| Over
the last few decades, we have been learning a great deal about how
people learn and the types of conditions that optimize long-term
retention and transfer, and numerous findings from this research
have important implications for ways in which we can improve instructional
practice. In this presentation, I focus on those results indicating
that in order to maximize the effectiveness of instruction and training,
we need to pay greater attention to an old distinction in psychology—namely,
the distinction between performance and learning—but in a
slightly different way than researchers thought about this distinction
in the past.
Early investigators
of learning were forced to make a distinction between performance
and learning when several, now classic, studies showed that—despite
the lack of any evidence in an animal’s performance during
training—learning had nonetheless occurred and could be revealed
under the right circumstances, such as when a food reward was introduced
into the situation. More recently, a variety of results suggest
that what we might think of as a corollary to this earlier distinction
needs to be made. Specifically, whereas learning can be occurring
with no apparent change in performance during training, improvements
in performance during training can occur with little or no durable
learning being achieved. Or, put slightly differently, conditions
of instruction that make performance improve rapidly often fail
to support long-term retention and transfer, while conditions of
instruction that appear to create difficulties for the learner,
often slowing the rate of apparent learning, can actually optimize
long-term retention and transfer.
As a consequence
of this corollary, performance during training can be a poor and
unreliable guide to whether the type of learning that is the goal
of our instruction—that is, learning that will be both durable
and support transfer—has actually occurred. But, of course,
what is readily observable to us as instructors is the performance
of our students during instruction and training. Consequently, as
instructors, we can easily be misled into using manipulations of
training and instruction having the property of enhancing performance
during training and instruction, but failing to support learning
as measured by long-term retention and the transfer of skills and
knowledge. And, conversely, as instructors, we can easily be led
away from using conditions that introduce difficulties for the learner
and appear to slow the rate of learning, but that are actually enhancing
post-training retention and transfer.
A discussion
of these latter types of conditions—originally labeled as
“desirable difficulties” by Robert A. Bork (1994) to
indicate their property of being conditions of instruction that
seem to present difficulties for the learner, that appear to slow
down the rate of acquisition, but actually result in better long-term
learning and transfer—constitutes the remainder of this presentation.
In this discussion, I hope to accomplish two main goals. First,
I hope to give you a feeling for a few types of desirable difficulties,
one of which I will also illustrate with experimental findings.
And, second, in this context, I want to point out the potential
for teachers and trainers—as well as students and trainees—to
be misled as to what are and are not good educational practices
or good conditions of learning.
As instructors,
we can often be misled in this determination because what is readily
available to us is the performance of our students during instruction,
which can be a poor indicator of whether durable learning is actually
occurring. If, for example, all we consider is the rapidity and
apparent ease of their learning during training and instruction,
we can easily be led into preferring poorer conditions of learning
to better conditions of learning. Additionally, as learners, it
seems that we do not develop—through the trials and errors
of everyday living—an accurate mental model, so to speak,
of those operations that result in learning and those that do not.
Furthermore, we are fooled by certain indices—such as how
fluently we process information during the re-reading of to-be-learned
material—into illusions of learning and/or competence that
then lead us to prefer poorer conditions of learning to better conditions
of learning.
So, what are
some of these manipulations or conditions of instruction that introduce
desirable difficulties for the learner? I briefly describe five
of them. Then, I illustrate one—providing contextual interference
for the learner—with some experimental findings. Finally,
I present a number of points that, as instructors, we should keep
in mind to try to introduce some of these desirable difficulties
into the design of our undergraduate courses and curricula.
Manipulations
that Introduce Desirable Difficulties for the Learner
- Varying
the Conditions of Practice. When instruction occurs
under conditions that are constant and predictable, learning appears
to become what might be called contextualized. That is, while
it looks very good in that context, the learning acquired in that
context does not support retention later when tested in other
contexts, and it does not transfer well to different contexts.
In contrast, varying conditions of practice—even just the
place where you study (as illustrated by Smith, Glenberg, &
Bjork, 1978, and by Smith & Rothkopf, 1984)—can enhance
recall at a later time. With respect to these findings, it is
interesting to note that a how-to-study hint frequently given
to students is that they should find a quiet, convenient place
to study and then do all their studying in that same place.
- Providing
Contextual Interference during Learning. If when
trying to learn several different things, you intertwine the learning
of those things in such a way as to cause interference among them
during acquisition, long-term performance on them will be enhanced.
