Cognitive Psychology and Its Implications, Ch. 6
6
Human Memory: Encoding
and Storage
Past chapters have discussed how we perceive and encode what is in our present.
Now we turn to discussing memory, which is the means by which we can
perceive our past. People who lose the ability to create new memories become
effectively blind to their past. I would recommend the movie Memento as providing
a striking characterization of what it would like to have no memory. The protagonist
of the film, Leonard, has anterograde amnesia, a condition that prevents him from
forming new memories. He can remember his past up to the point of a terrible
crime that left him with amnesia, and he can keep track of what is in the immediate
present, but as soon as his attention is drawn to something else, he forgets what
has just happened. So, for instance, he is constantly meeting people he has met
before, who have often manipulated him, but he does not remember them, nor can
he protect himself from being manipulated further. Although Leonard incorrectly
labels his condition as having no short-term memory, this movie is an accurate
portrayal of anterograde amnesia—the inability to form new long-term memories. It
focuses on the amazing ways Leonard tries to connect the past with the immediate
present.
This chapter and the next can be thought of as being about what worked and
did not work for Leonard. This chapter will answer the following questions: • How do we maintain a short-term or working memory of what just happened?
This is what still worked for Leonard. • How does the information we are currently maintaining in working memory
prime knowledge in our long-term memory? • How do we create permanent memories of our experiences? This is what did
not work any more for Leonard. • What factors influence our success in creating new memories?
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 146
Memory and the Brain | 147
•Memory and the Brain
Throughout the brain, neurons are capable of changing in response to experience.
This neural plasticity provides the basis for memory. Although all of the
brain plays a role in memory, there are two regions, illustrated in Figure 6.1,
that have played the most prominent role in research on human memory. First,
there is a region within the temporal cortex that includes the hippocampus,
whose role in memory was already discussed in Chapter 1 (see Figure 1.7). The
hippocampus and surrounding structures play an important role in the storage
of new memories. This is where Leonard had his difficulties. Second, research
has found that prefrontal brain regions are strongly associated with both the
encoding of new memories and the retrieval of old memories. These are the same
regions that were discussed in Chapter 5 with respect to the meaningful encoding
of pictures and sentences. This area also includes the prefrontal region from
Figure 1.15 that was important in retrieval of arithmetic and algebraic facts.
With respect to the prefrontal regions shown in Figure 6.1, note that memory
research has found laterality effects similar to those noted at the beginning
of Chapter 5 (Gabrielli, 2001). Specifically, study of verbal material tends to involve
mostly the left hemisphere. Study of pictorial material, in contrast, tends
to involve sometimes the right hemisphere and sometimes both hemispheres.
Gabrielli notes that the left hemisphere tends to be associated with pictorial
material that is linked to verbal knowledge, such as pictures of famous people
or common objects. It is as if people are naming these objects to themselves.
Human memory depends heavily on frontal structures of the brain for the
creation and retrieval of memories and on temporal structures for the
permanent storage of these memories.
FIGURE 6.1 The brain
structures involved in the
creation and storage of
memories. Prefrontal regions
are responsible for the creation
of memories. The hippocampus
and surrounding structures
in the temporal cortex are
responsible for the permanent
storage of these memories.
Brain Structures
Prefrontal regions
that process information
for storage in posterior regions
Internal hippocampal
regions that store
memories permanently
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 147
•Sensory Memory Holds Information Briefly
Before beginning the discussion of more permanent memories, it is worth
noting the visual and auditory sensory memories that hold information briefly
when it first comes in.
Visual Sensory Memory
Many studies of visual sensory memory have used a procedure in which participants
are presented with a visual array of items, such as the letters shown in
Figure 6.2, for a brief period of time (e.g., 50 ms). When asked to recall the
items, participants are able to report three, four, five, or at most six items. One
might think that only this much material can be held in visual memory—yet
participants report that they were aware of more items but the items faded
away before they could attend to them and report them.
An important methodological variation on this task was introduced by
Sperling (1960). He presented arrays consisting of three rows of four letters.
Immediately after this stimulus was turned off, participants were cued to attend
to just one row of the display and to report only the letters in that row. The cues
were in the form of different tones (high for top row, medium for middle, and
low for bottom). Sperling’s method was called the partial-report procedure, in
contrast to the whole-report procedure, which was what had been used until
then. Participants were able to recall all or most of the items from a row of four.
Because participants did not know beforehand which row would be cued, Sperling
argued that they must have had most or all of the items stored in some sort
of short-term visual memory. Given the cue right after the visual display was
turned off, they could attend to that row in their short-term visual memory and
report the letters in that row. The reason participants could not report more
items in the full-report procedure was that these items had faded from this
memory before the participants could attend to them.
In the procedure just described, the tone cue was presented immediately
after the display was turned off. Sperling also varied the length of the delay
between the removal of the display and the tone. The results he obtained, in
terms of number of letters recalled, are presented in Figure 6.3. As the delay
increased to 1 s, the participants’ performances decayed back to what would be
expected from the original whole-report level of four or five items. That is,
participants were reporting about one-third as many items from the cued row
as they could report from three rows in the whole-report procedure.
Thus, it appears that the memory of the actual display decays very
rapidly and is essentially gone by the end of 1 s. All that is left after
that is what the participant has had time to attend to and convert to
a more permanent form.
Sperling’s experiments indicate the existence of a brief visual sensory
store—a memory system that can effectively hold all the information
in the visual display. While information is being held in this
store, a participant can attend to it and report it. This sensory store
appears to be particularly visual in character. In one experiment that
148 | Human Memory: Encoding and Storage
X
C
V
N K P
F L B
M R J
FIGURE 6.2 An example of
the kind of display used in
a visual-report experiment.
The display is presented briefly
to participants, who are then
asked to report the letters
it contains.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 148
demonstrated the visual character of the sensory store, Sperling (1967) varied
the postexposure field (the visual field after the display). He found that when the
postexposure field was light, the sensory information remained for only 1 s, but
when the field was dark, it remained for a full 5 s. Thus, a bright postexposure
field tends to “wash out’’ memory for the display. Moreover, following a display
with another display of characters effectively “overwrites’’ the first display and so
destroys the memory for the first set of letters. The brief visual memory revealed
in these experiments is sometimes called iconic memory. Unless information
in the display is attended to and processed further, it will be lost.
Auditory Sensory Memory
There is similar evidence for a brief auditory sensory store, which is sometimes
called echoic memory. There are behavioral demonstrations (e.g.,Moray,
Bates, & Barnett, 1965; Darwin, Turvey, & Crowder, 1972; Glucksberg & Cowan,
1972) of an auditory sensory memory, similar to Sperling’s visual sensory
memory, by which people can report an auditory stimulus with considerable
accuracy if probed for it soon after onset. One of the more interesting measures
of auditory sensory memory involves an ERP measure called the mismatch
negativity. When a sound is presented that is different from recently heard
sounds in such features as pitch or magnitude (or is a different phoneme), there
is an increase in the negativity of the ERP recording 150 to 200 ms after the
discrepant sound (for a review, read Näätänen, 1992). In one study, Sams,
Hari, Rif, and Knuutila (1993) presented one tone followed by another at various
intervals. When the second tone was different from the first, it produced a
Sensory Memory Holds Information Briefly | 149
0.0
Mean number of letters reported
Delay of tone (s)
0.2 0.4 0.6 0.8 1.0
FIGURE 6.3 Results from
Sperling’s experiment demonstrating
the existence of a brief
visual sensory store. Participants
were shown arrays consisting of
three rows of four letters. After
the display was turned off, they
were cued by a tone, either
immediately or after a delay,
to recall a particular one of the
three rows. The results show
that the number of items reported
decreased as the delay in
the cuing tone increased. (After
Sperling, 1960. Adapted by permission of
the publisher. © 1960 by Psychological
Monographs.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 149
mismatch negativity as long as the delay between the two tones was less than 10 s.
This indicates that an auditory sensory memory can last up to 10 s, consistent
with other behavioral measures. It appears that the source of this neural
response in the brain is at or near the primary auditory cortex. Similarly, it
appears that the information held in visual sensory memory is in or near the
primary visual cortex. Thus, these basic perceptual regions of the cortex hold
a brief representation of sensory information for further processing.
Sensory information is held briefly in cortical sensory memories so that
we can process it.
•A Theory of Short-Term Memory
A very important event in the history of cognitive psychology was the development
of a theory of short-term memory in the 1960s. It clearly illustrated the
power of the new cognitive methodology to account for a great deal of data in a
way that had not been possible with previous behaviorist theories. Broadbent
(1958) had anticipated the theory of short-term memory, and Waugh and
Norman (1965) gave an influential formulation of the theory. However, it was
Atkinson and Shiffrin (1968) who gave the theory its most systematic development.
