prepara-
not considered that the way rules were repre- tory learning has taken place.
sented would affect performance. In fact, Zhang
A clear example of a short- vs. long-term effi-
and Norman point out, the content of these rules
ciency trade-off is Chinese vs. Roman script.
could be represented entirely verbally, posing a
Chinese characters allow faster reading than
heavy m e m o r y demand, or else also represented
Roman script at comparable levels of literacy,
in the environment (for example, by using disks
because the mind processes shapes and pictures
too large to lift more than one of conveniently),
faser than it does graphemes. In other words,
reducing the memory demand to an unspecified
Chinese is more efficient than Roman because it
degree--but with no experimental distinction
does more of the cognitive processing. However,
made.
Chinese also involves a longer learning process
The reasons for this, Zhang and Norman con- before reading can begin. Learning the charac-
clude, was that early cognitive theorists had lit- ters proceeds by memorization on a largely
tle awareness of the nature of external piecemeal basis over many years (Martin, 1972),
representations, and indeed "no means of while learning Roman script, after the initial dif-
accommodating them" within their assump- ficulty of recoding sounds as letters, proceeds to
tions or methodology. External objects, if they maturity on a productive (Perfetti, 1985, p. 208)
had anything to do with cognition at all, were or auto-instructional basis (Adams, 1990, p. 38).
"at most peripheral aids" such as mnemonics (p. Efficiency of eventual performance must be
88). Clark's view of instructional media is clearly weighed against efficiency of learning. The clas-
compatible with this outlook. sic problem in China has been that the learning
process was too arduous and lengthy to be com-
pleted by more than a scholarly elite (Balmuth,
Distribution and Efficiency 1982, p. 31), leaving the mass of folks illiterate
and the script's potential efficiencies unrealized.
The classic example of an external symbol sys- China's periodic interest in pinyin, a romanized
tem that does cognitive work for us is multipli- script, must be seen in this context--a case of
cation in Arabic numerals (Tzeng & Hung, 1981; "media selection" on a grand scale.
Marr, 1982). It is well known that all numerical In diverse research areas from the evolution
32 ETR&D,Vo145,No. 4

of human cognition (Donald, 1993) to con- now, their research has looked only at well-
nectionist models of learning (A. Clark, 1993) a structured toy problems like the Tower of
vastly expanded role is now regularly granted to Hanoi, but their findings are suggestive.
the invention and use of symbolic media, exter- Their experimental design used four versions
nal representations, and cognitive tools--the of the Hanoi puzzle, each a carefully specified
"things that make us smart" (Norman, 1993). An proportion of information held inside and out-
answer may even be in sight to Miller's (1956) side of memory, in contrast to the uncontrolled
ancient riddle, that if working memory is con- proportions of the original experiments. For
fined to seven bits of information then how is example, one rule was that "a disk can be placed
complex cognition possible? The answer may lie only on another disk smaller than itself": in one
less with in-the-head strategies (like chunking, experimental version, the rule was represented
automatization, top-down processing, forward verbally so that it was held in memory over the
reasoning, and skilled memory), and more in course of the task; in another version the rule
people's ability to offioad or "circumvent" cog- was encoded externally, for example as a stack
nitive work (Salthouse, 1991) through the skilled of three sizes of full coffee-cups of three sizes,
invention and employment of symbolic media. such that if a cup was placed on a larger cup it
Even Einstein said that the concept of relativity would fall in and spill coffee--creating no mem-
would never have occurred to him had he not ory burden for subjects who already believed
been working with a particular notation called that spilling coffee was undesirable.
curved-space geometry (reported in Pagels,
Zhang and Norman's levels of distributed-
1988).
ness were able to predict in detail subjects'
problem-solving performance: with more infor-
mation processed out of working memory, tasks
A Methodology for were easier, performance faster, and errors
Distributed Cognition? fewer. In other words, cognitive efficiency was
greater.