This type of desirable difficulty, often accomplished by interleaving
the practice of the various things to be learned, rather than
blocking their practice, is the desirable difficulty that I will
illustrate with some relevant experimental findings.
- Distributing
or spacing study and practice. The effects of distributed
practice on learning are somewhat complex. Although massing practice
(e.g., cramming for exams) supports short-term performance, spacing
practice (e.g., distributing presentations, study attempts, or
training trials) supports long-term retention. That the spacing
of practice enhances long-term performance is among one of the
more robust and general findings in learning research, holding
across a variety of spacing intervals, types of materials, and
types of learners. Unfortunately, however, because massed practice
or study can support short-term performance, students can be rewarded
by good test performance following an all-night cramming session.
Little of what they were able to recall after such a short delay,
however, will still be recallable after a more substantial delay;
whereas, had they distributed their study, much more of the to-be-learned
material would still be recallable after a long delay. If throughout
the duration of a course, students simply cram for each exam and
there is no cumulative final for which they must go back and re-study
information already tested, it is little wonder that most students
appear to retain very little of the content of a course they had
presumably mastered within even a moderate delay from having completed
it.
- Reducing
feedback to the Learner. That reducing feedback
to the learner during acquisition could be a desirable difficulty
seems very strange. Indeed, for many years in the area of motor-skills
learning, it was thought that the more feedback you give the learner,
the faster and better the learning would be. More recent work,
however, has shown that by reducing the feedback you actually
enhance the long-term retention and generalizability of motor
skills—that is, the ability to produce those skills accurately
after a long delay and under different circumstances. (For reviews
of the work supporting this new view of feedback and why reduced
feedback leads to more durable and flexible learning, see Schmidt
& Bjork, 1992, and Christina & Bjork, 1991.)
- Using
Tests (rather than presentations) as Learning Events.
Much research in the laboratory (e,g., Landauer & Bjork, 1978;
Carrier & Pashler, 1992) has demonstrated the power of tests
as learning events and, indeed, in terms of long-term retention,
such research has demonstrated that a test or retrieval attempt,
even when no corrective feedback is given, can be far more effective
than a second presentation or study opportunity. In addition,
much current research is being addressed to questions concerning
test effects, such as the optimal distribution of tests, the optimal
form of tests for different types of delays and materials, and
the optimal use of feedback with respect to testing outcomes.
I do not have time to cover this work in today’s talk, but
before leaving this topic, I do want to make two points relevant
to testing effects.
First, it seems
clear that the value of tests as learning events is greatly underappreciated
in most educational contexts, where, instead, tests are primarily
viewed as assessment tools. Clearly, those of us who study learning
in the laboratory must do a more effective job of communicating
to teachers and instructors, in general, about the power of tests
to promote learning, not just assess it. To address this need, Roediger
and Karpicke (2005) at Washington University are currently looking
at testing effects with educationally realistic materials and are
obtaining dramatic and compelling evidence concerning the benefits
of testing over representations of material. As more of these types
of results, obtained with such materials, become available, our
ability to communicate to teachers and instructors regarding the
effectiveness of tests as learning events should be greatly improved.
(For references demonstrating the effectiveness of tests as learning
events and discussions of why tests are so effective, see Bjork,
1975; Bjork & Bjork, 1992; & Carrier & Pashler, 1992;
and for a review of this literature, see Dempster, 1996.)
Second, because
students, by and large, do not realize that tests—or attempts
to retrieve information—are more effective in promoting learning
than are repeated presentations of the material to be learned, they
are led to adopt highly inefficient study activities. Were we, for
example, to follow some typical students around campus and watch
how they went about studying, we would find that they spend way
too much time representing information to themselves—reading
a chapter over and over again, highlighting passages in different
colors, and so forth—and far too little time trying to retrieve
information. Or, put slightly different, they would be spending
far too much time on the input side of learning and far too little
time on the output side of learning. That this mode of studying
is so typical among students stems, at least in part, from a faulty
mental model of how we learn and remember. They, as many of us do,
tend to think of memory as being too much like a tape recorder.
Thus they feel that if they just present materials over and over
again to themselves, eventually it will write itself on their memories.
As it turns out, however, nothing could be further from the way
we actually learn and remember.