It has had an enormous influence on psychology, and although few
researchers still accept the original formulation, similar ideas play a crucial role
in some of the modern theories that we will be discussing.
Figure 6.4 illustrates the basic theory. As we have just seen, information coming
in from the environment tends to be held in transient sensory stores from
which it is lost unless attended to. The theory of short-term memory proposed
that attended information went into an intermediate short-term memory where
it had to be rehearsed before it could go into a relatively permanent long-term
memory. Short-term memory had a limited capacity to hold information. At
one time, its capacity was identified with the memory span. Memory span
refers to the number of elements one can immediately repeat back. Ask a friend
to test your memory span. Have that friend make up lists of digits of various
lengths and read them to you. See how many digits you can repeat back. You
will probably find that you are able to remember no more than around seven or
eight perfectly. The size of the memory span was considered convenient in
those days when American phone numbers were seven digits. One view was
that short-term memory has room for about seven elements, although other
theorists (e.g., Broadbent, 1975) proposed that its capacity is smaller and that
memory span depends on other stores as well as short-term memory.
In a typical memory experiment, it was assumed that participants rehearsed
the contents of short-term memory. For instance, in a study of memory span,
participants might rehearse the digits by saying them over and over again to
themselves. It was assumed that every time an item was rehearsed, there was a
probability that the information would be transferred to a relatively permanent
long-termmemory. If the item left short-termmemory before a permanent longterm
memory representation was developed, however, it would be lost forever.
150 | Human Memory: Encoding and Storage
Sensory
store
Attention
Rehearsal
Short-term
memory
Long-term
memory
FIGURE 6.4 A model of
memory that includes an
intermediate short-term
memory. Information coming in
from the environment is held in
a transient sensory store from
which it is lost unless attended
to. Attended information goes
into an intermediate short-term
memory with a limited capacity
to hold information. The
information must be rehearsed
before it can move into a
relatively permanent long-term
memory.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 150
One could not keep information in short-term
memory indefinitely because new information
would always be coming in and pushing out
old information from the limited short-term
memory.
An experiment by Shepard and Teghtsoonian
(1961) is a good illustration of these ideas.
They presented participants with a long sequence
of 200 three-digit numbers. The task
was to identify when a number was repeated.
The investigators were interested in how participants’
ability to recognize a repeated number
changed as more numbers intervened between
the first appearance of the number and its repetition.
The number of intervening items is referred
to as the lag. If the participant tended to
keep only the most recent numbers in shortterm
memory, memory for the last few numbers
would be good but would get progressively worse as the numbers were pushed
out of short-term memory. The results are presented in Figure 6.5. Note that
recognition memory drops off rapidly over the first few numbers, but then the
drop-off slows to the point where it appears to be reaching some sort of asymptote
at about 60%. The rapid drop-off can be interpreted as reflecting the decreasing
likelihood that the numbers are being held in short-term memory. The
60% level of recall for the later numbers reflects the amount of information
that got into long-term memory.1
A critical assumption in this theory was that the amount of rehearsal controls
the amount of information transferred to long-term memory. For instance,
Rundus (1971) asked participants to rehearse out loud and showed that
the more participants rehearsed an item, the more likely they were to remember
it. Data of this sort were perhaps most critical to the theory of short-term memory
because they reflected the fundamental property of short-termmemory: It is
a necessary halfway station to long-term memory. Information has to “do time”
in short-term memory to get into long-term memory, and the more time done,
the more likely it is to be remembered.
In an influential article, Craik and Lockhart (1972) argued that what was critical
was not how long information is rehearsed, but rather the depth to which it is
processed. This theory, called depth of processing, held that rehearsal improves
memory only if the material is rehearsed in a deep and meaningful way. Passive
rehearsal does not result in better memory.A number of experiments have shown
that passive rehearsal results in little improvement in memory performance. For
instance, Glenberg, Smith, and Green (1977) had participants study a four-digit
number for 2 s, then rehearse a word for 2, 6, or 18 s, and then recall the four
digits. Participants thought that their task was to recall the digits and that they
A Theory of Short-Term Memory | 151
0 10 20 30 40 50 60
.9
.8
.7
.6
Lag
p(“old” old)
FIGURE 6.5 Results from
Shepard and Teghtsoonian’s
experiment demonstrating that
information cannot be kept in
short-term memory indefinitely
because new information will
always be coming in and
pushing out old information.
The probability of an “old”
response to old items is plotted
as a function of the number of
intervening presentations (the
lag) since the last presentation
of a stimulus. (From Shepard &
Teghtsoonian, 1961. Reprinted by
permission of the publisher. © 1961
by the American Psychological
Association.)
1 This level of memory is not really 60% because participants were incorrectly accepting more than 20% of
new items as repeated. The level of memory is really the difference between this 60% hit rate and the 20%
false alarm rate.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 151
were just rehearsing the word to fill the time. However, they were given a final
surprise test for the words. On average, participants recalled 11%, 7%, and 13%
of the words they had rehearsed for 2, 6, and 18 s. Their recall was poor and
showed little relationship to the amount of rehearsal.2 On the other hand, as we
saw in Chapter 5, participants’ memories can be greatly improved if they process
material in a deep and meaningful way. Thus, it seems that there may be no shortterm,
halfway station to long-term memory. Rather, it is critical that we process
information in a way that is conducive to setting up a long-term memory trace.
Information may go directly from sensory stores to long-termmemory.
Kapur et al. (1994) did a PET study of the difference between brain correlates
of the deep and shallow processing of words. In the shallow processing
task, participants judged whether the words contained a particular letter;
in the deep processing task, they judged whether the words described living
things. Even though the study time was the same, participants remembered
75% of the deeply processed words and 57% of the shallowly processed words.
Kapur et al. (1994) found that there was greater activation during deep processing
in the left prefrontal regions indicated in Figure 6.1. A number of subsequent
studies have also shown that this region of the brain is more active
during deep processing (for a review, see Wagner, Bunge, & Badre, 2004).
Atkinson and Shiffrin’s theory of short-term memory postulated that as
information is rehearsed in a limited-capacity short-term memory, it is
deposited in long-term memory; but what turned out to be important
was how deeply the material is processed.
•Working Memory Holds the Information
Needed to Perform a Task
Baddeley’s Theory of Working Memory
Baddeley (1986) proposed a theory of the rehearsal processes that did not tie
them to storage in long-term memory. He hypothesized that there are two systems,
a visuospatial sketchpad and a phonological loop, that are what he
called “slave systems” for maintaining information, and he speculated that there
might be more such systems. These systems compose part of what he calls
working memory, which is a system for holding information that we need to
perform a task. For instance, try multiplying 35 by 23 in your head. You may
find yourself developing a visual image of part of a written multiplication
problem (visuospatial sketchpad) and you may find yourself rehearsing partial
products like 105 (phonological loop). Figure 6.6 illustrates Baddeley’s overall
conception of how these various slave systems interact. A central executive
controls how the slave systems are used. The central executive can put information
into any of the slave systems or retrieve information from them. It can
152 | Human Memory: Encoding and Storage
2 Although recall memory tends not to be improved by the amount of passive rehearsal, Glenberg et al. (1977)
did show that recognition memory is improved by rehearsal. Recognition memory may depend on a kind
of familiarity judgment that does not require creation of new memory traces.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 152
also translate information from one system to
another. Baddeley claimed that the central executive
needs its own temporary store of information
to make decisions about how to control
the slave systems.
The phonological loop has received much
more extensive investigation than the visuospatial
sketchpad. Baddeley proposed that the
phonological loop consists of multiple components.
One is an articulatory loop by which an
“inner voice” rehearses verbal information. A classic example of how we use the
articulatory loop is in remembering a phone number.When one is told a phone
number, one rehearses the number over and over again to oneself until one dials
it. Many brain-imaging studies (see E. E. Smith & Jonides, 1995, for a review)
have found activation in Broca’s area (the region labeled “J” in the frontal portion
of the Figure 4.1 brain illustration) when participants are trying to remember a
list of items. This activation occurs even if the participants are not actually talking
to themselves. Patients with damage to this region show deficits in tests of shorttermmemory
(Vallar,Di Betta,& Silveri, 1997).
Another component of the phonological loop is the phonological store.
Baddeley proposed that this store is in effect an “inner ear” that hears the inner
voice and stores the information in a phonological form. It has been proposed
that this region is associated with the parietal-temporal region of the brain (the
region labeled “J” in the parietal-temporal region of the Figure 4.1 brain illustration).
A number of imaging studies (Henson, Burgess, & Frith, 2000; Jonides
et al., 1998) have found activation of this region during the storage of verbal
information. Also, patients with lesions in this region suffer deficits of shortterm
memory (Vallar et al., 1997).
