3 Working Memory

Learning Objectives
1. Define working memory, short-term memory, primary memory, and
sensory memory.
2. Assess encoding, duration, and forgetting in working memory.
3. Appraise Baddeley’s working memory model.
4. Describe the applications of working memory theory to everyday life.

Imagine you are driving your car to a friend’s house. You’ve never
been there before and you forgot to program your phone with the
directions, so you pull off to the side of the road to speak on your cell
phone with your friend. She is giving you directions: “Make a left onto
Martin Luther King Boulevard, go straight for two miles, and then,
when you are just past the university, make a right onto Canseco
Street.” Your friend gets another call, signs off, and you are on your
own. Will you remember those directions? You are not sure, so you
decide to repeat the directions over and over until you get to
Canseco Street. What many people do is mentally rehearse those
directions to keep them fresh—available, that is, in their working
memory until they do not need them anymore. “Left on MLK, two
miles, right on Canseco,” might keep running through your head. The
goal of rehearsing is to keep information in working memory, the
active contents of our consciousness, so we can make use of that
information immediately. The concept of working memory is
important for practical reasons: you need to remember directions,
the phone number to the pizza shop, and where you just put your car
keys.

Working memory: The neural structures and cognitive processes that

maintain the accessibility of information for short periods of time in an
active conscious state.
Working memory systems are the neurocognitive systems that allow
us to maintain information over short periods of time. Working
memory used to be called short-term memory, but for a number of
reasons, that term has fallen out of favor. Short-term memory is
more often used to describe a stage in an information-processing
model rather than a functional neurocognitive system (Atkinson &
Shiffrin, 1968). Given current theory, working memory is a more
appropriate term (Baddeley, 2012; Shipstead & Nespodzany, 2019).
However, working memory is also one of the most philosophically
loaded terms in modern experimental psychology because of its
relation to conscious processing. Most cognitive psychologists today
think of working memory as the active contents of consciousness
(Jacobs & Silvanto, 2015). Working memory indeed can be
considered consciousness itself. Whatever you are conscious of
right now (hopefully this paragraph) is exactly what your working
memory is representing right now. Direct your attention to the
television running in the background or your roommate’s game of
Call of Duty, and this paragraph will cease to be maintained in
working memory as your attention and conscious awareness are
directed elsewhere.

What Is Working Memory?

Although there is still controversy about the theories used to
describe what working memory is and explain how it works, almost
all cognitive scientists would agree on certain basic tenets
(Barrouillet & Camos, 2015). Working memory is a short-term
memory system. Working memory’s function is to temporarily hold
information over a short period of time. Estimates of this period may
vary, but most run somewhere between 15 and 30 seconds (1/2
minute). If information is continually refreshed or rehearsed, it can be
maintained indefinitely in working memory. However, as soon as
rehearsal stops, information will be lost from working memory within
that timeframe. Interestingly, the process by which information is lost
from working memory has to do with interference, not time per se. In
contrast, long-term memory can store information for minutes, years,
and even an entire lifetime. Some researchers consider short-term
memory storage—that is, the representation of information within
working memory—an important part of working memory (Shipstead,
Lindsey, Marshall, & Engle, 2014; Shipstead & Nespodzany, 2019).
Working memory is a limited capacity system. It can only hold so
much information. G. A. Miller (1956) identified short-term memory
(now working memory) as maintaining about seven units of
information. Subsequent research has modified this conception of
working memory; nonetheless, only a small, finite amount of
information can be active at any particular point in time. In essence,
working memory capacity is the amount of information that can be
spoken in 1.5 seconds. This contrasts with long-term memory, which
appears to have a virtually limitless capacity. Research has never
been able to document the maximum amount that human long-term
memory can hold.

Working memory is readily contrasted with the typical

conceptualization of long-term memory. Working memory maintains
information for brief periods of time in an active conscious state. In
contrast, long-term memory stores information for long periods
before it is activated when called for. Working memory can maintain
only a limited number of items in conscious awareness at any point
in time. Long-term memory has seemingly limitless capacity. Indeed,
most research suggests that the more someone knows, the easier it
is for that person to learn more. However, the contents of long-term
memory are not conscious until they are activated into working
memory. Despite some popular misconceptions to the contrary, it is
impossible to “fill up” one’s long-term memory, as if it were a gas
tank. Finally, what is in our working memory is what we are thinking
about now. Thus, working memory is often thought of as being
equivalent to conscious awareness. In long-term memory, there may
be information we have not thought of for years and that may be very
difficult to retrieve into an active form.

Some Terminological Clarifications

Prior to the late 1980s (Baddeley, 1986), the term short-term
memory was used more often to describe the phenomena covered
in this chapter. Cognitive psychologists now seldom use the term
short-term memory. First, it is associated with theory that is no longer
considered to be correct. Second, the term short-term memory is
now used in everyday speech in a way that is different from its
former use in psychology. As a consequence, nowadays, memory
science prefers the term working memory. You may see the term
primary memory as well. Primary memory was the term originally
used by William James to refer to working memory and has been
used by some researchers since then (James, 1890; see Baddeley,
2012). This term also fell out of favor because current conceptions of
working memory postulate that it is both the active area where we
rehearse new information and the area that holds information after it
has been retrieved from long-term memory. Thus, it is neither
primary nor secondary to long-term memory.

Short-term memory: An older term used to describe the memory

system that holds information for a short period of time, up to 15
seconds.

Primary memory: A term used to mean short-term memory.

Terminology is often tricky, as other subfields of psychology may use

the same term to mean a different thing. Such is the case with the
term working memory. Many animal behavior researchers are
interested in animal memory systems and how they compare to each
other and to human memory. Animal behavior researchers use the
term working memory, but they use the term in a different manner. In
animal memory research, the term working memory refers to
memory of the most recent trial, regardless of the time course of that
memory (see Shettleworth, 2010). This is quite different from its
usage in human memory research. In keeping with current
terminology, this textbook will use the term working memory in the
manner in which it is used in the study of human memory.
Sensory Memory
Although the focus of this chapter will be on working memory, it is
also important to describe a memory system that functions at even
shorter time spans than does working memory—that is, sensory
memory. We include sensory memory here, but make sure to
understand that sensory memory is thought to be a different system
than working memory, though they are sometimes confused.
Sensory memory refers to a very brief memory system that holds
an exact representation of what you have seen for a fraction of a
second to allow cognitive processing to occur. Unlike working
memory, sensory memory occurs prior to conscious access. Sensory
memory is thought to be composed of separate memory systems for
each perceptual system. Iconic memory is visual sensory memory,
whereas echoic memory is auditory sensory memory. Sensory
memory is a system that operates prior to consciousness and thus
before working memory begins.

Sensory memory: A very brief memory system that holds literal

information for a fraction of a second to allow cognitive processing.

Iconic memory: Visual sensory memory.

Echoic memory: Auditory sensory memory.

Sensory memory is sometimes considered a buffer system, because

it holds information from our senses for a brief period of time so that
we can extract more information from what we are seeing or hearing.
Our eyes and ears are constantly being bombarded with information,
only some of which is relevant at any particular time. In theory,
sensory memory creates a buffer that allows us to maintain sensory
information—in a sensory format—long enough for cognitive
mechanisms to detect and attend to relevant information.

In a now-classic experiment, the cognitive psychologist George

Sperling (1960) demonstrated the hypothetical existence of iconic
memory or visual sensory memory. Participants were shown a matrix
of 12 letters in a 4 × 3 grid for a brief period (50 msec). Participants
were then asked to report letters in one of two ways: whole report
and partial report. In the whole-report technique, participants were
asked to retrieve all of the letters from the matrix. Under the whole-
report technique, participants could only recall about five letters. In
the partial-report technique, they heard a tone that indicated which
line to report. A high tone indicated the top line, a medium tone
indicated the middle line, and a low tone indicated the bottom line.
The tones occurred just after the stimulus grid disappeared.
Participants in the partial-report condition could remember three from
the specified line, suggesting that nine letters were accessible
visually at the time of recall rather than the five one might have
estimated from the whole-report technique (see Figure 3.1). In this
study, the independent variable was the kind of report—partial or
whole. The dependent measure was the amount recalled from the
matrix.

Figure 3.1 ■ Participants in Sperling’s 1960 experiment would see

the following grid for a very brief flash. In the partial report, they were
required to report just one line. In the whole report, they were asked
to recall the entire matrix.
Source: Sperling (1960).