Contextual
Interference as a Desirable Difficulty
I turn now
to the desirable difficulty of contextual interference and to demonstrate
it with some empirical studies. In the first study I discuss, by
Shea and Morgan (1979), contextual inference during learning was
provided by having some subjects learn three different movement
patterns in an interleaved manner, while others learned them in
a blocked manner. The apparatus used by Shea and Morgan looked somewhat
like a pinball machine, having two vertical rows of hinged paddles
on each side with a start button and a hole containing a tennis
ball located between these two rows. In addition, located at the
back of the apparatus were three differently patterned stimulus
lights, each of which was associated with a different movement pattern
that the participant was to learn. When one of the lights came on,
the participant was to: 1) push the start button; 2) pick up the
tennis ball; 3) while holding it, knock down the paddles in the
manner associated with that particular light (e.g., knocking down
the first paddle in the left row, then the middle paddle in the
right row, and then the rear paddle in the left row); and, 4) when
finished, return the ball to its initial location, which turned
off a response timer.
In the blocked
condition, participants learned the three movements by practicing
only one pattern at a time in a blocked manner. For example, a given
participant would practice the first pattern to be learned, say
A, for many times in a row, then movement pattern B for the same
number of trials, and then movement pattern C, also for the same
number of trials. For participants learning in the interleaved (or
random) condition, the light designating a given movement, say A,
might come on for the first practice trial, then the light designating
movement C, then A again, then B, then C, and so forth, in a random
order, until the participant had practiced each movement pattern
for the same number of trials as had the participants in the blocked
condition.
As might be
expected, during training, the performance of the participants given
blocked practice improved much more rapidly than did that of the
participants in the interleaved or random condition. Although performance
in the interleaved condition eventually caught up to that in the
blocked condition, it took quite a while for it to do so—essentially,
twice as long to attain the same asymptotic level of performance.
If Shea and Morgan had ended their study at this point, and, thus,
all the results available to us would have been the participants’
performance during acquisition or training, it would seem clear
that blocking of practice trials was the superior leaning procedure.
But, fortunately, Shea and Morgan did not stop their study at this
point. Rather, they had participants return after 10 days at which
time they were given a retention test on the movement patterns—a
final exam, so to speak. What happened on this exam was quite dramatic!
Shea and Morgan
tested their participants in two ways: either under conditions that
matched those present during training or under conditions that did
not. Thus, for participants trained initially in the blocked condition,
half were tested under blocked conditions again and half were tested
under interleaved or random conditions. Similarly, for participants
trained under interleaved or random conditions, half were tested
under the interleaved conditions again and half under blocked conditions.
When testing was done under interleaved conditions, the participants
who had been trained under those conditions performed essentially
as well as they had on their last day of training—that is,
they showed little or no forgetting of the three movement patterns.
In dramatic contrast, those participants who had been trained under
blocked conditions—the participants who had looked the best
during training—performed exceptionally poorly on the test.
Indeed, their performance was so poor as to look like they had never
been trained in the first place. When participants were tested under
blocked conditions, the performance of participants trained under
blocked conditions was much better, showing only a small amount
of forgetting, but—of greater importance—the performance
of participants trained under interleaved conditions also showed
little or no forgetting. Indeed, if anything, their performance
was better—even when tested under blocked conditions—than
that of the participants originally trained in that manner.
In other words,
when participants trained under blocked conditions were later tested
under conditions not identical to those present during their training,
their performance was extremely poor, essentially looking like they
had never been trained at all. In contrast, participants trained
under interleaved conditions were not only able to perform with
little or no forgetting when tested under the same conditions, they
were also able to perform with little or no forgetting under changed
conditions. This pattern of results thus provides a dramatic illustration
of the benefits of introducing contextual interference into the
learning process. Although slowing acquisition during training relative
to blocked practice, the contextual interference introduced by the
random practice procedure served to enhance performance at a delay
and in a different context.
Several possibilities
have been advanced in the literature to explain why interleaving
might be so beneficial for long-term retention and transfer. One
of these (e.g., Battig, 1966) is in terms of the learner having
to resolve the interference among the different things that he or
she is trying to learn. To accomplish this resolution, the learner
has to notice similarities and differences among them and to schematize
or develop a more abstract representation of each item or movement.