One of the most compelling pieces of evidence for the existence of the articulatory
loop is the word-length effect (Baddeley, Thomson, & Buchanan, 1975).
Read the five words below and then try to repeat them back without looking
at the page:
• wit, sum, harm, bay, top
Most people can do this. Baddeley et al. found that participants were able to
repeat back an average of 4.5 words out of 5 such one-syllable words. Now read
and try to repeat back the following five words:
• university, opportunity, aluminum, constitutional, auditorium
Participants were able to recall only an average of 2.6 words out of 5 such fivesyllable
words. The crucial factor appears to be how long it takes to say the word.
Vallar and Baddeley (1982) looked at recall for words that varied from one to
five syllables. They also measured how many words of the various lengths participants
could say in a second. Figure 6.7 shows the results. Note that the percentage
of sequences correctly recalled almost exactly matches the reading rate.
Trying to maintain information in working memory is analogous to the
circus act that involves spinning plates on a stick. The circus performer will get
one plate spinning on one stick, then another on another stick, then another,
Working Memory Holds the Information Needed to Perform a Task | 153
Central
executive Phonological
loop
Visuospatial
sketchpad
FIGURE 6.6 Baddeley’s theory
of working memory in which a
central executive coordinates
a set of slave systems. (From
Baddeley, 1986. Reprinted by permission
of the publisher. © 1986 by Oxford
University Press.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 153
and so on. Then he runs back to the first before it slows
down and falls off. He respins it and then respins the rest.
He can keep only so many plates spinning at the same time.
Baddeley proposed that it is the same situation with respect
to working memory. If we try to keep too many items in
working memory, by the time we get back to rehearse the
first one, it will have decayed to the point that it takes too
long to retrieve and re-rehearse. Baddeley proposed that
we can keep about 1.5 to 2.0 seconds’ worth of material
rehearsed in the articulatory loop.
There is considerable evidence that this articulatory
loop truly involves speech. For instance, the research of
R. Conrad (1964) showed that participants suffered more
confusion when they tried to remember spans that had a
high proportion of rhyming letters (such as BCTHVZ) than
when they tried to remember spans that did not (such as
HBKLMW). Also, as we just discussed, there is evidence for
activation in Broca’s area, part of the left prefrontal cortex,
during the rehearsal of such memories.
One might wonder what the difference is between
short-term memory and Baddeley’s articulatory loop. The crucial difference is
that processing information in the phonological loop is not critical to getting
it into long-term memory. Rather, the phonological loop is just an auxiliary
system for keeping information available.
Baddeley proposed that we have an articulatory loop and a visuospatial
sketchpad, both of which are controlled by a central executive, which are
systems for holding information and are part of working memory.
The Frontal Cortex and Primate Working Memory
The frontal cortex gets larger in the progression from lower mammals, such as
the rat, to higher mammals, such as the monkey; and it shows a proportionately
greater development between the monkey and the human. It has been
known for some time that the frontal cortex plays an important role in tasks
that can be thought of as working-memory tasks in primates. The task that has
been most studied in this respect is the delayed match-to-sample task, which
is illustrated in Figure 6.8. The monkey is shown an item of food that is placed
in one of two identical wells (Figure 6.8a). Then the wells are covered, and the
monkey is prevented from looking at the scene for a delay period—typically 10 s
(Figure 6.8b). Finally, the monkey is given the opportunity to retrieve the
food, but it must remember in which well it was hidden (Figure 6.8c). Monkeys
with lesions in the frontal cortex cannot perform this task (Jacobsen,
1935, 1936). A human infant cannot perform similar tasks until its frontal cortex
has matured somewhat, usually at about 1 year of age (Diamond, 1991).
When a monkey must remember where a food item has been placed, a
particular area of the frontal cortex is involved (Goldman-Rakic, 1988). This
154 | Human Memory: Encoding and Storage
1 2 3
Number of syllables
Items correct (%)
% correct
Reading rate
Reading rate (words/s)
4 5
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
2.5
2.3
10
20
30
40
50
60
70
80
90
100
FIGURE 6.7 Results of
Vallar and Baddeley’s (1982)
experiment showing the
existence of the articulatory
loop. Mean reading rate and
percentage of correct recall
of sequences of five words are
plotted as a function of word
length. (From Baddeley, 1986. Reprinted
by permission of the publisher. © 1986
by Oxford University Press.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 154
small region, called area 46 (Figure 6.9), is found on the side of the frontal
cortex. Lesions in this specific area produce deficits in this task. It has been
shown that neurons in this region fire only during the delay period of the task,
as if they are keeping information active during that interval. They are inactive
before and after the delay.Moreover, different neurons in that region seem tuned
to remembering objects in different portions of the visual field (Funahashi,
Bruce,& Goldman-Rakic, 1991).
Goldman-Rakic (1992) examined monkey performance on other tasks that
require maintaining different types of information over the delay interval. In
one task, monkeys had to remember different objects. For example, the animal
would have to remember to select a red circle, and not a green square, after a
delay interval. It appears that a different region of the prefrontal
cortex is involved in this task. Different neurons in
this area fire when a red circle is being remembered than
when a green square is being remembered. Goldman-Rakic
speculated that the prefrontal cortex is parceled into many
small regions, each of which is responsible for remembering
a different kind of information.
Like many neuroscience studies, these experiments are
correlational—they show a relationship between neural activity
and memory function, but they do not show that the
neural activity is essential for the memory function. In an
effort to show a causal role, Funahashi, Bruce, and Goldman-
Rakic (1993) trained monkeys to remember the location of
objects in their visual field and move their eyes to these
locations after a delay—an oculomotor equivalent of the
task described in Figure 6.8. They selectively lesioned this
Working Memory Holds the Information Needed to Perform a Task | 155
(a) Cue (b) Delay (c) Response
Wrong Right
10
10 12
11 47
45
45
44
8A
8B
46
46
FIGURE 6.8 An illustration of the delayed match-to-sample task. (a) Food is placed in the well
on the right and covered. (b) A curtain is drawn for the delay period. (c) The curtain is raised,
and the monkey can lift the cover from one of the wells. (From Goldman-Rakic, 1987. Reprinted by
permission of the publisher. © 1987 by the American Physiological Society.)
FIGURE 6.9 Lateral views
of the cerebral cortex of a
human (top) and of a monkey
(bottom). Area 46 is the region
shown in darker color. (From
Goldman-Rakic, 1987. Reprinted by
permission of the publisher. © 1987
by the American Physiological Society.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 155
area of the prefrontal cortex in the left hemisphere. They found that monkeys
were no longer able to remember the locations in the right visual field (recall
from Chapter 2 that the left visual field projects to the right hemisphere; see
Figure 2.5), but their ability to remember objects in the left visual field was
unimpaired. When they lesioned the right hemisphere region, their ability to
remember the location of objects in the left visual field was also impacted.
Thus, it does seem that activity in these prefrontal regions is critical to the
ability to maintain these memories over delays.
E. E. Smith and Jonides (1995) used PET scans to see whether there are similar
areas of activation in humans. When participants held visual information
in working memory, there was activation in right prefrontal area 47, which is
adjacent to area 46. The monkey brain and the human brain are not identical
(see Figure 6.8), and we would not necessarily expect a direct correspondence
between regions of their brains. Smith and Jonides also looked at a task in
which participants rehearsed verbal labels; they found that left prefrontal area 6
was active in this task. This region of the prefrontal cortex is associated with
linguistic processing (Petrides, Alvisatos, Evans, & Meyer, 1993). One might
view these two tasks as invoking Baddeley’s two slave systems, with the visuospatial
sketchpad associated with right prefrontal regions and the phonological
loop associated with left prefrontal regions.
Different areas of the frontal cortex appear to be responsible for maintaining
different types of information in working memory.
•Activation and on Long-Term Memory
So far, we have discussed how information from the environment comes into
working memory and is maintained by rehearsal. There is another source of
information besides the environment, however: long-term memory. For instance,
rather than reading a phone number and holding it in working memory,
we can retrieve a familiar number and hold it in working memory. A number
of theories have assumed that different pieces of information in long-term
memory can vary from moment to moment in terms of how easy it is to retrieve
them into working memory. Various theories use different words to describe the
same basic idea. The language I use in this chapter is similar to that used in my
ACT (AdaptiveControl of Thought) theory (J. R.Anderson, 1983; J. R.Anderson
& Lebiere, 1998). In ACT, one speaks of memory traces as varying in their
activation. Another well-known theory, SAM (Search of Associative Memory)
(Gillund & Shiffrin, 1984; Raaijmakers & Shiffrin, 1981), speaks of images
(memory traces) as varying in their familiarity (activation).