Sperling (1960) demonstrated the existence and the function of

sensory memory by comparing performance in the whole-report
condition and the partial-report condition. In each condition,
participants saw the 4 × 3 grid for the same amount of time, and in
each condition, the interval between the offset of the stimulus and
the repeating of the letters was the same. However, in the partial-
report condition, participants received information about which line to
report immediately after they could no longer see the stimulus. Thus,
they needed to be storing a visual record of the grid mentally in order
to be able to report 75% of the line that the tone told them to. The
whole-report condition showed lower per-line memory, because of
working memory problems. As these participants reported letters
from the grid, the representation faded from sensory memory, and
because of the short presentation time, few letters had entered
working memory. Thus, sensory memory is a low-level system
separate from working memory. Despite the success of this
experiment and others, sensory memory has not been a major
concern of memory researchers for some time (see G. R. Loftus,
1983).

Working Memory Capacity

Here’s the phone number for the best pizza in the city: 555-3756.
Most of us do not have a problem juggling that seven-digit number in
our head. We repeat the numbers over and over until we dial the
number and get our “everything but anchovies” pizza (see Figure
3.2). Now consider when you also have to remember an unfamiliar
area code. Now try keeping this number in working memory, 324-
555-3756. Try mentally rehearsing that number. It is likely that you
will have forgotten bits of it by the time you actually get your phone
out of your pocket. The task of remembering the numbers becomes
much more difficult for most of us as the number of digits passes
seven.
Figure 3.2 ■ You will need to remember the digits of the phone
number before you can have this mouth-watering pizza delivered.
© Dick Luria/Photodisc/Thinkstock

George Miller (1956) claimed that the capacity of human working

memory (or primary memory, to use Miller’s term) is, on average,
seven items of information. Capacity is the amount of information
that can be maintained in working memory. Miller described the
“magic number 7” as critical to working memory. It was his view that
working memory could hold 7 ± 2 “items” at any point in time (we
shall return to the concept of items shortly). This makes the seven-
digit number relatively easy but the slightly longer 10-digit number
extremely difficult. Likewise, for most people, remembering a nine-
digit Social Security number on the first try is very difficult. The “plus
or minus two” refers to individual differences in the capacity of
working memory. This capacity limitation is why we need to keep
glancing back at our airplane confirmation number on our ticket
when we enter it into the boarding pass machine at the airport. This
also works for other stimuli—it is easy to remember a bar or two of a
great song you hear on the radio for the first time, but it is very
difficult to sing back the entire song after listening to it once.

One method originally employed to look at the capacity of short-term

memory, but since adapted to working memory, is called the digit
span task, a standard test in the memory researcher’s toolkit. The
digit span task is an easy but very valuable research tool. This task
is similar to trying to remember the digits in the phone number for the
pizza restaurant. An experimenter reads a list of numbers to a willing
participant or group of participants. As soon as the list is read, the
participants must repeat back the numerals in order, either by
speaking the words aloud or writing them down. The experimenter
can then vary the number of digits that she reads to the participants.
The ability to recall the digits can be examined as a function of how
long the set was. See the demonstration below and try it yourself.

Capacity: The amount of information that can be maintained in working

memory.

Digit span task: A task in which a person must remember a list of digits
presented by an experimenter.

Demonstration: Try your hand at the following digit sequences. Read

them. Then close the book and try to repeat them.

5-digit: 9 0 3 6 7
6-digit: 4 7 6 2 9 1
7-digit: 3 7 8 5 2 6 0
8-digit: 1 9 8 7 2 0 4 6
10-digit: 1 2 0 7 3 9 6 8 2 9
12-digit: 4 1 3 5 6 7 5 8 1 8 4 6

Results in digit span experiments show that average (arithmetic

mean) performance is just about seven digits, consistent with Miller’s
“magic” number. Given that educated and younger people tend to
have longer digit spans than less educated and older people, most
college students doing this demonstration will find seven digits
relatively easy and succeed at the eight-digit sequence (G. Jones &
Macken, 2015). The 10-digit sequence, however, will likely be
outside their capacity. In such a digit span task, the independent
variable is the number of digits presented in a sequence. The
dependent measure is whether or not these sequences are recalled.

Let’s return to the concept of an “item.” G. A. Miller (1956) suggested

that the capacity of working memory was about seven items. For
Miller, however, the word item was intentionally vague. Indeed, he
argued that a lot of information could be packed into a single item.
People can use strategies to make each item of information
decomposable into several parts. For example, rather than
remembering the number sequence 6-1-9 as three digits, each a unit
of information, participants could simply rehearse “San Diego.” Then,
when it is time to retrieve the digits, the item “San Diego” could be
converted into 6-1-9 because 619 is the area code for that city.
Others might encounter the sequence 8-6-7-5-3-0-9 and encode it as
one item (“Jenny,” from the 1982 pop song; thanks to a reviewer for
suggesting this example). Using this strategy is called chunking.
Miller introduced the word chunk to represent the basic unit of
information in working memory. Each chunk in memory represents a
unit of related components (Baddeley, 2012; Cowan, 2001). Indeed,
researchers now may be more inclined to use letters instead of
numbers to measure capacity, as this may reduce chunking.

Chunk: Basic unit of information in working memory. A chunk may be

decomposable into more information.

Individuals can use many strategies to chunk digits in a digit span

task. You can use area codes, home addresses, and even the jersey
numbers of your favorite athletes. For example, another sequence,
2-3, can be encoded as “Michael Jordan” rather than as two digits,
as the basketball player was famous for that jersey number. In this
way, we can extend our working memory capacity by using more and
more informative chunking strategies. For example, your author can
remember up to 14 digits by taking each two-digit sequence and
processing it as a jersey number for a famous athlete.
In a classic demonstration of how effective chunking can be, Anders
Ericsson and his colleagues trained a first-year college student in
digit span tasks (Ericsson, Chase, & Faloon, 1980). At the start, the
first-year college student had a normal digit span of about seven
numbers. The experimenters did not suggest any methods to him.
They just gave him lots of practice. However, by the end of a year of
training, the young man had an 80-number digit span! That is, an
experimenter could read off a list of 80 numbers, and the young man
could repeat them all right back without any mistakes. How could this
man with a normal digit span increase his abilities so dramatically
over the course of an academic year?

First, it was his job—he spent about 10 hours a week for 40 weeks
engaging in digit span tasks. That means a lot of practice and hard
work. But it was not simply practice—the student used a complex
mnemonic chunking strategy to build his digit span. The young man
was a star runner on the university track team. He therefore chunked
numbers by thinking of them in terms of times in track-and-field
races. The sequence 1-9-1-9 might be encoded as “world record in
200-meter dash.” The sequence 3-4-3 might be remembered as near
world record pace for the mile. Because most of these sequences
were already stored in his long-term memory, he was able to apply
them to the new digit sequences (Ericsson et al., 1980). The
complex chunking strategy combined with lots of practice made a
mnemonist out of this young runner. Just to be sure there was
nothing unique about this young man, Ericsson and colleagues
recruited another runner from the track team the next year and gave
him the same amount of practice using the same chunking strategy.
This participant went from a normal seven-digit span to a 70-digit
span over the course of the experiment. The implication is that
anyone can use chunking strategies to boost memory performance
(see Ericsson, 2003).

Mnemonic Improvement Tip 3.1

If you need to keep arbitrary lists of information accessible in working

memory, use chunking strategies as best you can. Try to encode
 multiple items of information using a common associative strategy. Use
well-learned information to guide your chunking strategy. Chunking is
useful in ordinary learning tasks that involve working memory, such as
remembering phone numbers, keeping directions in mind, and
remembering the names of new acquaintances.

Memory Span Tasks

With evolving models of working memory, it became clear that digit
spans were too simple to address many of the issues in working
memory. In particular, they could not address attentional issues (or
executive-function issues). As such, more complex tasks were
developed that could allow researchers to look at attentional issues
as well as storage issues. These tasks are called memory span
tasks or running-memory span tasks (Shipstead & Nespodzany,
2019). In a memory span task, participants must observe a series
of items presented to them one after the other. For example, they
may see or hear a series of letters presented one by one. The list
may be of any length, and the participant may not know how long the
list will be. When the list finally ends, the participant must be able to
then report on the last 7 digits (sometimes fewer). Because the
participant does not know when the list is going to end, he or she
must continually track and update the items on the list. Thus, unlike
the normal digit span task, the memory span task requires
attentional resources to be devoted to continually updating the items
(Harrison et al., 2013).

Memory span task: Participants must track a series of letters, digits,

words, or other items as they are continuously presented. At the end of
the list, participants must report on the last few items.