Despite the plausability of cognition being
shared between individuals and symbolic
media, up to now the idea has received little Distributed Task Analysis
empirical validation. Proponents have located
and described striking instances of shared cogni- Cognitive efficiency, then, is a measure of how
tion more often than they have proposed mech- much cognitive work is performed outside of
anisms or ways of calculating the contributions working memory in a given task, by a symbol
of agents in particular distributed systems. system (multiplying in Arabic), constraint of
There has been no established methodology for nature (disks too large to move two of), or cul-
studying the phenomenon in general, much less ture (avoid spilling coffee). Efficient instruc-
one that might be adapted to media studies. Per- tional media are symbol systems that do some of
haps that is why media researchers interested in the learners' cognitive work for them.
distributed cognition like Kozma (1994) and Ull- It goes without saying that the most efficient
mer (1994) have adopted qualitative-descriptive medium would not necessarily be ideal for
approaches to their subject. every stage of learning. The goal is to have a
However, at least one concrete step has been principled and empirical way to calculate opti-
taken toward an empirical methodology of dis- mal information distributions at various points
tributed representations. Zhang and Norman in different types of learning processes, includ-
(1994) have proposed and demonstrated' a meth- ing of course terminal distributions. Airline
odology of "representational analysis" consist- pilots are destined always to share major parts
ing of the identification, separation, and of their cognitive work with their instruments,
principled reintegration of all the internal and trapeze artists to get most of the work packed
external representations and computations that into their heads. The way forward in media
are relevant to a particular cognitive task. Up to design is to model learner and medium as dis-
TOWARDA REVISEDTHEORYOF MEDIA 33

tributed information systems, with principled, for problem solving. But unfamiliar relations
empirically determined distributions of infor- between decontextualized letters and numbers
mation storage and processing over the course are fully processed, in working m e m o r y with
of learning. Zhang and Norman's experiment predictably poor results.
shows that in principle this is possible. Clearly,
ways of calculating efficiencies and distributions
will be needed for problem spaces far more com- CONCLUSION: AN INTERIM APPR()ACH
plex and ill structured than the Tower of Hanoi TO MEDIA
puzzle. One can only hope that Zhang and
N o r m a n ' s methodology will be further devel- The notion that external stimuli, representa-
oped and extended. tions, symbol systems, and media are peripheral
However, even a conceptual version of their to cognition, and therefore to learning, is an idea
methodology can shed new light on some old attached to a body of cognitive theory that has
media conundrums. For example, it gives now been substantially modified. There is still
Petkovich and Tennyson an answer to Clark's no fully elaborated learning theory from which a
remark that their computer program was hardly media theory would follow, but there are none-
necessary if blind people could learn to land air- theless some points where media research can
planes. Clark intends a comparison between two be usefully aligned with recent cognitive
instructional media, voice vs. computer, imply- research.
ing there is no interesting difference between 1. There is no further reason for media
them. But the comparison may be more usefully researchers to accept that their work has "noth-
represented as two information distributions: ing to do with learning." First, it is n o w gener-
one involves verbal learning via the medium of ally accepted that the ability to interface with
voice or braille, with all information held in symbolic m e d i a and integrate their outputs is
memory, while the other involves verbal-visual nearer the heart of h u m a n cognition than its
learning via the computer program, with con- periphery. Second, different representational
trolled amounts of information processed out- forms of the same underlying information
side the mind and remembered on a computer clearly affect how the information can be pro-
screen. As in Zhang and Norman's experiment, cessed and learned. Therefore, the design of
the default prediction is that the greater propor- such forms is an activity that can be aided by an
tion of work performed in memory, the more understanding of cognitive processes.
arduous and error-prone the learning (a view 2. There is no further reason for media dis-
apparently shared by the airline industry, which cussions to be limited by the idea that only
invests heavily in flight Simulators, less in books unique or necessary media solutions are worth
and lectures). talking about. There are clearly m a n y media for
Or, take Rumelhart's problem solvers. Sub- any instructional job, but this does not mean
jects could decide which cards to turn over more they all do it at the same level of efficiency--
easily when the problem was phrased in terms whether economic, logistic, social, or cognitive.
of a furniture store than when presented as It is precisely the job of the media specialist to
decontextualized vowels and consonants. Nor- know the range of media that can realize any
mally this is attributed to the presence of a "store instructional methodology, and to find the ones
schema" in the former condition. The schema that best match all the resources of their target
explanation, however, merely begs another learners.