An Example of Activation Calculations
Activation determines both the probability and the speed of access to memory.
The free-association technique is sometimes used to get at levels of activation
in memory. Whatever ideas come to mind as you are free-associating can be
156 | Human Memory: Encoding and Storage
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 156
taken as reflecting the things that are currently most active in your long-term
memory.What do you think of when you read the three words below?
Bible
animals
flood
If you are like the students in my classes, you will think of the story of Noah.
The curious fact is that when I ask students to associate to just the word Bible,
they come up with such terms as Moses and Jesus—almost never Noah. When
I ask them to associate to just animals, they come up with farm and zoo, but
almost never Noah; and when I ask them to associate to just flood, they come up
with Mississippi and Johnstown (the latter being perhaps a Pittsburgh-specific
association), but almost never Noah. So why do they come up with Noah when
given all three terms together? Figure 6.10 represents this phenomenon in terms
of activation computations and shows three kinds of things:
1. Various words that might come to mind, such as Jesus,Moses, and
Mississippi.
2. Various terms that might be used to prime memory, such as Bible,
animals, and flood.
3. Associations between the primes and the responses. In this illustration,
these associations are indicated by the triangular connections where the
input line touches the output line.
Activation and on Long-Term Memory | 157
Strengths of association (Sji )
Baseline activation
(Bi )
Input weight
(Wj )
2 2
1 Noah
Moses
Jesus
Farm
Zoo
Mississippi
Johnstown
1 0 1 0 0 1 0 0
Bible Animals Flood
FIGURE 6.10 A representation
of how activation accumulates
in a neural network such as that
assumed in the ACT theory.
Activation coming from various
stimulus words—such as Bible,
animals, and flood—spreads
activation to associated
concepts, such as Noah,
Moses, and farm.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 157
The ACT theory has an equation to represent how the activation, Ai , of any
particular element i reflects the structure and circumstance of this network:
This equation relates the activation Ai to the three components in the network:
1. Bi reflects the base-level activation of the words that might be recalled.
Some concepts, such as Jesus and Mississippi, are more common than
others, such as Noah, and so would have greater base-level activation.
Just to be concrete, we might assign the base-level activations for Jesus
and Mississippi to be 3 and for Noah to be 1.
2. Wj reflects the weight given to various terms j that might prime the
memory. For instance, we might assume that the Wj for any word we
present is 1 and the Wj for words we do not present is 0. The indicates
that we are summing over all of the potential primes j.
3. Sji reflects the strength of associations between potential primes j in item
2 above and potential responses i in item 1 above. To keep things simple,
we might assume that the strength of association is 2 in the case of
related pairs such as Bible-Jesus and flood-Mississippi and 0 in the case
of unrelated pairs such as Bible-Mississippi and flood-Jesus.
With this equation, these concepts, and these numbers, we can explain why the
students in my class associate Noah when prompted with all three words but
never do so when presented with any word individually. Consider what happens
when I present just the word Bible. There is only one prime with a positive
Wj, and this is Bible. In this case, the activation of Noah is
ANoah _ 1 _ (1 _ 2) _ 3
where the first 1 is its base-level activation Bi, the second 1 is its weight Wj of
Bible, and the 2 is the strength of association Sji between Bible and Noah. The
activation for Jesus is different because it has a higher base-level activation,
reflecting its greater frequency:
AJesus _ 3 _ (1 _ 2) _ 5
The reason Jesus and not Noah comes to mind in this case is that Jesus has
higher activation. Now let’s consider what happens when I present all three
words. The activation of Noah will be
ANoah _ 1 _ (1 _ 2) _ (1 _ 2) _ (1 _ 2) _ 7
where there are three (1 _ 2)’s because all three of the terms—Bible, animals,
and flood—have associations to Noah. The activation equation for Jesus remains
AJesus _ 3 _ (1 _ 2) _ 5
because only Bible has the association with Jesus. Thus, the extra associations to
Noah have raised the current activation of Noah to be greater than the activation
of Jesus, despite the fact that it has lower base-level activation.
©
Ai = Bi + a
j
WjSji
158 | Human Memory: Encoding and Storage
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 158
There are two critical factors in this activation equation: the base-level activation,
which sets a starting activation for the idea, and the activation received
through the associations, which adjusts this activation to reflect the current
context. The next section will explore this associative activation, and the section
after that will discuss the base-level activation.
The speed and probability of accessing a memory are determined by the
memory’s level of activation, which in turn is determined by its base-level
activation and the activation it receives from associated concepts.
Spreading Activation
Spreading activation is the term often used to refer to the process by which currently
attended items can make associated memories more available.Many studies
have examined how terms to which we currently attend can prime memories.
One of the earliest was a study by Meyer and Schvaneveldt (1971) in which participants
were asked to judge whether or not pairs of items were words. Table 6.1
shows examples of the materials used in their experiments, along with participants’
judgment times. The items were presented one above the other. If either
item in a pair was not a word, participants were to respond no. It appears from
examining the negative pairs that participants judged first the top item and then
the bottom item.When the top item was not a word, participants were faster to
reject the pair than when only the bottom item was not a word. (When the top
item was not a word, participants did not have to judge the bottom item and so
could respond sooner.) The major interest in this study was in the positive
pairs. There were unrelated items, such as nurse and butter, and pairs with an
associative relation, such as bread and butter. Participants were 85 ms faster on
the related pairs. This result can be explained by a spreading-activation analysis.
When the participant read the first word in the related pair, activation would
spread from it to the second word. This would make information about the
spelling of the second word more active and make that word easier to judge.
The implication of this result is that the associative spreading of information
activation through memory can facilitate the rate at which words are read.
Activation and on Long-Term Memory | 159
TABLE 6.1
Examples of the Pairs Used to Demonstrate Associative Priming
Positive Pairs Negative Pairs
Nonword Nonword Both
Unrelated Related First Second Nonwords
Nurse Bread Plame Wine Plame
Butter Butter Wine Plame Reab
940 ms 855 ms 904 ms 1,087 ms 884 ms
From Meyer and Schvaneveldt, 1971. Reprinted by permission of the publisher. © 1971 by the Journal
of Experimental Psychology.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 159
Thus, we can read material that has a strong associative coherence more rapidly
than we can read incoherent material in which the words are unrelated.
Kaplan (1989), in his dissertation research, reported an effect of associative
priming at a very different time scale of information processing. The “participants”
in the study were members of his dissertation committee. I was one of
these participants, and it was a rather memorable and somewhat embarrassing
experience. He gave us riddles to solve, and each of us was able to solve about
half of them. One of the riddles that I was able to solve was
What goes up a chimney down but can’t come down a chimney up?
The answer is umbrella. Another faculty member was not able to solve this
one, and he has his own embarrassing story to tell about it—much like the one
I have to tell about the following riddle that I could not get:
On this hill there was a green house. And inside the green house there was a
white house. And inside the white house, there was a red house. And inside
the red house there were a lot of little blacks and whites sitting there. What
place is this?
More or less randomly, different faculty members were able to solve various
riddles.
Then Kaplan gave us each a microphone and tape recorder and told us that
we would be beeped at various times over the next week and that we should then
record what we had thought about the unsolved riddles and whether we had
solved any new ones. He said that he was interested in the steps by which we
came to solve these problems. That was essentially a lie to cover the true purpose
of the experiment, but it did keep us thinking about the riddles over the week.
What Kaplan had done was to split the riddles each of us could not solve
randomly into two groups. For half of these unsolved problems, he seeded our
environment with clues to the solution. He was quite creative in how he did
this: In the case of the riddle above that I could not solve, he drew a picture of a
watermelon as graffiti in the men’s restroom. Sure enough, shortly after seeing
this graffiti I thought again about this riddle and came up with the answer—
watermelon! I congratulated myself on my great insight, and when I was next
beeped I proudly recorded how I had solved the problem—quite unaware of
the role the bathroom graffiti had played in my solution.
Of course, that might just be one problem and one foolish participant.
Averaged over all the problems and all the participants (which included a
Nobel Laureate), however, we were twice as likely to solve a problem that had
been primed in the environment than one that had not been. Basically, activation
from the picture of the watermelon was spreading to my knowledge about
watermelons and priming it when I thought about this problem. We were all
unaware of the manipulation that was taking place. This example illustrates the
importance of priming to issues of insight (a topic we will consider at length in
Chapter 8) and also shows that one is not aware of the associative priming that
is taking place, even when one is trained to spot such things, as I am.
Activation spreads from presented items through a network to memories
related to that prime item.
160 | Human Memory: Encoding and Storage
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 160
•Practice and Memory Strength
Spreading activation concerns how the context can make some memories more
available. However, some memories are just more available because they are
used frequently in all contexts. So, for instance, you can recall the names of
close friends almost immediately, anywhere and anytime. The quantity that
determines this inherent availability of a memory is referred to as its strength.