Pronunciation Time
Regardless of one’s ability to chunk information, most of us are still
constrained by the fact that working memory usually can contain only
about seven items, even if these items can be chunked. Therefore,
the extraordinary digit spans of these participants do not take away
from Miller’s basic finding—that seven items is the average amount
of information in working memory. However, factors other than the
number of items in a chunk are important in determining the capacity
of working memory. One such factor is phonological processing, also
called pronunciation time. Pronunciation time refers to the amount
of time it would take to say aloud the items being rehearsed in
working memory. The limit on working memory is the number of
words that can be pronounced, either aloud or subvocally, in about
1.5 seconds (Schweickert & Boruff, 1986). Most of us can say seven
digits in American English in 1.5 seconds; hence, the digit span of
seven.

Pronunciation time: The amount of time it would take to say aloud the
items being rehearsed in working memory.

Because longer words require more of working memory than do

shorter words, the capacity of working memory is influenced by word
length. For example, you can maintain the names of European cities
with one syllable (Prague, Nice, Rome, Bruge) more easily in
working memory than the names of European cities with multiple
syllables (Amsterdam, Bratislava, Barcelona, Manchester). For
another example, consider the names of trees. Try to keep the
following words rehearsed in working memory: oak, birch, pine,
palm. Then try to keep the following words rehearsed in working
memory: eucalyptus, bottlebrush, Poinciana, sycamore. Most of us
will find it easier to maintain the trees with short names in working
memory than the longer ones. This phenomenon is referred to as the
word length effect.

Word length effect: Longer words are more difficult to maintain in

working memory than are shorter words.

To test the effect of word length, Ellis and Hennelly (1980) looked at
digit spans in children in Great Britain. They focused their testing in
the part of Great Britain called Wales. In Wales, there are still people
whose first language is Welsh (a Celtic language) rather than
English, although practically all Welsh speak English. In Welsh, the
words for digits take longer to pronounce than they do in English.
This is largely because Welsh vowels take longer to pronounce than
their English counterparts. Therefore, it takes more time to count
from one to 10 in Welsh than it does in English. Ellis and Hennelly
examined digit spans in Welsh–English bilingual children.
Consistently, the children’s digit span was longer in English than it
was in Welsh. Because the words mean the same thing in both
languages and because most of the targeted students were more
fluent in Welsh, the digit spans must be different because, at least
for the names of the digits, one can say the digits faster in English
than in Welsh. In the word-length effect, the independent variable is
the length of the words. The dependent measure is whether or not
the words are recalled.

A short time later, a second study confirmed this finding using

different languages. Naveh-Benjamin and Ayres (1986) also made
use of the fact that digits take a different amount of time to say in
different languages. For example, in English, the numeral 1 is
pronounced one, which is one syllable. In Spanish, however, the
numeral 1 is pronounced uno, which is two syllables and takes a little
more time to say than one does in English. Therefore, uno should
occupy slightly more of the capacity of working memory than one
does. Naveh-Benjamin and Ayres tested students at a university in
Israel, where they were able to examine fluent speakers of four
languages: English, Spanish, Hebrew, and Arabic. In terms of
pronunciation time, it takes the least amount of time to count from 1
to 10 in English and the most amount of time to count from 1 to 10 in
Arabic. When the digit span task was presented to speakers of each
language, the digit spans reflected the pronunciation times. The digit
spans were longest in English, followed by, in order, Spanish,
Hebrew, and Arabic (see Figure 3.3). This decrease in digit spans is
related to the increase in pronunciation time of the digits in each
language. Therefore, these results support the idea that
pronunciation time is relevant in evaluating the capacity of working
memory. And for your own practice, identify the independent and
dependent variables in the Naveh-Benjamin and Ayres study.

Description

Figure 3.3 ■ Digit span as a function of language. y-axis represents

the number of words repeated.
Source: Based on Naveh-Benjamin and Ayres (1986).

The Duration of Information in

Working Memory
Working memory holds information for only brief periods of time.
Because of this, a frustrating memory experience is the rapid loss of
information from working memory. Consider having just heard the
phone number of your client on your answering machine. Before you
can write the entire number down, you have lost the first three digits.
For professors, working memory often constrains the efficiency with
which we learn the names of new students. For example, I try to
learn the names of many new students on the first few days of class
each semester. I am well aware how insulting it can be when
somebody forgets your name immediately. A student may introduce
herself as “Christa Rodriguez, a sophomore, from South Miami,
Florida. Majoring in psychology.” I look at their face and silently
repeat the name to myself. Then the next student introduces himself
or herself. After a few more students, I look back at the first person,
and I have no recollection of the name they just spoke mere minutes
before. This is because the information pertaining to Christa
Rodriguez is no longer in working memory, having been replaced by
subsequent names. Because I failed to transfer their name to long-
term memory, the subsequent names, faces, and other information
have displaced the information in my working memory, leaving me
with a complete blank on the student’s name. Because the
information was never encoded into long-term memory, it can no
longer be recalled.

This leads us to one of the earliest issues tackled by cognitive

psychologists: How long does information persist in working
memory? And it also leads us to the closely related question: What
causes forgetting from working memory?

To address the first question, the short answer is that it depends.

And what it depends on is rehearsal, an important concept in
memory science. Rehearsal here means actively maintaining the
item in working memory by repeating it over and over (maintenance
rehearsal) or by relating the item to some other concept
(elaborative rehearsal). If you wish to maintain a new name in
working memory, you can simply continue to mentally rehearse it.
For example, as long as I am repeating the name “Christa
Rodriguez” over and over, I can maintain that name in working
memory. Indeed, if I continuously repeat it, I can keep that name for
hours in my working memory. However, as soon as my thoughts drift
off elsewhere, the new information replaces the rehearsed name in
working memory.

Rehearsal: Actively maintaining the items in working memory by

repeating them over and over (maintenance rehearsal) or by elaborating
on the item to some other concept (elaborative rehearsal).

Maintenance rehearsal: Repeating information over and over.

 Elaborative rehearsal: Processing the meaning of information in
working memory.

Elaborative rehearsal means associating the item in working memory

to existing long-term memory structures. So, for example, if instead
of repeating the name over and over, I thought about whether Mx.
Rodriguez looked like other people I know or tried to link them with
other people with the same first name or same last name or even
thought about whether they look honest or not to me, I would be
engaging in elaborative rehearsal. Elaborative rehearsal takes more
attention but produces better encoding into long-term memory
because it creates more retrieval cues that are useful for later recall.

Once you stop maintenance rehearsal, there is only a limited amount

of time before information is forgotten (or replaced) in working
memory. When you move on from “Christa Rodriguez” to “Sanjay
Parekh,” it is the new name that is now being maintained in working
memory rather than the old one. Therefore, one important question
is, what is the rate of forgetting once you stop rehearsing the
information?

Most estimates of the duration of information unrehearsed in working

memory are between 15 and 30 seconds. After that, information is
lost from working memory. This has a counterintuitive implication: If
you remember something after not thinking about it for one minute,
you are retrieving that information from long-term memory, not
working memory. Thus, for example, if I remember “Christa
Rodriguez” just one minute after hearing their name for the first time
when that minute was consumed with hearing other names, I am
remembering them from long-term memory. This is an important
point, because the science here diverges from popular thinking
about the nature of memory. In this sense, working memory is a very
short-term system, and uses of memory over the short term in our
everyday lives are often subsumed by long-term memory function
rather than from retrieving them from the limited duration of working
memory.
What seems to be important with respect to the duration of
information in working memory is whether information becomes
activated that interferes with it. If we could somehow keep our minds
completely blank, information might not spontaneously decay from
working memory. Keeping our minds blank while awake is
notoriously difficult. However, people who meditate are particularly
good at keeping their minds free of interfering information, and it
turns out there is a correlation between meditation and good working
memory (Moss, Monti, & Newberg, 2013). But for most of us, new
information is always entering working memory. This new information
engages working memory, becomes part of the items being kept
there, and displaces other information that is not being rehearsed.

The basis of the estimate of the duration of information in working

memory comes from two classic sets of experiments done by John
Brown (1958) in the United Kingdom and Lloyd and Margaret
Peterson in the United States (1959). The two groups worked largely
independently but published similar experiments within months of
each other. Both researchers were interested in forgetting from
working memory but also demonstrated the duration of information in
working memory (or what they referred to as primary memory).

In the experimental paradigms, participants were presented with

three words to remember (i.e., apple, hammer, shell). Following the
presentation of the three words, the participants were given a
number, such as 417, and asked to count backward by threes from
that number (i.e., 417, 414, 411, 408, and continuing down). The
counting task served as a rehearsal prevention task, preventing
participants from repeating the words previously shown. The
research teams then varied the amount of time required to count
backward. In some conditions, the participant might only count
backward by threes for three seconds, but in other conditions, the
participant might be required to count backward for two minutes.
After the rehearsal prevention period was over, the participants were
asked to retrieve the three words given to them prior to rehearsal
prevention. After retrieving the three words, the participants were
given three new words to remember and then given a new number
from which to start counting backward. In this way, the researchers
could look at the effect of the amount of time spent on rehearsal
prevention and the recall of the words. The method is illustrated in
Figure 3.4.