This higher-order type of learning is what permits both long-term
retention and transfer. Another explanation assumes that what is
beneficial in the interleaving procedure is that it forces us, as
learners, to reload our memories for the different things we are
trying to learn over and over again. If required to do A, then B,
then C, and then B again, the memory for how to do B is not just
sitting there in short-term memory waiting for us to access with
no effort. Instead, we have to retrieve it again from long-term
memory. These successive attempts to retrieve things that have been
forgotten from short-term memory are what lead to the enhanced long-term
retention in the interleaved situation. (For a discussion of forgetting
as a condition for learning, see Bjork, 1994; Estes, 1955; and Cuddy
& Jacoby, 1982.)
While the results
of the Shea and Morgan study illustrate how we, as instructors,
could easily be misled by the performance of our students during
instruction or training into preferring a condition of instruction
that is actually not supportive of long-term retention and transfer
over one that is, the next study I describe illustrates how we,
as learners, can similarly be misled into preferring poorer conditions
of learning to better conditions of learning. In this study, conducted
by Simon and Bjork (2001), participants also learned three different
movement patterns, and they also learned them in either a blocked
or interleaved (random) order. Rather than knocking down paddles,
however, participants in the Simon and Bjork study learned to execute
three different movement patterns on a computer number pad in a
specific amount of time (i.e., 900, 1200, and 1500 milliseconds),
and they were given feedback on how close they had come to the required
duration after each trial. Twenty-four hours after their training,
participants returned to the lab and were tested on the three movements.
Consistent with the results of Shea and Morgan, participants who
learned under blocked training performed better during acquisition;
but 24 hours later, they performed more poorly than the participants
who had received the random or interleaved training.
The new wrinkle
in the Simon and Bjork study was that participants were periodically
stopped during training and asked to take a reading on how well
they were learning the task. They were asked, if you were to stop
training right now and come back in 24 hours, how well do you think
you would do—that is, how close do you think you could come
to the correct movement time. Participants in the blocked condition
all predicted that they would do better than the participants in
the interleaved condition predicted that they would do. In other
words, their meta-cognitive assessment of how well they were going
to do later was exactly wrong. Participants in the blocked condition
most likely mistook the rapidity and apparent ease of their being
able to perform the required movement patterns—made possible
by the blocking of practice trials—as indicating that they
were actually learning them well; whereas, the participants in the
interleaved condition most likely mistook the slowness and apparent
difficulty with which they were being able to perform the required
motor pattern as indicating that they were not learning them well.
(For a relevant discussion of such confusion between performance
and learning in terms of the difference between the retrieval strength
and storage strength of memories, as hypothesized in a new theory
of disuse, see Bjork & Bjork, 1992.)
Thus, taken
together, these two studies illustrate both how we, as instructors,
can be misled if we only attend to or only have available to us
the performance of our students during acquisition, and how we,
as learners, can be misled into thinking that we are learning better
under one condition than another when, in fact, the opposite is
true. Unfortunately, as learners, we do not seem to be very good
at assessing our actual state of competence or knowledge during
training and seem easily misled concerning the conditions of training
and instruction that are optimal. We seem, for example, to intuit
that we are learning better under massed as opposed to spaced conditions
of practice, or when the conditions of learning are kept constant
as opposed to varied, or when we are given more rather than less
feedback. Apparently, these conditions—because they support
our performance during training—give us a sense of ease and
a sense of learning that turns out to be misleading as far as the
actual long-term learning that we are achieving. Whether or not,
we, as learners, could be made to be more meta-cognitively sophisticated
with respect to when we are or are not learning well is a topic
of considerable research interest right now. (For a more thorough
discussion of factors that can lead to such “illusions”
of knowledge and/or competency, see Bjork, 1999, and Jacoby, Bjork,
& Kelley, 1994).