In contrast to the activation level of a trace, which can have rapid fluctuations
depending on whether associated items are being focused upon, the strength of
a trace changes more gradually. Each time we use a memory trace, it increases a
little in strength. The strength of a trace determines in part how active it can
become and hence how accessible it will be. The strength of a trace can be gradually
increased by repeated practice.
The Power Law of Learning
The effects of practice on memory retrieval are extremely regular and very
large. In one study, Pirolli and Anderson (1985) taught participants a set of
facts and had them practice the facts for 25 days; then they looked at the speed
with which the participants could recognize these facts. Figure 6.11a plots how
participants’ time to recognize a fact decreased with practice. As can be seen, participants
sped up from about 1.6 s to 0.7 s, cutting their retrieval time by more
than 50%. The illustration also shows that the rate of improvement decreases
Practice and Memory Strength | 161
Recognition time (s)
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0 10 20 30
Days of practice
0.5
(1.5)
(1.0)
(0.5)
0.3
0.1
– 0.1
– 0.3
– 0.5
– 0.7
0 1 2 3 4
(1) (5) (25)
Log (days of practice)
Log (recognition time)
(a) (b)
FIGURE 6.11 Results of Pirolli and Anderson’s study to determine the effects of practice
on recognition time. (a) The time required to recognize sentences is plotted as a function
of the number of days of practice. (b) The data in (a) are log–log transformed to reveal a
power function. The data points are average times for individual days, and the curves are
the best-fitting power functions. (From Pirolli & Anderson, 1985. Reprinted by permission of the publisher.
© 1985 by the Journal of Experimental Psychology: Learning, Memory, and Cognition.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 161
with more practice. Increasing practice has diminishing returns. The data are
nicely fit by a power function of the form
T _ 1.40 P_0.24
where T is the recognition time and P is the number of days of practice. This
is called a power function because the amount of practice P is being raised
to a power. This power relationship between performance (measured in terms
of response time and several other variables) and amount of practice is a ubiquitous
phenomenon in learning. One way to see that data correspond to a
power function is to plot logarithm of time (the y-axis) against the logarithm of
practice (the x-axis). If we have a power function in normal coordinates, we
should get a linear function in log–log coordinates:
ln T _ 0.34 _ 0.24 ln P
Figure 6.11b shows the data so transformed. As can be seen, the relationship
is quite close to a linear function (straight line).
Newell and Rosenbloom (1981) refer to the way that memory performance
improves as a function of practice as the power lawof learning. Figure 6.12 shows
some data from Blackburn (1936), who looked at the effects of practicing addition
problems for 10,000 trials by two participants. The data are plotted in log–log
terms, and there is a linear relationship. On this graph and on some others in this
book, the original numbers (i.e., those given in parentheses in Figure 6.11b) are
plotted on the logarithmic scale rather than being expressed as logarithms. Blackburn’s
data show that the power law of learning extends to amounts of practice
far beyond that shown in Figure 6.11. Figures 6.11 and 6.12 reflect the gradual
162 | Human Memory: Encoding and Storage
2 2
0.5
1.0
2.0
5.0
5 10 20
Problem number (logarithmic scale)
Time (s) (logarithmic scale)
50 100 200 500 1000 10,000
FIGURE 6.12 Data from Blackburn’s study on the effects of practicing addition problems for
10,000 trials. The results are presented as improvement with practice in the time taken to add
two numbers. Data are plotted separately for two participants. Both the time required to solve
the problem and the number of problems are plotted on a logarithmic scale. (Plot by Crossman, 1959,
of data from Blackburn, 1936. Adapted by permission of the publisher. © 1959 by Ergonomics.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 162
increase in memory trace strength with practice. As memory traces become
stronger, they can reach higher levels of activation and so can be retrieved more
rapidly.
As a memory is practiced, it is strengthened according to a power function.
Neural Correlates of the Power Law
One might wonder what really underlies the power law of practice. Some evidence
suggests that the law may be related to basic neural changes involved in
learning. One kind of neural learning that has attracted much attention is called
long-term potentiation (LTP), which occurs in the hippocampus and cortical
areas. This form of neural learning seems to be related to behavioral measures
of learning like those in Figures 6.11 and 6.12. When a pathway is stimulated
with a high-frequency electric current, cells along that pathway show increased
sensitivity to further stimulation. Barnes (1979) looked at this phenomenon in
rats by measuring the percentage increase in excitatory postsynaptic potential
(EPSP) over its initial value.3 Barnes stimulated the hippocampus of her rats
each day for 11 successive days and measured the growth in LTP as indicated by
the percentage increase in EPSP. Figure 6.13a displays her results in a plot of percentage
change in LTP versus days of practice. There appears to be a diminishing
increase with amount of practice. Figure 6.13b, which plots log percentage
change in LTP against log days of practice, shows that the relationship is approximately
linear, and therefore the relationship in Figure 6.13 is approximately a
power function. Thus, it does seem that neural activation changes with practice,
just as behavioral measures do.
Note that the activation measure shown in Figure 6.13a increases more
and more slowly, whereas recognition time (see Figure 6.11a) decreases more and
more slowly. In other words, a performance measure such as recognition time
is an inverse reflection of the growth of strength that is happening internally.
As the strength of the memory increases, the performance measures improve
(which means shorter recognition times and fewer errors). You remember something
faster after you’ve thought about it more often.
The hippocampal region being observed here is the area that was damaged
in the fictional Leonard character discussed at the beginning of the chapter.
Damage to this region often results in amnesia. Studies of the effects of practice
on participants without brain damage have found that activation in the hippocampus
and the prefrontal regions decreases as participants become more
practiced at retrieving memories (Kahn & Wagner, 2002).4
Practice and Memory Strength | 163
3 As discussed in Chapter 1, the difference in electric potential between the outside and inside of the cell
decreases as the dendrite and cell body of a neuron become more excited. EPSP is described as increasing
when this difference decreases.
4 Note that neural activation decreases with practice because it takes less effort to retrieve the memory. This
can be a bit confusing—greater trace activation resulting from practice results in lower brain activation. This
happens because trace activation reflects the availability of the memory, whereas brain activation reflects
the hemodynamic expenditure required to retrieve the memory. Trace activation and brain activation refer
to different concepts.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 163
The relationship between the hippocampus and regions of the prefrontal
cortex is interesting. In normal participants, these regions are often active at the
same time, as they were in the Kahn and Wagner study. It is generally thought
(e.g., Paller & Wagner, 2002) that processing activity in prefrontal regions
regulates input to hippocampal regions that store the memories. Patients with
hippocampal damage show the same prefrontal activation as normal people do;
but because of the hippocampal damage, they fail to store these memories
(R. L. Buckner, personal communication, 1998).
Two studies illustrating the role of the prefrontal cortex in forming new
memories appeared back to back in the same issue of Science magazine. One
study (Wagner et al., 1998) investigated memory for words; the other (J. B. Brewer
et al., 1998) studied memory for pictures. In both cases, participants remembered
some of the items and forgot others. Using fMRI measures of the hemodynamic
response, the researchers contrasted the brain activation at the time of study for
those words that were subsequently remembered and those that were subsequently
forgotten.Wagner et al. found that left prefrontal regions were predictive
of memory for words, whereas Brewer et al. found that right prefrontal regions
were predictive of memory for pictures. Figure 6.14a shows the results for words;
Figure 6.14b, the results for pictures. In both cases, the rise in the hemodynamic
response is plotted as a function of time from stimulus presentation.As discussed
in Chapter 1, the hemodynamic response lags, so that it is at maximum about 5 s
after the actual neural activity. The correspondence between the results from the
two laboratories is striking. In both cases, remembered items received greater
164 | Human Memory: Encoding and Storage
Change (%)
50
40
30
20
10
0 2 4 6 8 10
Days of practice
Log (change)
3.8
3.6
3.4
3.2
3.0
2.8
2.6
0 1 2 3
(a) (b) Log (days of practice)
FIGURE 6.13 Results from Barnes’s study of long-term potentiation (LTP) demonstrating that
when a neural pathway is stimulated, cells along that pathway show increased sensitivity to
further stimulation. The growth in LTP is plotted as a function of number of days of practice
(a) in normal scale and (b) in log–log scale. (From Barnes, 1979. Reprinted by permission of the publisher.
© 1979 by the Journal of Comparative Physiology.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 164
Practice and Memory Strength | 165
activation from the prefrontal regions, supporting the conclusion that prefrontal
activation is indeed critical for storing a memory successfully.5 Also, note that
these studies are a good example of the lateralization of prefrontal processing,
with verbal material involving the left hemisphere to a greater extent and visual
material involving the right hemisphere to a greater extent.