Duration of information in working memory: The amount of time

information will remain in working memory if not rehearsed.

Rehearsal prevention task: A task that prevents a participant from

maintaining information in working memory.

Description

Figure 3.4 ■ Example rehearsal prevention task.

The findings are illustrated graphically in Figure 3.5. More time spent
in the rehearsal prevention task led to lower recall of the words
presented. Indeed, much forgetting occurs after just five seconds of
rehearsal prevention. L. R. Peterson and Peterson (1959) estimated
that within 18 seconds, all information stored in working memory was
lost, and only information that had somehow gotten into long-term
memory accounted for the 20% recall. They noted that after 18
seconds, the curve had hit an asymptote (flattened out) and
remained essentially the same even at much longer retention
intervals.
Description

Figure 3.5 ■ Typical results from the Brown-Peterson task. y-axis

represents the percentage recalled.
Source: Peterson 1959

It should be noted that typically, even at longer retention intervals,

the participants remembered all three words from the very first trial.
Apparently, the number counting did not produce sufficient
interference to cause forgetting of the first trial, as these items were
easy to encode into long-term memory. This is a common problem in
studies of working memory; one always has to be careful that the
retrieval is from working memory, not long-term memory, and indeed,
in the Brown-Peterson task, retrieval from long-term memory can be
seen on the first trial. However, on subsequent trials, memory for the
words was interfered with by the combination of words from earlier
trials and the rehearsal prevention task. Thus, the results shown in
Figure 3.5 are a function of the average across many trials.
Both the Petersons and Brown favored an explanation based on
decay—that is, information not being rehearsed naturally vanishes or
decays after 20 seconds or so. However, subsequent research
strongly supports an explanation based on interference.
Interference means that new information enters working memory
and displaces information already present. Because the capacity of
working memory is limited, new information will necessarily displace
old information. In the case of the Brown-Peterson-Peterson
experiment, the numbers being spoken during the rehearsal
prevention task entered working memory and displaced the words
that the participant initially was asked to remember. The more time
spent counting backward, the more likely those words were
displaced by the interfering numbers. Thus, the best explanation for
the results is that the numbers being spoken during rehearsal
prevention replaced the words in working memory, thus interfering
with their presence in working memory. Indeed, a few years later,
Keppel and Underwood (1962) showed that part of the interference
that created forgetting in working memory came not simply from the
rehearsal prevention but also from previous to-be-remembered
items. Keppel and Underwood showed that on the first trial,
participants could remember all of the words, even when the
retention interval was long. However, as more trials occurred, it was
more difficult for participants to remember digits at any retention
interval because earlier items were now interfering with later items.
Thus, interference played a bigger role in forgetting than did decay.

Interference: New information enters working memory and displaces

information already present.

Another classic experiment demonstrated the role of interference in

working memory. Waugh and Norman (1965) presented participants
with a sequential list of 16 digits. After viewing all 16 digits, the
participants were presented with one of the digits that they had seen
in the list. Their task was to recall the digit that occurred just prior to
the probe digit during the sequence. In other words, if part of the
sequence was 1-5-6-2-9 and the probe digit was 6, the participant
should reply with the digit 5. Waugh and Norman could then examine
performance on this task as a function of where in the sequence the
probe digit was. If more digits occurred after the probe, then more
new digits were likely to interfere with and replace the memory of the
probe digit and the digit that preceded it. If the probe digit occurred
toward the end of the sequence, there should have been less
information to interfere with it, and therefore memory of the digit that
preceded it should be better. This is exactly what Waugh and
Norman found—the fewer items that followed the probe digit, the
better memory was for the item that preceded it. This cemented the
view that forgetting from working memory originates from
interference.

The Serial Position Curve and Its

Implication for Working Memory
In light of what we know now about the functioning of the brain and
memory, it is quite clear that separate neurocognitive memory
systems handle different kinds of information. For example, memory
should be divided into short-term memory systems and long-term
memory systems. With respect to long-term memory, most
researchers agree on the distinction between semantic and episodic
memory. However, 50 years ago, on the basis of the principle of
parsimony (that is, opting for the simplest theory when possible),
many researchers argued that the brain had just one memory
system that could be used in many different ways. Thus, one of the
goals of early researchers in the field of working memory was to
show how it was different from long-term memory (i.e., Atkinson &
Shiffrin, 1968; Baddeley & Hitch, 1974; see Baddeley, 2018). Some
of the earliest evidence for the separation of working memory from
long-term memory came from a deceptively simple procedure, called
“free recall,” of single-item lists, which is typically measured by a
serial position curve. A serial position curve plots the order of items
presented and how well each of those items was recalled across
participants.
 Serial position curve: The observation that participants remember
items well from the beginning and end of a list but not from the middle.

In a free-recall test, participants are read (or read themselves) a list

of words, usually randomly chosen and with little associative
structure. Immediately following the reading of the list, participants
are asked to recall (usually by writing down) as many of the words as
possible. As you can see, nothing could be simpler. Yet this test is
quite powerful.

After the participants can no longer recall any more words from the
list, the experimenters can examine the amount recalled as a
function of serial position—that is, the order in which the words
appear. So the first word on the list in Figure 3.6 is medal. The
experiments would examine the percentage of participants who
successfully recalled the word medal at serial position 1. The
experimenters then examine recall at the next serial position—that
is, for the word paintbrush in this list. Thus, the percentage recall can
be examined as a function of what serial position a word occupies in
a given list. The standard result of such a test is illustrated in Figure
3.7.

Description

Figure 3.6 ■ Read aloud the following words. Immediately after

reading the words, write down as many as you can. Then plot your
total recall as a function of input order. However, you do not have to
recall the words in order.
Description

Figure 3.7 ■ Recall as a function of item position in a serial position

curve. y-axis represents the percentage of words recalled.

There are a few important aspects of the serial position curve. First,
you can see that recall is very good for the first few items on the list.
This effect is called the primacy effect. You then see a big dip in
performance for items in the middle of the list. Then right at the end
of the list, the words become easier to recall again. This good
performance for words at the end of the list is called the recency
effect. Primacy and recency effects can be seen under a wide range
of conditions in which people must recall words or other items
immediately. Indeed, monkeys and other primates also show
primacy and recency effects in experiments that test them for serial
memory (Buchanan, Gill, & Braggio, 1981; Wright, Santiago, &
Sands, 1984).

Primacy effect: The observation that memory is usually superior for

items at the beginning of a serial position curve; thought to be caused by
the encoding of those items into long-term memory.

Recency effect: The observation that memory is usually superior for

items at the end of a serial position curve; thought to be caused by the
maintenance of those items in working memory.
Primacy effects appear to result from the encoding of information
into long-term memory, even though the memory test may be
administered less than a minute after the participants originally heard
the words. Much research supports this point of view. In terms of
people’s self-reports, many participants describe trying to remember
the words by constructing a story. For example, referring to the list of
words in Figure 3.6, they might think of a story in which a “medal” is
awarded to a person who invented a “paintbrush” key on a
“typewriter,” which can be used on the “sofa.” The elaborative
encoding necessary to create this story promotes storage in long-
term memory. However, if such elaborative rehearsal is prevented,
such as by presenting words at a particularly fast rate, the primacy
effect can be reduced or eliminated. Reducing the amount of time
per word lowers the primacy effect in free-recall studies. In contrast,
speeded presentation does not affect the recency portion of the
serial position curve, which is based on working memory (Glanzer &
Cunitz, 1966; Murdock, 1962).

Other evidence also supports the claim that the primacy effect is
caused by retrieval from long-term memory. Mistakes made on early
items tend to be related to their meaning, a key component of
representation in long-term memory. For example, medal might be
mistakenly recalled as award, whereas paintbrush might be
remembered as toothbrush. The mistake here is because we encode
information into long-term memory mostly in terms of what the
information means. Because medal and award overlap considerably
in meaning, errors of this sort tend to be thought of as resulting from
long-term memory (Conrad & Hull, 1964).

Recency effects appear to be based on retrieval from working

memory. As with the primacy effect, a great deal of research now
supports this idea. Think about doing the task yourself. If the first few
items you wrote down were the last items, you were probably likely
to remember several of them. If, however, you tried to recall the list
in order from start to finish, you probably did not recall the last two or
three words from the list. If you ask study participants to wait 30
seconds (that is, if you introduce a retention interval) before writing
down the words that they can recall, the recency effect disappears
(Glanzer & Cunitz, 1966). However, this delay does not affect the
primacy effect at all. This is because the participants’ working
memory has been engaged in other activities, causing interference
with the items at the end of the list. This effect is stronger if you give
participants a rehearsal prevention task so that they cannot rehearse
the last few items in working memory before the test. In contrast,
items recalled from the beginning of the list are in long-term memory,
so the extra 30 seconds do not affect their strength in memory.