Now, in case
by the studies I have used so far to illustrate the benefits of
contextual interference, I have created the impression that this
desirable difficulty only works with motor learning or simple materials,
I end by describing two studies using more educationally relevant
materials. In the first study, Mannes and Kintsch (1987) examined
the effects of contextual interference on learning from the reading
of text. Participants were given a certain period of time to study
a technical, but somewhat interesting article on the industrial
use of microbes and bacteria with the clever title, Industry
in Ferment. Prior to studying this article, however, participants
had either been given a consistent or an inconsistent outline to
read. The consistent outline had the same structure as the article
and 25% of the information in the article was presented in the outline;
thus, it was very much like the type of advanced organizer frequently
used in educational settings. The inconsistent outline had all the
same factual information—thus it too had 25% overlap with
the Industry in Ferment article—but it was actually
the outline of an Encyclopedia Britannica article on microbes and,
thus, it mismatched the article in a number of ways. After participants
had studied their assigned outline and then the article, different
types of tests were administered. When given a straightforward,
verbatim recall kind of test, participants who had received the
consistent outline performed better. When given a test that involved
problem solving and a deeper understanding of the article, however,
the participants who had received the inconsistent outline performed
better.
How can we
explain this pattern of results? Mannes and Kintsch argued that
the inconsistent outline created contextual interference for the
readers, forcing them to engage in more active processing of the
material in order to resolve this interference. To make peace, so
to speak, between the two sources of information, these readers
were forced to notice similarities and differences between them
and to make inferences in order to bridge gaps between them. Consequently,
the readers in the inconsistent-outline condition achieved a deeper
understanding of the material than did those in the inconsistent-outline
condition.
Although Mannes
and Kintsch did not do so in this study, it is interesting to speculate
what they would have discovered had they asked their participants
how helpful they had found their outlines to be. Participants receiving
the consistent outline would probably have given the outline high
marks. But what about the participants in the inconsistent condition?
Most likely, they would not have given their outline high marks.
In fact, they would probably have complained about its inconsistency
with the article, even though it was probably in the resolution
of these inconsistencies between the outline and the article that
learning of a deeper kind was taking place. Almost certainly, however,
like the participants in the interleaved versus the blocked conditions
of the Simon and Bjork study, these participants too would not have
been able to appreciate the better learning being produced by the
inconsistent versus the consistent condition.
Finally, in
the last study that I want to share with you, McNamara, Kintsch,
Songer, and Kintsch (1996) introduced desirable difficulties into
their participants processing of text by creating two different
levels of coherency in a text about heart disease. Additionally
and interestingly, they also had participants with different levels
of background knowledge in the domain of biology read the two different
levels of text. They then tested their participants regarding the
text in a variety of ways by asking them different types of questions:
Some text-based and some requiring the making of inferences or the
solving of problems. Although a more complicated study than I am
describing now, the two hypotheses of relevance to the present discussion
were that (a) for both types of students, the consistent outline
should be better for the straight recall of text information, but
(b) for students with the requisite background knowledge, the text
with low coherence could be more beneficial than the test with high
coherence. Similar to the reasoning as to why the inconsistent outline
was beneficial for deeper learning, the idea behind the second hypothesis
was that such students may learn better when they have to provide
the coherence themselves (e.g., make the inferences and provide
the explanatory connections that are not explicitly provided in
the text, thus integrating the information in the text with the
information they already have stored in long-term memory.) In contrast,
students without the requisite background knowledge would not be
able to make the necessary inferences nor fill in the gaps. For
them, then, the low coherence in the text would not be a desirable
difficulty, as it would present them with difficulties that they
would not be able to overcome.
As predicted,
for text-based recall of information, the high-coherence text was
found to be better for both high and low knowledge students. And,
also as hypothesized, for questions requiring problem solving or
the making of bridging inferences, the high-knowledge students did
profit from having to deal with the low coherent text. In contrast,
but as predicted, for the low-knowledge students, the low-coherence
text created difficulties that they could not overcome. Thus, for
them, the low coherency of the text was not a desirable difficulty.
Concluding
Comments
I hope in this
discussion, I have been able to convince you of the need for us
to take a new look at our own methods of instruction and how we
design and organize our courses with an eye for introducing desirable
difficulties for our students. In doing so, however, we need to
keep a few points in mind. First, we need to be mindful of how easy
it is for us, as instructors, to be misled regarding the optimal
conditions of instruction. In particular, we need to be wary of
preferring conditions that speed acquisition and seem to make the
learning process too easy, as these conditions may simply be propping
up the temporary performance of our students and not creating the
type of learning that can lead to long-term retention and transfer.
Furthermore, in making decisions regarding how to optimize the learning
of our students, we must keep in mind that we cannot rely on the
meta-cognitive reports of our students, who themselves—as
learners—are often misled into preferring non-optimal to optimal
conditions of learning. We want to introduce procedures that present
difficulties for the learner—in general, difficulties that
force the learner to be a more active participant in the acquisition
process. At the same time, however, we need to insure that the difficulties
we introduce are, in fact, desirable difficulties, that
is, ones that the learner is capable of overcoming.