Activation in prefrontal regions appears to drive long-term potentiation in
the hippocampus. This activation results in the creation and strengthening
of memories.
(a)
100
90
70
80
60
50
40
30
20
10
2 4 6 8 10 12
Remembered
Forgotten
Time from stimulus (s)
Hemodynamic response (% of maximum)
(b)
100
90
80
70
60
50
40
30
20
10
2 4
Time from stimulus (s)
Hemodynamic response (% of maximum)
6 8 10 12
Remembered
Forgotten
FIGURE 6.14 Results from two studies illustrating the role of the prefrontal cortex in forming
new memories. (a) Data from the study by Wagner et al. show the rise in the hemodynamic
response in the left prefrontal cortex while participants studied words that were subsequently
remembered or forgotten. (After Wagner et al., 1998. Adapted by permission of the publisher. © 1998 by Science.)
(b) Data from the study by Brewer et al. show the rise in the hemodynamic response in the right
prefrontal cortex while participants studied pictures that were subsequently remembered or
forgotten. (After J. B. Brewer et al., 1998. Adapted by permission of the publisher. © 1998 by Science.)
5 Greater hemodynamic activation at study results in a stronger memory—which, as we noted, can lead
to reduced hemodynamic activation at test.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 165
•Factors Influencing Memory
A reasonable inference from the preceding discussion might be that the only thing
determining memory performance is how much we study and practice the material
to be remembered. However, we earlier reviewed some of the evidence that
mere study of material will not lead to better recall.How one processes the material
while studying it is important. We saw in Chapter 5 that more meaningful
processing of material results in better recall. Earlier in this chapter, with respect
to Craik and Lockhart’s (1972) depth-of-processing proposal, we reviewed the
evidence that shallow study results in little memory improvement. As a different
demonstration of the same point, D. L. Nelson (1979) had participants read
paired associates that were either semantic associates (e.g., tulip-flower) or rhymes
(e.g., tower-flower). Better memory (81% recall) was obtained for the semantic
associates than for the rhymes (70% recall). Presumably, participants tended to
process the semantic associates more meaningfully than the rhymes. In Chapter 5,
we also saw that people retain more meaningful information better. In this section,
we will review some other factors, besides depth of processing and meaningfulness
of the material, that determine our level of memory.
Elaborative Processing
There is evidence that more elaborative processing results in better memory.
Elaborative processing involves creating additional information that relates
and expands on what it is that needs to be remembered. J. R. Anderson and
Bower (1973) did an experiment to demonstrate the importance of elaboration.
They had participants try to remember simple sentences such as The doctor
hated the lawyer. In one condition, participants just studied the sentence; in the
other, they were asked to generate an elaboration of their choosing—such as
because of the malpractice suit. Later, participants were presented with the subject
and verb of the original sentence (e.g., The doctor hated) and were asked to
recall the object (e.g., the lawyer). Participants who just studied the original sentences
were able to recall 57% of the objects, but those who generated the elaborations
recalled 72%. The investigators proposed that this advantage resulted
from the redundancy created by the elaboration. If the participants could not
originally recall lawyer but could recall the elaboration because of the malpractice
suit, they might then be able to recall lawyer.
A series of experiments by B. S. Stein and Bransford (1979) showed why selfgenerated
elaborations are often better than experimenter-provided elaborations.
In one of these experiments, participants were asked to remember 10 sentences,
such as The fat man read the sign. There were four conditions of study.
• In the base condition, participants studied just the sentence. • In the self-generated elaboration condition, participants were asked to
create an elaboration of their own. • In the imprecise elaboration condition, participants were given a continuation
of the sentence poorly related to the meaning of the sentence, such
as that was two feet tall. • In the precise elaboration condition, they were given a continuation that
gave context to the sentence, such as warning about the ice.
166 | Human Memory: Encoding and Storage
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 166
After studying the material, participants in all conditions were presented with
such sentence frames as The _______ man read the sign, and they had to recall
the missing adjective. Participants recalled 4.2 of the 10 adjectives in the base
condition and 5.8 of the 10 when they generated their own elaborations.
Obviously, the self-generated elaborations had helped. They could recall only
2.2 of the adjectives in the imprecise elaboration condition, replicating the
typical inferiority found for experimenter-provided elaborations relative to selfgenerated
ones. However, participants recalled the most (7.8 of 10 adjectives) in
the precise elaboration condition. So, by careful choice of words, experimenter
elaborations can be made better than those of participants. (For further research
on this topic, read Pressley,McDaniel, Turnure,Wood, & Ahmad, 1987.)
It appears that the critical factor is not whether the participant or the experimenter
generates the elaborations but whether the elaborations prompt
the material to be recalled. Participant-generated elaborations are effective because
they reflect the idiosyncratic constraints of each particular participant’s
knowledge. As Stein and Bransford demonstrated, however, it is possible for
the experimenter to construct elaborations that facilitate even better recall.
Memory for material improves when it is processed with more meaningful
elaborations.
•Techniques for Studying Textual Material
Frase (1975) found evidence of the benefit of elaborative processing with text material.
He compared how participants in two groups remembered text: One group
was given questions to think about before reading the text—sometimes called
advance organizers (Ausubel, 1968)—and a control group that simply studied the
text without advance questions. Participants in the first group were asked to find
answers to the advance questions as they read the text. This requirement should
have forced them to process the text more carefully and to think about its implications.
In a subsequent test, the advance-organizer group answered 64% of the
questions correctly, whereas the control group answered only 57% correctly. The
questions in the test were either relevant or irrelevant to the advance organizers.
For instance, a test question about an event that precipitated America’s entry into
WorldWar II would be considered relevant if the advance questions directed participants
to learn why America entered the war. A test question would be considered
irrelevant if the advance questions directed participants to learn about the
economic consequences of WorldWar II. The advance-organizer group correctly
answered 76% percent of the relevant questions and 52% of the irrelevant ones.
Thus, they did only slightly worse than the control group on topics for which they
had been given only irrelevant advance questions but did much better on topics
for which they had been given relevant advance questions.
Many college study-skills departments, as well as private firms, offer courses
designed to improve students’ memory for text material. These courses teach
study techniques mainly for texts such as those used in the social sciences, not
for the denser texts used in the physical sciences and mathematics or for literary
Techniques for Studying Textual Material | 167
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 167
materials such as novels. The study techniques from different programs are
rather similar, and their success has been fairly well documented. One example
of such a study technique is the PQ4R method (Thomas & Robinson, 1972).
The Implications box in Chapter 1 described a slight variation on this technique
as a method for studying this book.
The PQ4R method derives its name from the six phases it advocates for
studying a chapter in a textbook:
1. Preview: Survey the chapter to determine the general topics being
discussed. Identify the sections to be read as units. Apply the next
four steps to each section.
2. Questions:Make up questions about each section. Often, simply
transforming section headings results in adequate questions.
3. Read: Read each section carefully, trying to answer the questions you
have made up about it.
4. Reflect: Reflect on the text as you are reading it. Try to understand it, to
think of examples, and to relate the material to your prior knowledge.
5. Recite: After finishing a section, try to recall the information contained
in it. Try to answer the questions you made up for the section. If you
cannot recall enough, reread the portions you had trouble remembering.
6. Review: After you have finished the chapter, go through it mentally,
recalling its main points. Again try to answer the questions you made up.
The central features of the PQ4R technique are the generation and answering of
questions. There is reason to think that the most important aspect of these features
is that they encourage deeper and more elaborative processing of the text material.
At the beginning of this section, we reviewed the Frase (1975) experiment that
demonstrated the benefit of reading a text with a set of advance questions in mind.
It seems that the benefit was specific to test items related to the questions.
Another experiment by Frase (1975) compared the effects of making up
questions with the effects of answering them. He asked pairs of participants to
study a text passage that was divided into halves. For one half, one participant in
the pair read the passage and made up study questions during the process. The
second participant then tried to answer these questions while reading the text.
The participants switched roles for the second half of the text. All participants
answered a final set of test questions about the passage. A control group, which
just read the text without doing anything special, correctly answered 50% of the
final set of test questions. Experimental participants who made up questions
while they read the text correctly answered 70% of the final test items that were
relevant to their questions and 52% of the items that were irrelevant. Experimental
participants who answered questions while they read the text correctly
answered 67% of the relevant test items and 49% of the irrelevant items. Thus, it
seems that both question generation and question answering contribute to good
memory. If anything, the creation of questions contributes the most benefit.
Reviewing the text with the questions in mind is another important component
of the PQ4R technique. Rothkopf (1966) compared the benefit of reading a
text with questions in mind with the benefit of considering a set of questions after
reading the text. Rothkopf instructed participants to read a long text passage with
168 | Human Memory: Encoding and Storage
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 168
questions interspersed every three pages. The questions were relevant to the three
pages either following or preceding the questions. In the former condition, participants
were supposed to read the subsequent text with these questions in mind.