Furthermore, errors in the recency effect part of the serial position

curve tend to be based on sensory errors, either visual or auditory
(Conrad & Hull, 1964; Laughery, Welte, & Spector, 1973). For
example, folder might be recalled as bolder, or market might be
recalled as markup. This kind of mistake is characteristic of working
memory, which is more dependent on sensory characteristics than is
long-term memory.

Finally, participants who write down the words from the recency
portion of the curve first remember more total items than those who
try to retrieve the words in order. Participants who retrieve in order
remember the words from early in the list based on long-term
memory, but by the time they get to the end of the list, those words
have already been eliminated from working memory.

For this reason, the serial position curve—the graph that shows good
memory for both the beginning of the list and the end of the list—in
immediate free-recall tests demonstrates that working memory and
long-term memory have different properties. Variables that affect the
primacy effect are variables that affect long-term memory, and these
variables do not influence the recency effect. By contrast, the
variables that affect the recency effect are implicated in working
memory and do not affect the primacy effect.

In free recall of recently presented lists, the recency effect is due to

retrieval from working memory. But it turns out that the primacy and
recency effects are not just restricted to free recall of lists of
randomly grouped words. Both primacy and recency effects exist in
other memory situations as well, including situations in which
retrieval is strictly from long-term memory. For example, try writing
down the names of as many U.S. presidents as you can. Take a
moment away from the textbook now—so you don’t see any of the
names below. It turns out that recalling the names of the U.S.
presidents shows a primacy effect and recency effect. Most people
remember Washington, Adams, and Jefferson on the one hand and
Bush, Obama, and Trump on the other. However, few people will get
Chester Alan Arthur, Millard Fillmore, or Grover Cleveland in the
middle of the list. Remembering the first presidents is equivalent to
the primacy effect, whereas the recall of the most recent presidents
can be considered a recency effect (Roediger & Crowder, 1976;
Roediger & Magdalena, 2015).

Another example comes from watching the advertisements during

the Super Bowl. The advertisements have become almost as
important a part of the show as the football game itself. But which
advertisements are remembered the best (an important piece of
information for the advertisers, who are paying a tremendous
amount of money for airtime)? It turns out that the ones at the
beginning and the ones at the end are recalled better than those
from the middle of the game (Brunel & Nelson, 2003). Primacy and
recency effects are also seen in the retention of information from
academic classes. Information from the beginning of the semester
and the end of the semester is remembered better than information
from the middle of the semester (Conway, Cohen, & Stanhope,
1992).
Section Summary and Quiz
Working memory is a short-term memory system that can hold
information for a short period of time for conscious introspection.
Originally, it was thought to hold about seven items, but more recent
views of working memory suggest that factors such as word length
and pronunciation time affect the amount that can be maintained in
working memory. If unrehearsed, information will fade from working
memory in approximately 15 to 30 seconds. When people learn a list
of words and then have to recall them immediately after
presentation, the recency effect is caused by retrieval from working
memory, whereas the primacy effect is because of the early items in
the list having entered long-term memory. Variables that affect
working memory affect the recency portion of the serial position
curve; errors here will reflect working memory processes. However,
serial position curves are common in memory, and primacy and
recency effects can also occur in retrieval from solely long-term
memory.

Quiz

1. Which is the current term used to refer to systems of memory that

hold information up to 30 seconds?
1. Primary memory
2. Short-term memory
3. Working memory
4. Unemployed memory
2. In Sperling’s classic 1960 experiment on sensory memory,
participants in the partial-report condition were expected to
1. Report all of the material when they heard any tone
2. Report only one line when cued by a particular tone
3. Report all the vowels they saw in the matrix
4. Report all the consonants they saw in the matrix
3. One factor that affects the amount of information currently in
working memory is
1. The pronunciation time of the items if they were spoken
aloud
2. The ability of a person to chunk information
3. Whether the words are long or short when spoken
 4. All of the above are factors
4. A person in a serial position curve experiment recalled the word
violin when actually the word violence had been presented. Based
on what you know about the serial position curve, where would you
expect the word violence to have been presented?
1. Right at the beginning of the list
2. In the middle of the list
3. As one of the last few items in the list
4. It is impossible to tell from the information provided

1. c
2. b
3. d
4. c

The Working Memory Model of

Baddeley
Alan Baddeley is a British memory psychologist who has contributed
to many areas in psychology. Though now well into his 80s, he still
continues to contribute to both research and theory. His most
influential theory is of working memory (Baddeley, 1986, 2007, 2012,
2018; Baddeley & Hitch, 1974), a theory that has both grown and
changed over a 40-year time course. Baddeley and his theory are
probably most responsible for the change of terminology from short-
term memory to working memory. For Baddeley, the term working
memory more genuinely reflected what working memory is for: to
work with the active contents of consciousness.

But another aspect of Baddeley’s working memory theory makes it

different from the theories of short-term memory that came before it
—he advanced the idea that there are actual multiple working
memory systems. He called these various systems “subsystems” or
“slave systems.” The word subsystems is obviously preferred now
and used in this book henceforth, but if one goes back and reads the
earliest papers, expect to see the less-sensitive term. For all intents
and purposes, what he proposed is that we have separate working
memory systems for each major perceptual modality. He called
visual working memory the visuospatial sketchpad, and he called
auditory working memory the phonological loop (Baddeley, 1986).
In later versions of the model, Baddeley introduced another system
called the episodic buffer, which coordinates overlap between the
auditory and visual systems (Baddeley, 2018). The episodic buffer is
also the link between working memory and long-term memory.
Coordinating the activities of these two systems is an attentional
mechanism that he called the central executive (see Figure 3.8).
We will now examine what each of these terms means and what
evidence exists to support the theoretical constructs developed by
Baddeley and his colleagues over the past 45 years.

Visuospatial sketchpad: Visual working memory.

Phonological loop: Auditory working memory.

Episodic buffer: Coordinates overlap between the auditory and visual

systems and also interfaces working memory with long-term memory.

Central executive: The attentional mechanism of working memory.

Description

Figure 3.8 ■ Baddeley’s working memory model.

Source: McBride (2016)

Baddeley originally stumbled onto the concept that working memory

must be composed of multiple systems while working on tasks that
were designed to look at how working memory handled multiple
simultaneous tasks. He and his colleagues were conducting
experiments in which they asked participants to engage in two
working memory tasks at the same time. All the old models predicted
interference between the two tasks, leading to diminished
performance in at least one, if not both, tasks. What Baddeley and
Hitch (1974) discovered, in contrast, is that this is not always the
case. In some instances, people can successfully do two working
memory tasks at the same time without interference—that is, they
can be successful at both tasks. When no interference was
discovered, it tended to be because one task was visual and one
task was auditory. This led Baddeley to the conclusion that working
memory was separable among sensory systems. For example,
consider the visual working memory required to drive a car. Your
visual attention must be on the road, which is why texting and driving
is such a life-threatening combination. However, the visual attention
needed for driving does not prevent you from devoting auditory
working memory to listening to music or talking to a passenger.
Similarly, in a psychological experiment, a participant can track a
moving arrow on a screen and rehearse digits without a deficit in the
ability to do either task.

The first published evidence to support the idea that working

memory was not unitary and might consist of multiple systems goes
back to a landmark study by Baddeley and Hitch (1974). In the study,
participants were read aloud strings of digits to remember. These
strings varied from simply one numeral to as many as eight.
Participants were required to speak back the digits. This task was a
standard digit span task, with spans ranging from trivially easy to
more difficult. However, Baddeley and Hitch also asked their
participants to do a concurrent task (a task to be done
simultaneously with the first task). In the concurrent task, the
participants had to judge whether simple sentences were true or
false. The participant might see the letters JK, and underneath would
be a statement, “The J is before the K” (true), or the participant might
see, “The K is before the J” (false). In such a study, the independent
variable is the presence or absence of the concurrent task, whereas
the dependent variable is performance in saying the digits one just
heard.

Concurrent task: A task to be done simultaneously with another task.

Based on the models of the time of working memory, Baddeley and

Hitch (1974) expected that the concurrent task would provide an
additional burden to working memory and interfere with recall in the
digit span task. However, that is not what they found. Instead, there
was almost no overlap between the two tasks. The retrieval of digits
was just as good as it would have been without a concurrent task,
and participants did as well on the concurrent task when they were
rehearsing eight digits as when they rehearsed only one digit. There
was no interference between the two tasks. For these two tasks, the
participants really could do two things simultaneously, much as
someone can drive and listen to music at the same time.