References
Battig, W.
F. (1972). Intratask interference as a source of facilitation in
transfer and retention. In R. F. Thompson and J. F. Voss (Eds.),
Topics in learning and performance, pp. 131-159. New York:
Academic Press.
Bjork, R. A.
(1975). Retrieval as a memory modifier. In R. Solso (Ed.), Information
processing and cognition: The Loyola Symposium, pp. 123-144.
Hillsdale, NJ: Erlbaum.
Bjork, R. A.
(1994). Memory and metamemory considerations in the training of
human beings. In J. Metcalfe and A. Shimamura (Eds.), Metacognition:
Knowing about knowing (pp. 185-205). Cambridge, MA: MIT Press.
Bjork, R. A.
(1999). Assessing our own competence: Heuristics and and illusions.
In D. Gopher and A. Koriat (Eds.), Attention and Performance
XVII. Cognitive regulation of performance: Interaction of theory
and application (pp. 435-459). Cambridge, MA: MIT Press.
Bjork, R. A.,
& Bjork, E. L. (1992). A new theory of disuse and an old theory
of stimulus fluctuation. In A. Healy, S. Kosslyn, and R. Shiffrin
(Eds.), From learning processes to cognitive processes: Essays
in honor of William K. Estes, vol. 2, pp. 35-67. Hillsdale,
NJ: Erlbaum.
Carrier, M.,
& Pashler, H. (1992). The influence of retrieval on retention.
Memory & Cognition, 20, 633-642.
Christina,
R. A., & Bjork, R. A. (1991). Optimizing long-term retention
and transfer. In D. Druckman and R. A. Bjork (Eds.), In the
mind’s eye: Enhancing human performance, pp. 23-56. Washington,
DC: National Academy Press.
Cuddy, L. J.,
& Jacoby, L. L. (1982). When forgetting helps memory: Analysis
of repetition effects. Journal of Verbal Memory and Verbal Behavior,
21, 451-456.
Dempster, F.
N. (1996). Distributing and managing the conditions of encoding
and practice. In E. L. Bjork and R. A. Bjork (Eds.), Handbook
of Perception and Cognition. Vol. 10, Memory, pp. 317-344.
New York: Academic Press.
Estes, W. K.
(1955). Statistical theory of distributional phenomena in learning.
Psychological Review, 62, 369-377.
Jacoby, L.
L., Bjork, R. A., & Kelley, C. M. (1994). Illusions of comprehension
and competence. In D. Druckman and R. A. Bjork (Eds.), Learning,
Remembering, Believing: Enhancing Individual and Team Performance.
(pp. 57-80). Washington, DC: National Academy Press.
Mannes, S.
M., & Kintsch, W. (1987). Knowledge organization and text organization.
Cognition and Instruction, 4, 91-115.
McNamara, D.
S., Kintsch, E., Songer, N. B., & Kintsch, W. (1996). Are good
texts always better? Text coherence, background knowledge, and levels
of understanding in lerning from text. Cognition and Instruction,
14, 1-43.
Roediger, H.
L., & Karpicke, J. D. (2005) Test-enhanced learning: Taking
memory tests improves long-term retention. Manuscript submitted
for publication.
Schmidt, R.
A., & Bjork, R. A. (1992). New conceptualization of practice:
Common principles in three paradigms suggest new concepts for training.
Psychological Science, 3,
Shea, J. B.,
& Morgan, R. L. (1979). Contextual interference effects on the
acquisition, retention, and transfer of a motor skill. Journal
of Experimental Psychology: Human Learning and Memory, 5, 179-187.
Simon, D.,
& Bjork, R. A. (2001). Metacognition in motor learning.
Journal of Experimental Psychology: Learning, Memory, and Cognition,
27, 907-912.
Smith, S. M.,
& Rothkopf, E. Z. (1984). Contextual enrichment and distribution
of practice in the classroom. Cognition and Instruction,
1, 341-358.
Smith, S. M.,
Glenberg, A. M., & Bjork, B. A. (1978). Environmental context
and human memory. Memory & Cognition, 6, 342-353.
|