In the latter condition, they were to review what they had just read and answer
the questions. The two experimental groups were compared with a control group,
which read the text without any special questions. The control group answered
30% of the questions correctly in a final test of the whole text. The experimental
group whose questions previewed the text correctly answered 72% of the test
items relevant to their questions and 32% of the irrelevant items—basically the
same results as those Frase (1975) obtained in comparing the effectiveness of relevant
and irrelevant test items. The experimental group whose questions reviewed
the text correctly answered 72% of the relevant items and 42% of the irrelevant
items. Thus, it seems that using questions to review the text just read may be more
beneficial than reading the textwith the questions in mind.
Study techniques that involve generating and answering questions lead
to better memory for text material.
Meaningful versus Nonmeaningful Elaborations
Although the research just reviewed indicates that meaningful processing leads
to better memory, other research demonstrates that other sorts of elaborative
processing also result in better memory. For instance, Kolers (1979) looked at
participants’ memory for sentences that were read in normal form versus sentences
that were printed upside down and found that participants remembered
more about the upside-down sentences. He argued that the extra processing
involved in reading the typography of upside-down sentences provides the
basis for the improved memory. It is not a case of more meaningful processing,
but one of more extensive processing.
A study by Slamecka and Graf (1978) demonstrated separate
effects on memory of elaborative and meaningful processing.
They contrasted two conditions. In the generate condition, participants
were given a word and had to generate either a synonym
that began with a particular letter (e.g.,What is a synonym of sea
that begins with the letter o? Answer: ocean) or a rhyme that began
with a particular letter (e.g.,What is a rhyme of save that begins
with the letter c? Answer: cave). In the read condition, they just
read the rhyme pair or the synonym pair and then were tested for
their recognition of the second word. Figure 6.15 shows the
results. Participants performed better with synonyms and better
when they had to generate either a synonym or a rhyme. Thus, it
appears that there are beneficial effects on memory of both
semantic processing and elaborative processing.
Otten, Henson, and Rugg (2001) noted that the prefrontal
and hippocampal regions involved in memory for material that
is processed meaningfully and elaborately seem to be the same
regions that are involved in memory for material that is
processed shallowly. High activity in these regions is predictive of
Techniques for Studying Textual Material | 169
0.0
Proportion recognized
Synonym Rhyme
0.2
0.4
0.6
0.8
1.0
Generate
Read
FIGURE 6.15 Results of a
study by Slamecka and Graf
demonstrating separate effects
on memory of elaborative and
meaningful processing. In the
generate condition, participants
were presented with a word
and had to generate either a
synonym or a rhyme that began
with a particular letter. In the
read condition, participants
read the synonym pair or rhyme
pair, then were tested for their
recognition of the second
word. The proportion of words
recognized is shown as a
function of the type of
elaboration (synonym or rhyme)
and whether it was generated
or read. (From Slamecka & Graf, 1978.
Reprinted by permission of the publisher.
© 1978 by the Journal of Experimental
Psychology: Human Learning and Memory.)
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 169
subsequent recall (see Figure 6.13). Elaborative, more meaningful processing
tends to evoke higher levels of activation than the shallow processing (Wagner et
al., 1998). Thus, it appears that meaningful, elaborate processing is effective
because it is better at driving the brain processes that result in successful recall.
170 | Human Memory: Encoding and Storage
bananas, and bread. To associate the milk with the
bookstore, we might imagine books lying in a puddle of
milk in front of the bookstore. To associate hot dogs
with the record shop (the next location on the path
from the bookstore), we might imagine
a package of hot dogs spinning on a
phonograph turntable. The pizza shop is
next, and to associate it with dog food,
we might imagine a dog-food pizza (well,
some people even like anchovies). Then
we come to an intersection; to associate
it with tomatoes, we can imagine an
overturned vegetable truck with tomatoes
splattered everywhere. Next we
come to the administration building and
create an image of the president coming
out, wearing only a hula skirt made of
bananas. Finally, we reach the library and
associate it with bread by imagining a
huge loaf of bread serving as a canopy
under which we must pass to enter. To re-create the
list, we need only take an imaginary walk down this
path, reviving the association for each location. This
technique works well even with very much longer lists;
all we need is more locations. There is considerable
evidence (e.g., Christen & Bjork, 1976) that the same
loci can be used over and over again in the learning of
different lists.
Two important principles underlie this method’s effectiveness.
First, the technique imposes organization on an
otherwise unorganized list. We are guaranteed that if we
follow the mental path at the time of recall, we will pass
all the locations for which we created associations. The
second principle is that imagining connections between
the locations and the items forces us to process the
material meaningfully, elaboratively, and by use of visual
imagery.
Implications
How does the method of loci help us organize recall?
Mental imagery is an effective method for developing
meaningful elaborations. A classic mnemonic technique,
the method of loci, depends heavily on visual imagery
and the use of spatial knowledge to organize recall. This
technique, used extensively in ancient
times when speeches were given without
written notes or teleprompters, is still
used today. Cicero (in De Oratore) credits
the method to a Greek poet,
Simonides, who had recited a lyric poem
at a banquet. After his delivery, he was
called from the banquet hall by the gods
Castor and Pollux, whom he had praised
in his poem. While he was absent, the
roof fell in, killing all the people at the
banquet. The corpses were so mangled
that relatives could not identify them.
Simonides was able to identify each
corpse, however, according to where each
person had been sitting in the banquet
hall. This feat of total recall convinced Simonides of the
usefulness of an orderly arrangement of locations into
which a person could place objects to be remembered.
This story may be rather fanciful, but whatever its true
origin, the method of loci is well documented (e.g., Christen
& Bjork, 1976; Ross & Lawrence, 1968) as a useful
technique for remembering an ordered sequence of
items, such as the points a person wants to make in a
speech.
To use the method of loci, one imagines a specific
path through a familiar area with some fixed locations
along the path. For instance, if there were such a path
on campus from the bookstore to the library, we might
use it. To remember a series of objects, we simply walk
along the path mentally, associating the objects with
the fixed locations. As an example, consider a grocery
list of six items—milk, hot dogs, dog food, tomatoes,
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 170
More elaborate processing results in better memory, even if that processing
is not focused on the meaning of the material.
Incidental versus Intentional Learning
So far, we have talked about factors that affect memory. Now we will turn to a
factor that does not affect memory, despite people’s intuitions to the contrary:
It does not seem to matter whether people intend to learn the material; what is
important is how they process it. This fact is illustrated in an experiment by
Hyde and Jenkins (1973). Participants saw groups of 24 words presented at the
rate of 3 s per word. One group of participants was asked to check whether each
word had a letter e or a letter g. The other group was asked to rate the pleasantness
of the words. These two tasks were called orienting tasks. It is reasonable to
assume that the pleasantness rating involved more meaningful and deeper processing
than the letter-verification task. Another variable was whether participants
were told that the true purpose of the experiment was to learn the words.
Half the participants in each group were told the true purpose of the experiment
(the intentional-learning condition). The other half of participants in
each group thought the true purpose of the experiment was to rate the words
or check for letters (the incidental-learning condition). Thus, there were four
conditions: pleasantness-intentional, pleasantness-incidental, letter checkingintentional,
and letter checking-incidental.
After studying the list, all participants were asked to recall as many words as
they could. Table 6.2 presents the results from this experiment in terms of percentage
of the 24 words recalled. Two results are noteworthy. First, participants’
knowledge of the true purpose of studying the words had relatively little effect
on performance. Second, a large depth-of-processing effect was demonstrated;
participants showed much better recall in the pleasantness rating condition,
independent of whether they expected to be tested on the material later. In
rating a word for pleasantness, participants had to think about its meaning,
which gave them an opportunity to elaborate upon the word.
The Hyde and Jenkins (1973) experiment illustrates
an important finding that has been proved over
and over again in the research on intentional versus
incidental learning: Whether a person intends to learn
or not really does not matter (see Postman, 1964, for a
review).What matters is how the person processes the
material during its presentation. If one engages in
identical mental activities when processing the material,
one gets identical memory performance whether
one is intending to learn the material or not. People
typically show better memory when they intend to
learn because they are likely to engage in activities
more conducive to good memory, such as rehearsal
and elaborative processing. The small advantage for
Techniques for Studying Textual Material | 171
TABLE 6.2
Words Recalled as a Function of Orienting Task
and Participant Awareness of Learning Task
Words Recalled (%)
Orienting Task
Learning-Purpose Rate
Conditions Pleasantness Check Letters
Incidental 68 39
Intentional 69 43
After Hyde and Jenkins, 1973. Adapted by permission of the publisher.