These results are interpretable when one thinks of the visuospatial

sketchpad and the phonological loop as separate subsystems. The
phonological loop handles the digit span task, whereas the
visuospatial sketchpad handles the reading and processing of the
simple reasoning task. Because both of these components did not
need to tax the attentional mechanism (the central executive), both
tasks are done at the same time without any interference. Thus, this
study counts as evidence that visual and auditory working memory
can function independently of each other.

In another experiment specifically designed to test the new model of

working memory, Logie (1986) asked participants to learn paired
associates (digit–word pairs, such as 23–typewriter or 12–
candlestick). Some participants were instructed to use a visual
strategy to encode the words (that is, to employ the visuospatial
sketchpad by making a mental image of the association), whereas
other participants were instructed to use rote encoding (to employ
the phonological loop). In the concurrent task, Logie either presented
pictures (visual) or required participants to listen to names (auditory).
The participants had to see or hear the items—they were not
required to encode them. Nonetheless, the results were striking: The
visual concurrent task interfered with the learning of paired
associates when the associates were learned using the visual
strategy, and the auditory concurrent task interfered with learning of
the paired associates when the associates were being learned using
rote encoding. In contrast, cross-sensory interference was much
less. Looking at pictures did not interfere with rote encoding, and
hearing names did not interfere with the visual learning strategy. To
restate the findings another way, when the learning task required
visual imagery, viewing pictures interfered with learning, but hearing
words did not. When the task required auditory processing, hearing
words interfered with learning, but seeing pictures did not. This
supports the idea that the visuospatial sketchpad is a different
system than the phonological loop.

Now contrast the Baddeley and Hitch (1974) and Logie (1986) data
with those from another experiment. L. R. Peterson and Johnson
(1971) also did a digit span task with a simultaneously performed
concurrent task. Peterson and Johnson asked participants to repeat
simple words over and over (e.g., the, the, the, the …) while they
were also supposed to be rehearsing the digits for the digit span
task. Because both tasks now involved the phonological loop, this
concurrent task reduced the number of digits that could be
remembered. This kind of interference is called articulatory
suppression. You might think that silently repeating the word the
would not interfere with processing, as it is such an easy task. But
because it requires some use of the phonological loop, it interferes
with other tasks that employ the phonological loop. In the Peterson
and Johnson experiment, articulatory suppression lowered
participants’ digit spans.

Thus, a basic principle can be derived from these experiments. As

long as the attentional demands are not great, visual working
memory (visuospatial sketchpad) tasks should not interfere with
auditory working memory (phonological loop) tasks. By the same
token, auditory working memory tasks should not interfere with visual
working memory tasks. In contrast, even relatively easy tasks within
the same working memory system will interfere with each other. That
is, two auditory working memory tasks will interfere with each other,
and two visual working memory tasks will interfere with each other.

Articulatory suppression: A concurrent task that prevents participants

from engaging in rehearsal within the phonological loop.

Mnemonic Improvement Tip 3.2

Keeping your working memory sharp can lead to general cognitive well-
being. Practicing digit spans can be considered mental exercise.
Working Memory Systems
The Phonological Loop
The phonological loop is our auditory working memory system. It
stores sounds, particularly language sounds, for a short period of
time. It also has processes that allow us to rehearse or otherwise
manipulate the information in the short-term store. In this way, it is
like an “inner ear” that stores the sounds we hear in a somewhat
literal format until we can process them in terms of their meaning
and store them in long-term memory. Critical to the idea of the
phonological loop is that it is a limited-capacity system that holds
auditory information for a brief period (Baddeley, 2018).

We have already discussed the initial findings of Baddeley and Hitch

(1974), which support the distinction between the phonological loop
and the visuospatial sketchpad. Only a visual concurrent task
interfered with the visuospatial sketchpad, and only a phonological
concurrent task interfered with the phonological loop. The word
length effect also supports the notion of the phonological loop. The
fact that words that take longer to say are more difficult to maintain in
working memory is consistent with the idea that an auditory-based
working memory system is responsible for that information.
Interestingly, Baddeley, Lewis, and Vallar (1984) eliminated the
advantage of short words over long words when participants were
simultaneously required to repeat a particular sound, such as the
word the, under their breath. Repeating the created articulatory
suppression, which involves using the phonological loop and forcing
the participants to rely on other systems, such as the sketchpad, to
remember the words. In the sketchpad, the shorter words are not an
advantage. Thus, the fact that articulatory suppression eliminates the
word length effect supports the idea of a separate phonological loop.

Research on articulatory suppression is critical for supporting the

idea that the phonological loop is independent of other working
memory systems. In an interesting variant on articulatory
suppression, Otsuka and Osaka (2015) asked participants to do
mentally a simple arithmetic task—that is, two-digit addition. While
participants were engaged in the addition task, some were asked to
engage in simultaneous secondary tasks. When the secondary tasks
involved articulatory suppression, participants who were relatively
poor at arithmetic showed a greater decrease in performance than
those who were more proficient in arithmetic. It is likely that the
better performers used visual methods to add the numbers and may
also have had less attention needed for the task, so their
performance remained high despite the articulatory suppression.
However, the participants who were less good at arithmetic used a
more auditory strategy to add the numbers and were therefore more
prone to the interference caused by the articulatory suppression.

Irrelevant speech effect: The observation that the phonological loop is

mildly impaired in the presence of background speech.

Other evidence that supports the existence of a phonological loop

comes from the research on irrelevant speech effect. The
irrelevant speech effect refers to the observation that the
performance of the phonological loop is mildly impaired when talking
occurs in the background. Irrelevant speech affects performance on
phonological loop tasks but not visuospatial sketchpad tasks. For
example, Salame and Baddeley (1989) asked participants to
maintain information in working memory while listening to singing,
music without singing, or no sounds at all (see Figure 3.9). The silent
group performed the best on the working memory task, but the group
that listened to music without singing outperformed the group that
listened to singing. Thus, sounds of any nature, particularly
meaningful sounds, interfere with our ability to maintain information
in the phonological loop. Irrelevant speech is also important in why
we find overheard cell phone conversations so distracting
(Emberson, Lupyan, Goldstein, & Spivey, 2010). Because the
overheard conversation occupies the phonological loop, it interferes
with the information we wish to maintain in our phonological loop.
Moreover, because we cannot hear the other person in the
conversation, we are lacking information to process the cell phone
conversation, and thus central-executive processes are drawn in.
Such central-executive processes would not be needed if we heard
both sides of the conversation. Thus, overheard cell phone
conversations are more distracting than overheard physically present
conversations (Emberson et al., 2010). In this study, the independent
variable is the kind of conversation heard. The dependent measure
is memory performance.

Description

Figure 3.9 ■ Errors as a function of condition. y-axis represents the

percentage of errors made during the task.
Source: Based on information from Salame and Baddeley (1989).

The irrelevant speech effect has implications for mnemonic

improvement as well: It suggests that studying in a quiet room leads
to a more efficient environment for learning, as information can be
processed in the phonological loop more efficiently. This contradicts
the so-called Mozart effect, which tells us that listening to Mozart can
help us learn new information. The Mozart effect is a widespread
myth of learning—there is no evidence that listening to Mozart can
improve memory efficiency. In fact, research suggests that listening
to Mozart did not have positive effects on immediate learning of
spatial information (Lilienfeld, Lynn, Ruscio, & Beyerstein, 2010; T. L.
Wilson & Brown, 1997).
Evidence supporting the existence of the phonological loop comes
from the neuropsychological domain as well. Vallar and Baddeley
(1984) examined a patient with selective damage to the phonological
loop as a function of damage to the left frontal cortex. The patient
tested normally in most areas of cognitive functioning. However, her
working memory, as measured by digit spans, was severely
impaired. Moreover, she did not show the word length effect,
meaning that longer words were not more difficult to keep in working
memory than shorter words. Vallar and Baddeley claimed that this
occurred because whatever words she could recall were coming
from the sketchpad rather than the loop. The longer words may have
taken longer to pronounce, but because she was not using the
phonological loop, whatever she did retain in working memory was
based on visual processing. Her working memory was also
unaffected by phonological similarity (that is, words that sound
similar are easier to maintain in working memory), also suggesting
that only the phonological loop was affected and not the sketchpad.

Visuospatial Sketchpad
The visuospatial sketchpad stores visual and spatial information for
short amounts of time in the activated contents of consciousness. It
is largely independent of the phonological loop as long as attentional
demands are low. Like the phonological loop, it is a short-term
memory system, designed to hold visual and spatial information for
short periods of time, up to about 30 seconds. Recent research
suggests that it is likely that visual and spatial information may be
represented in different subsystems (Baddeley, 2018), but we will
treat them together here.