question: what does a schema do? A distribu- 3. There is no further reason for media
tion-of-information analysis suggests an answer, researchers to accept that the only methodology
that schemafized information (including for available to them is qualitative. As useful as
example the idea that large purchases may be qualitative studies m a y be for exploring new
subject to special controls) is to a large extent technologies and formulating relevant hypothe-
preprocessed in a consumer culture, and so ses, the hypotheses themselves should be tested
imposes a low memory demand when called up empirically. At present, there is no reason w h y
34 ETR&D,Vo145, No, 4

the cognitive efficiencies of otherwise equivalent In M. Wittrock (Ed.), Third handbook of research on
media cannot be compared empirically, for teaching (pp. 464-478). Chicago: Rand McNally.
example on uncontroversial measures like ease, Clark, R.E. and Snow, R.E. (1975). Alternative designs
for instructional technology research. Audio-Visual
speed, a n d effectiveness of learning. For the Communications Review, 23(4), 373-394.
future, empirical methodologies are being Conrad, K.B. (1996). CALL--Non-EnglishL2 instruc-
developed for exploring distributed cognition tion. Annual Review of Applied Linguistics, 16, 158-
that m a y be adaptable to the goal of modeling 181.
learners and media as distributed systems, and Coulmas, F. (1989). The writing systems of the world.
Oxford: Blackwell.
this is clearly a promising area for further
Dick, W. (1991). An instructional designer's view of
research. constructivism. Educational Technology, May, 41-44.
[] Donald, M. (1993). Precis of Origins of the modern
mind: Three stages in the evolution of culture and
cognition. Behavioral and Brain Sciences, 16, 737-791.
Tom Cobb is in the Ecole de langues, at the Dunkel, P. (1991). The effectiveness research on com-
University~ du Qu6bec ~ Montr6al, in Montr6al, puter-assisted instruction and computer-assisted
Canada. His e-mail address is cobb.tom@uqam.ca. language learning. In P. Dunkel (Ed.), Computer-
assisted language learning and testing: Research issues
and practice. New York: Newbury House.
Glaser, R. (1990). The reemergence of learning theory
REFERENCES within instructional research. American Psychologist,
45(1), 29-39.
Abu Rabia, S. and Siegel L.S. Different orthographies, Gleitman, L.R. and Rozin, P. (1973). Teaching reading
different context effects: The effects of Arabic sen- by use of a syllabary. Reading Research Quarterly, 8,
tence context in skilled and poor readers. Reading 447--483.
Psychology, 16(1), 1-19. Hayes, J.R. and Simon, H.A. (1977). Psychological dif-
Adams, M.J. (1990). Beginning to read: Thinking and ferences in problem isomorphs. In N.J. Castellan,
learning about print. Cambridge, Mass. : MIT Press. D.B. Pisoni, and G.R. Potts (Eds.), Cognitive theory.
Anderson, J.R. (1995). Learning and memory. New York: Hillsdale, NJ Erlbaum.
Wiley. Jamieson, J. and Chappelle, C. (1988). Using CALL
Balmuth, M. (1982). The roots of phonics. New York: effectively: What do we need to know about stu-
Teachers College Press. dents? System, •6(2), 151-162.
Chappelle, C. (1996). CALL--English as a second lan- Koda, K. (1988). Cognitive processes in second-lan-
guage. Annual Review of Applied Linguistics, 16, 139- guage reading: Transfer of L1 reading skills and
157. strategies. Second Language Research, 4(2), 133-156.
Chi, M.T.H., Glaser, R., and Rees, E. (1982). Expertise Kozrna, R.B. (1991). Learning with media. Review of
in problem solving. In R.J. Sternberg (Ed.), Advances Educational Research, 6•(2), 179-211.
in the psychology of human intelligence, Vol. 1. Kozma, R.B. (1994). Will media influence learning?
Hillsdale, NJ: Erlbaum. Refraining the debate. Educational Technology
Chomsky, N. (1975). Reflections on language. New York: Research and Development, 42(2), 7-19.
Pantheon. Marr, D. (1982). Vision: A computational investigation
Clark, A. (1993). Symbolic invention: The missing into the human representation and processing of visual
(computational) link? Behavioral and Brain Sciences, information. New York: Freeman.