© 1973 by the Journal of Verbal Learning and Verbal Behavior.
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 171
participants in the intentional learning condition of the Jenkins and Hyde experiment
may reflect some small variation in processing. Experiments in which great
care is taken to control processing find that intention to learn or amount of motivation
to learn has no effect (see T.O.Nelson, 1976).
There is an interesting everyday example of the relationship between intention
to learn and type of processing. Many students claim they find it easier to
remember material from a novel, which they are not trying to remember, than
from a textbook, which they are trying to remember. The reason is that students
find a typical novel much easier to elaborate, and a good novel invites
such elaborations (e.g.,Why did the suspect deny knowing the victim?).
Level of processing, and not whether one intends to learn, determines the
amount of material remembered.
Flashbulb Memories
Although it does not appear that intention to learn affects memory, various sets
of data support the conclusion that people display better memory for events that
are important to them. One class of research involves flashbulb memories—
events so important that they seem to burn themselves into memory forever
(R. Brown & Kulik, 1977). The event these researchers used as an example was
the assassination of President Kennedy in 1963, which was a particularly traumatic
event for Americans of their generation. They found that most people had
vivid memories of the event 13 years later. They proposed that we have a special
biological mechanism to guarantee that we will remember those things that are
particularly important to us. The interpretation of this result is problematic,
however, because R. Brown and Kulik did not really have any way to assess the
accuracy of the reported memories.
Since the Brown and Kulik proposal, a number of studies have been done to
determine what participants remembered about a traumatic event immediately
after it occurred and what they remembered later. For instance, McCloskey,
Wible, and Cohen (1988) did a study involving the 1986 space shuttle Challenger
explosion. At that time, many people felt that this was a particularly traumatic
event they had watched with horror on television. McCloskey et al. interviewed
participants 1 week after the incident and then again 9 months later.Nine months
after the accident, one participant reported:
When I first heard about the explosion I was sitting in my freshman dorm
room with my roommate and we were watching TV. It came on a news flash
and we were both totally shocked. I was really upset and I went upstairs to talk
to a friend of mine and then I called my parents. (Neisser & Harsch, 1992, p. 9)
McCloskey et al. found that although participants reported vivid memories
9 months after the event, their reports were in fact inaccurate. For instance, the
participant just quoted had actually learned about the Challenger explosion in
class a day after it happened and then watched it on television.
Palmer, Schreiber, and Fox (1991) came to a somewhat different conclusion
in a study of memories of the 1989 San Francisco earthquake. They compared
172 | Human Memory: Encoding and Storage
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 172
participants who had actually experienced the earthquake firsthand with those
who had only watched it on TV. Those who had experienced it in person
showed much superior long-term memory of the event. Conway et al. (1994)
argued that McCloskey et al. (1988) failed to find a memory advantage in the
Challenger study because their participants did not have true flashbulb memories.
They contended that it is critical for the event to have been consequential
to the individual remembering it. Hence, only people who actually experienced
the San Francisco earthquake, and not those who saw it on TV, had flashbulb
memories of the event. Conway et al. studied memory for Margaret Thatcher’s
resignation as prime minister of the United Kingdom in 1990. They compared
participants from the United Kingdom, the United States, and Denmark, all
of whom had followed news reports of the resignation. It turned out that
11 months later, 60% of the participants from the United Kingdom showed
perfect memory for the events surrounding the resignation, whereas only 20%
of those who did not live in the United Kingdom showed perfect memory.
Conway et al. argued that they obtained this result because the Thatcher resignation
was really consequential only for the U.K. participants.
On September 11, 2001, Americans suffered a particularly traumatic event.
The terrorist attacks of that day have come to be known simply as “9/11.”
A number of studies were undertaken to study the effects of these events on
memory. Talairco and Rubin (2003) report a study of the memories of students
at Duke University for details of the terrorist attacks (flashbulb memories) versus
details of ordinary events that happened that day. They were contacted and
tested for their memories the morning after the attacks. They were then tested
again either 1 week later, 6 weeks later, or 42 weeks later. Figure 6.16 shows the
results. The figure shows both recall of details that are consistent with what
they said the morning after and recall of details that were inconsistent (presumably
false memories). By neither measure is there any evidence that the flashbulb
memories were better retained than the everyday memories. Of course,
these were students at Duke and not people who were experiencing the terrorist
attacks firsthand.
Sharot, Martorella, Delgado, and Phelps (2007) reported
a study of people who were in Manhattan, where the Twin
Towers were struck. The study was performed three years
after the attack and people were asked to recall the events
from the attack and events from the summer before. Because
the study was 3 years after the event and they could not
verify participants’ memories for accuracy, but they could
study their brain responses while they were recalling the
events, Sharot et al. also interviewed the participants to find
out where they were in Manhattan when the twin towers
were struck. They broke the participants into two groups—
a downtown group who were approximately 2 miles away
and a midtown group who were approximately 5 miles away.
They focused on activity in the amygdala, which is a brain
structure known to reflect emotional response. They found
greater amgydala activation in the downtown group when
Techniques for Studying Textual Material | 173
Number of details
12
10
1 7 42
Inconsistent
Consistent
Flashbulb
Everyday
224
Days since 9/11 (log scale)
FIGURE 6.16 The mean number
of consistent and inconsistent
details for the flashbulb and
everyday memories (from Talairco &
Rubin, 2003).
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 173
they were recalling events from September 11 than in the midtown group. This is
significant because there is evidence that amygdala activity enhances retention
(Phelps, 2004). In a state of arousal, the amygdala releases hormones that influence
the processing in the hippocampus that is critical in forming memories
(McGaugh & Roozendaal, 2002).
Behaviorally there is considerable evidence that memories learned in high
arousal states are better retained (J. R. Anderson, 2000). This evidence can be
interpreted in two alternative ways. The first is that we have some special biological
mechanism that reinforces memory for events that are important to us.
The second is that people simply rehearse the material that is more important
to them more often (as argued by Neisser et al., 1996). Both factors probably
play a role in better retention of information learned in a high arousal state.
People certainly do tend to rehearse important memories, but evidence about
the amygdala involvement in memory suggests that there is an effect of arousal
over and above rehearsal.
People report better memories for particularly traumatic events, and there is
evidence that memories in high arousal states are better retained.
•Conclusions
This chapter has focused on the processes involved in getting information into
memory. We saw that a great deal of information gets registered in sensory
memory, but relatively little can be maintained in working memory and even
less survives for long periods of time. However, an analysis of what actually gets
stored in long-term memory really needs to consider how that information is
retained and retrieved—which is the topic of the next chapter. For instance, the
effects of arousal on memory that we have just reviewed seem to be more about
retention than encoding. Likewise, many of the issues considered in this chapter
are complicated by retrieval issues. This is certainly true for the effects of elaborative
processing that we have just discussed. There are important interactions
between how a memory is processed at study and how it is processed at test.
Even in this chapter, we were not able to discuss the effects of such factors as
practice without discussing the activation-based retrieval processes that are
facilitated by these factors. Chapter 7 will also have more to say about the activation
of memory traces.
174 | Human Memory: Encoding and Storage
1. Many people write notes on their bodies to remember
things like phone numbers. In the movie Memento,
Leonard tattoos information that he wants to remember
on his body. Describe instances where storing information
on the body works like sensory memory,where it is like
working memory, and where it is like long-termmemory.
2. The chapter mentions a colleague of mine who was
stuck solving the riddle “What goes up a chimney down
but can’t come down a chimney up?”How would you
have seeded the environment to subconsciously prime
a solution to the riddle? To see what Kaplan did, read
J. R. Anderson (2007, pp. 93–94).
Questions for Thought
Anderson7e_Chapter_06.qxd 8/20/09 9:45 AM Page 174
Key Terms | 175
3. Figures 6.11 and 6.12 describe situations in which the
experiment arranges for a participant to study memories
many times to improve its strength. Can you describe
situations in your schooling where this sort of
practice happened to improve your memory for facts?
4. Think of the most traumatic events you have experienced.
How have you rehearsed and elaborated upon
these events? What influence might such rehearsal and
elaboration have on these memories? Could it cause
you to remember things that did not happen?
Key Terms
activation
ACT (Adaptive Control
of Thought)
anterograde amnesia
articulatory loop
associative spreading
auditory sensory store
central executive
depth of processing
echoic memory
elaborative processing
flashbulb memories
iconic memory
long-term potentiation
(LTP)
memory span
method of loci
partial-report
procedure
phonological loop
power function
power law of learning
SAM (Search of Associative
Memory)
short-term memory
spreading activation
strength
visual sensory store
visuospatial sketchpad
whole-report
procedure
working memory