The visuospatial sketchpad is open to conscious introspection. For

example, we experience the visuospatial sketchpad when we
retrieve what a familiar person looks like. If you have a mental image
of your best friend’s face in your mind’s eye, it is being represented
in your visuospatial sketchpad. When you close your eyes and think
of the Manhattan skyline, you are using your visuospatial sketchpad.
If you glance briefly at a map while driving and then try to figure out
where you are supposed to get off the highway, you are using your
visuospatial sketchpad to represent that information. Like the
phonological loop, the visuospatial sketchpad occupies a role in
cognition that brings thoughts and perceptions to our conscious
awareness. Thus, whatever you are attending to visually at the
moment is the contents of your visuospatial sketchpad.

A classic experiment on the nature of the visuospatial sketchpad

comes from an experiment done by Lee Brooks (1968). Brooks’s
original purpose was to explore the nature of imagery, but his
experiment provides an excellent example of the independence of
the visuospatial sketchpad from the phonological loop. Brooks asked
participants to imagine letters that were not actually present, such as
the letter F (see Figure 3.10). Because participants were asked to
make a visual image, we can assume that the representation of this
image is being held in the visuospatial sketchpad. Keep in mind that
the participants could not actually see the letter—they only had a
mental image on which to base their decisions. Participants were
then asked to make judgments about the letter, such as whether the
angles in the letter were obtuse (greater than 180 degrees) or acute
(less than 180 degrees). Look at the following F: Is the angle made
by the line that goes straight up and the line that goes out to the right
acute or obtuse? Then try answering the same question with a T
without actually looking at one. Participants had to rely on their
image of the letter and were not allowed to draw one. Most
participants found this task challenging yet possible.
Figure 3.10 ■ Imagine the letter F. Examine the corners where the
lines making up the letter come together. Are the angles acute or
not?
Source: Based on Brooks, L. (1968).

Brooks’s (1968) important experimental manipulation concerned how

participants made their responses. Responses were made by
speaking the answers aloud (phonological), tapping them out with
the hands (i.e., one tap for yes; two taps for no; motor), or pointing to
a field with an array of Ys for yes and Ns for no (visuospatial). Note
that neither speaking nor tapping requires use of the visuospatial
sketchpad. That is, the independent variable is the form of the
response, whereas the dependent measure is the speed and
accuracy of the response. Thus, these response options should not
interfere with the imagery task. Pointing to letters on a display,
however, does require visual processing, so the pointing response
also employs the visuospatial sketchpad. When Brooks examined
his results, he found that participants were more accurate and faster
when they had to speak the answer or tap the answer than when
they had to point to the array. Thus, only the visual response task
interfered with the imagery task. Because of the interference across
the visual domain, performance decreased in the imagery task.
Thus, as we have seen throughout, interference works greater within
domains. Visual interference hurts performance on tasks that use the
visuospatial sketchpad, but auditory tasks interfere less with tasks
that use the visuospatial sketchpad.

In the Otsuka and Osaka (2015) study, the researchers also looked
at how visual tasks interfered with arithmetic performance. As
described earlier, high performers in the arithmetic task were more
likely to use both visual and auditory strategies to help them with
mental arithmetic, whereas the low performers were more likely to
use auditory strategies alone. As such, it was the higher performers
whose performance was more interfered with when Otsuka and
Osaka introduced a secondary visuospatial task (tapping). So again,
we see that visuospatial secondary tasks interfere with the
visuospatial sketchpad more than they interfere with other working
memory systems.

The Episodic Buffer

The episodic buffer is a short-term memory system that holds
information integrated across auditory working memory, visual-
spatial working memory, and long-term memory for a brief period of
time. Because it maintains information already integrated across
modalities, it serves as the point of interface between working
memory and long-term memory systems (see Baddeley, 2000,
2018). The episodic buffer provides needed meaning or semantic-
based information to the working memory system. For example,
when you retrieve information from long-term memory, such as the
directions to your cousin’s apartment, this information is briefly
maintained in the buffer before it is converted into directions (in the
phonological loop) or a visual map (in the visuospatial sketchpad).
The buffer also allows new information to be integrated before it
reaches long-term memory during encoding. For example, watching
a ballet performance will involve dance information being processed
by the visuospatial sketchpad and music information being
processed by the phonological loop. The episodic buffer will
integrate the dance and music with information from long-term
memory related to meaning in the buffer before the experience is
stored in long-term episodic memory.

Current evidence for the episodic buffer comes from amnesic

patients who have very impaired encoding into long-term memory
but normal working memory (Baddeley & Wilson, 2002). When
encoding stories, they seem to be able to maintain more than seven
items of information in mind while reading the stories. The patients
forget the stories later, but their “intermediate memory” extends past
the usual capacity and duration of working memory. Baddeley and
Wilson (2002) argued that this is because of the integrated nature of
the episodic buffer. The buffer allowed for understanding even if the
patient later forgot the information.
In an experimental study on healthy individuals, Langerock,
Vergauwe, and Barrouillet (2014) examined the episodic buffer by
asking participants to maintain in working memory associations that
crossed the boundaries of working memory systems. Thus, the
participants were required to maintain an association between a
spoken word (phonological loop) and a location on a map
(visuospatial sketchpad). They compared memory on these cross-
system associations to within-system associations (e.g., word to
word association). Because the within-system associations did not
tax the capacity of each system, there was no within-system
interference. But because the episodic buffer was necessary for the
cross-system associations, some attention was needed to maintain
that association, resulting in less accurate performance for the cross-
system associations than for the within-system associations. Thus,
drawing on the episodic buffer required more attention than if only
phonological loop or the visuospatial sketchpad was required, and
therefore the task that required the episodic buffer was more difficult
to do.

The Central Executive

The central executive is an attentional system that supervises and
coordinates the actions of the other working memory components.
Baddeley (2012) argued that the central executive has three main
functions: It can direct attention to a particular source of information,
it can divide attention among sources of input when appropriate, and
it can allow us to switch our attention among competing sources of
input. In this way, the central executive determines what information
will enter the phonological loop, the visuospatial sketchpad, and the
episodic buffer. This makes the central executive an extremely vital
aspect of cognition, as it controls what goes into and out of
conscious awareness. Thus, the goal of the central executive is to
allocate limited attentional resources to the working memory
subsystems. Most researchers think of the central executive as part
of a broader supervisory network, probably located in the prefrontal
lobes, which has many other roles in cognition in addition to
coordinating working memory.
One interesting study designed to probe the role of the central
executive in working memory comes from the work of Teasdale et al.
(1995). They asked participants to generate random numbers at a
rate of one number per second. This is not as easy as it seems. You
must regulate the generation of numbers to ensure that you are not
making any obvious pattern. You must also do it in rhythm. If you try
it yourself, you will find it does require attention—that is, a little bit of
focus to make sure a clear pattern is not developing and that you are
keeping up the beat. Teasdale et al. probed participants about every
two minutes and asked them what they were thinking about. When
participants reported “daydreaming” (not surprising, given the
repetitive nature of the random number–generation task), their
patterns tended to be nonrandom (i.e., noticeable sequences, such
as 1, 2, 4, 8). When they reported concentration on the task, their
numbers were better approximations of randomness. Thus, when
people’s attention was diverted, as when they were daydreaming,
insufficient attentional resources remained to devote to the number-
generation task. The implication is that the central executive—when
directing attention to the task—allowed participants to better produce
the digits in working memory.

A common way of examining the effects of the central executive on

working memory performance is to vary the tasks that a person is
engaging in, so that he or she has to frequently make attentional
adjustments to what is being done (Shipstead & Nespodzany, 2019).
One way of doing this is with something called n-back task. In the n-
back task, a participant is getting a long string of new information,
such as digits. A cue will indicate to the participant that he or she
must report a digit that occurred n digits before (where n will vary
depending on the task requirements). This requires the central
executive system to continually update what is being represented in
working memory, as older items become superfluous, but new items
may be needed to report.

n-back task: In the n-back task, a participant is getting a long string of

new information, such as digits. A cue will indicate to the participant that
 he or she must report a digit that occurred n digits before (where n will
vary depending on the task requirements).

For example, Scharinger, Soutschek, Schubert, and Gerjets (2015)

examined the central executive using the n-back task in an
interesting way. Their participants were given a 2-back test, in which
the digit reported two earlier needed to be reported, a relatively
difficult task. In addition, irrelevant stimuli, known as flankers, were
included in the display. In order to succeed, participants needed to
inhibit processing of the irrelevant flankers and continually update
the digits being stored in the phonological loop. Scharinger et al.
found that inhibiting the flankers activated the central executive, thus
allowing greater attentional control, leading to better n-back
performance when flankers were present than when they were
absent. Thus, the central executive was able to direct attentional
control to the necessary task.