16, 753-754. Martin, S.E. (1972). Nonalphabetic writing systems:
Clark, R.E. (1983). Reconsidering research on learning Some observations. In J.F., Kavanagh and I.G. Mat-
from media. Review of Educational Research, 53(4), tingly (Eds.), Language by eye and by ear: The relation-
445--459. ship between speech and reading (pp. 81-102).
Clark, R.E. (1984). A reply to Petkovich and Tennyson. Cambridge MA: MIT Press.
Educational Communications and Technology Journal, McLuhan, M. (1962). The Gutenberg galaxy. Toronto:
32, 238-241. University of Toronto Press.
Clark, R.E. (1985). Confounding in educational com- Miller, G.A. (1956). The magical number seven, plus-
puting research. Journal of Educational Computing or-minus two: Some limitations on our capacity for
Research, 1(2), 137-148. informationprocessing. Psychological Review, 63, 81-
Clark, R.E. (1994a). Media will never influence learn- 97.
ing. Educational Technology Research and Development, Norman, D.A. (1993). Things that make us smart: Defend-
42(2), 21-29. ing human attributes in the age of the machine. Reading,
Clark, R.E. (1994b). Media and method. Educational MA: Addison-Wesley.
Technology Research and Development, 42(3), 7-10. Pagels, H.R. (1988). The dreams of reason: The computer
Clark, R.E. and Salomon, G. (1986). Media in teaching. and the rise of the sciences of complexity. New York:
TOWARD A REVISEDTHEORYOF MEDIA 35

Bantam Books. Salthouse, T.A. (1991). Expertise as the circumvention

Perfetti, C.A. (1985). Reading ability. New York: Oxford of human processing limitations. In K.A. Ericsson
University Press. and J. Smith, Toward a general theory of expertise: Pros-
Petkovich, M.D. and Tennyson, R.D. (1984). Clark's pects and limits. New York: Cambridge University
Learning from Media: A critique. Educational Com- Press.
munication and Technology Journal, 32(4), 233-241. Sasanuma, S. (1975). Kana and Kanji processing in Jap-
Pinker, S. (1994). The language instinct: How the mind anese aphasics. Brain and Language, 369-383.
creates language. New York: William Morrow and Scribner, S. and Cole, M. (1981). The psychology of liter-
Co. acy. Cambridge MA: Harvard University Press.
Randall, M., and Meara, P. (1988). How Arabs read Shrock, S. (1994). The media effects question: Read the
Roman letters. Reading in a Foreign Language, 4(2), fine print, but don't lose sight of the big picture. Edu-
133-145.
cational Technology Research and Development, 42(2),
Romizowski, A. (1988). The selection and use of instruc- 49-53.
tional media: For improved classroom teaching and for
interactive, individualized instruction. London: Kogan Tzeng, O.J.L. and Hung, D.L. (1981). Linguistic deter-
minism: A written language perspective. In O.J.L.
Page.
Tzeng and H. Singer (Eds.), Perception of print: Read-
Rosch, E. (1978). Principles of categorization. In E.
ing research in experimental psychology. Hillsdale, NJ:
Rosch a n d B. Lloyd (Eds.), Cognition and categoriza-
Erlbaum.
tion. Hillsdale, NJ: Erlbaum.
Rumelhart, D.E. (1980). Schemata: The building blocks Ullmer, E. J. (1994). Media and learning: Are there two
of cognition. In R.J. Spiro, B.C.Bruce, and W.F. kinds of truth? Educational Technology Research and
Brewer (Eds.), Theoretical issues in reading comprehen- Development, 42(1), 21-32.
sion. Hillsdale,NJ: Erlbaum. Whorl B.L.. (1956). Language, thought, and reality.
Salomon, G. (1991). Transcending the qualitative- Edited by J.B. Carroll. Cambridge, MA: MIT Press.
quantitative debate: The analytic and systemic Zhang, J. and Norman, D.A. (1994). Representations of
approaches to educational research. Educational distributed cognitive tasks. Cognitive Science, 18, 87-
Research, 20(6), 10-18. 122.