The central executive also plays a role in the impairment of driving

that occurs when drivers are simultaneously talking on cell phones
(Sanbonmatsu, Strayer, Biondi, Behrends, & Moore, 2016; Strayer,
Watson, & Drews, 2011; Tillman, Strayer, Eidels, & Heathcote,
2017). Driving can be an automated task, and Strayer et al. found
little impairment of driving for cell phone users when road conditions
were normal. However, when hazards suddenly presented
themselves, such as swerving cars, rapidly changing lights, and
other obstacles, the drivers using cell phones took longer to respond
and consequently had more crashes than those not using cell
phones. Moreover, Sanbonmatsu et al. found that when people are
talking on cell phones, they are less aware of their driving skills than
when they are not using phones. Thus, ironically, even though
driving performance is impaired, drivers will continue to think they
are driving well, despite mistakes they may make, because they are
not monitoring these mistakes. Thus, accidents will be attributed to
factors other than their cell phone use.

In the event of a dangerous situation, it is likely that the central

executive must disengage attention from the cell phone
conversation, switch it to the current situation, and then allow for a
response. The extra time involved in switching attention from one
task to another can be detrimental and dangerous in driving because
of the high speeds involved. Strayer et al. (2011) also showed that
conversations with passengers in the car do not require as much
attention and thus do not result in more accidents. This is likely
because there is some degree of joint attention. If the passenger
sees a dangerous situation, he or she may disengage in the
conversation, giving the driver more time to react. Because under
normal driving circumstances cell phone use does not interfere with
driving, most drivers ignore data such as those from the Strayer et
al. study. However, the results suggest that, due to demands on the
central executive, using a cell phone slows one’s responses when
one needs to respond most quickly.
Section Summary and Quiz
The working memory model proposed by Baddeley (2018) states
that working memory is actually composed of a number of systems
bound together by an attentional mechanism. The phonological loop
is our auditory working memory system, responsible for maintaining
auditory and speech as conscious content to allow us to process that
information. The visuospatial sketchpad has the same function for
visual and spatial information. And episodic buffer links the two
systems together and is activated when meaning needs to be
extracted from the phonological loop and the visuospatial sketchpad.
A central executive system maintains attentional control across
these systems, allowing some systems to be activated while others
need to be inhibited. Much of the data presented in this section are
dissociations between these systems, presented to convince the
reader of the reality of these separate systems within working
memory.

Quiz

1. If a person is doing two rather easy concurrent tasks, one requiring

the person to monitor sounds and the other requiring the person to
follow a moving cursor, one would expect
1. Massive interference from the central executive
2. Only the visuospatial sketchpad requires use of the episodic
buffer
3. Little to no interference, because the two tasks tap into
different systems
4. Interference will only occur with the visuospatial sketchpad
2. Articulatory suppression occurs when
1. Rehearsal is prevented, because the phonological loop is
occupied by a concurrent task
2. The visuospatial sketchpad is forced into handling
phonological information
3. The episodic buffer is inhibited
4. All of the above
3. In Brooks’s (1968) imagery task, people were asked to imagine an
image of a letter. They were then required to make a decision
about that letter. Brooks found that
 1. Performance was worse when the test task required
participants to use the visuospatial sketchpad
2. Performance was better when the test task required
participants to use the visuospatial sketchpad
3. Performance was worse when the test task required
participants to use the phonological loop
4. Performance was better when the test task required
participants to use the phonological loop
4. One reason why talking on cell phones decreases one’s driving
ability is that
1. Attention must be disengaged in the conversation before it
can be reengaged on the hazards of driving
2. The cell phone conversation uses the phonological loop,
which is critical for driving
3. The cell phone conversation causes peturbations in working
memory, which are hard for the attentional system to allocate
elsewhere
4. All of the above are true

1. c
2. a
3. a
4. a

Working Memory and the Brain

Working memory has also become the intense focus of cognitive
neuroscientists recently, with the central executive being of particular
interest and importance. Baddeley’s (2018) multiple-component
model has found great support in brain-based studies. Indeed, the
phonological loop, the visuospatial sketchpad, and the central
executive appear not to be based on fine distinctions in areas of
adjacent cortex but seemingly housed in different lobes of the
cerebral cortex! So let’s start looking at the correlations between
behavioral measures of working memory and the brain. We will start
with a neuropsychological study and then consider the more recent
neuroimaging research.
Warrington and Shallice (1969) studied a young man with brain
damage identified in their paper by the initials K. F. His brain was
injured as the result of a motorcycle accident in England. K. F.’s
long-term memory was unaffected, both in terms of new learning and
retrieving prior knowledge. However, his working memory was
severely impaired. On a digit span task, he could not recall spans
longer than two digits when tested auditorily. Think about this—if you
read out the numbers 5-8-9 to him, he would not be able to repeat
them back to you. His visual working memory, however, was
somewhat better, though still impaired. This pattern suggests that the
problem was more with the phonological loop than with the
visuospatial sketchpad. Subsequent examination of his brain
confirmed this pattern—K. F.’s damage was in the left temporal lobe,
in areas typically associated with speech comprehension.

Jonides (1995), in one of the earliest positron emission tomography

(PET) studies to look at memory phenomena, tested Baddeley’s
model. Jonides examined the brains of participants engaged in tasks
designed to measure the phonological loop or the visuospatial
sketchpad. In the task designed to examine the phonological loop,
participants watched a sequence of letters at a rate of one letter
every three seconds. If a given letter was the same as one two
spaces back (i.e., six seconds ago), the participant indicated this by
saying yes. If not, the participant said no (this is the n-back task
again). Most people will do this task by mentally rehearsing the digits
as they appear. Thus, even though the presentation is visual, it is
sensible to consider this a task for the phonological loop. To occupy
the visuospatial sketchpad, participants saw three dots presented in
different locations for a 200-msec interval. The dots then
disappeared, and three seconds later, a circle appeared. The
participants indicated whether this circle marked a spot where one of
the dots had been just prior.

The PET data showed that the two tasks led to different patterns of
activation in the cerebral cortex. The phonological loop task was
associated with activity in Broca’s area of the left frontal lobe (known
as an important area in speech) as well as areas in the left parietal
lobe. In contrast, the visuospatial sketchpad task led to activation in
the right occipital lobe (visual processing) as well as the right parietal
and right prefrontal lobes. Thus, the distinction between the
visuospatial sketchpad and the phonological loop is supported by
PET studies, which show different activation patterns for each task
(see Jonides, Lacey, & Nee, 2005; Jonides et al., 2008). Indeed,
these data suggest that the basis of the phonological loop may be in
the left hemisphere, but the basis for the visuospatial sketchpad may
be in the right hemisphere. So, the early neuroimaging data suggest
strong differences between the neural bases of these two systems.

Neuroimaging data from functional magnetic resonance imaging

(fMRI) confirm that the central executive is an important component
of working memory. Areas of the prefrontal cortex that are known to
be involved in other attentional or monitoring tasks also appear to be
active during working memory tasks, including digit span and other
verbal working memory tasks. For example, PET studies and fMRI
studies agree that verbal working memory tasks activate the right
dorsolateral prefrontal cortex (Chen et al., 2015; Ruchkin, Grafman,
Cameron, & Berndt, 2003; Rypma & D’Esposito, 2003) as well as the
anterior cingulate (Otsuka & Osaka, 2005, 2015). The anterior
cingulate is located toward the back of the prefrontal regions of the
brain. Both of these brain regions (right dorsolateral prefrontal cortex
and anterior cingulate) are important in self-regulation, attention, and
cognitive monitoring as well as verbal working memory. In an
interesting study of the neural correlates of the central executive,
Chen et al. (2015) required participants to scan visual arrays or
auditory lists for “oddball” items—that is, items that did not fit into a
pattern being presented. The oddball items for visual tasks could be
a green dot presented among red dots. The oddball item for an
auditory task could be a note of higher pitch presented among other
notes all of the same lower pitch. These two tasks are both within-
system, so they require less supervision from the central executive.
In a multisensory task, the participant had to detect when both the
color and pitch were oddballs. This task requires more intervention
from the central executive. In this study, the anterior cingulate
(prefrontal) showed a crucial role in detecting the oddball item and
even more so when the task became more difficult, as in the
multisensory condition. Thus, this study confirms the importance of
prefrontal regions to the control exerted by the central executive.

Thus, the following pattern of activity emerges from the

neuroimaging studies: The phonological loop is mainly housed in the
language-related areas of the brain, particularly in areas associated
with the production of speech. The visuospatial sketchpad appears
to be located in areas of the right hemisphere associated with vision
and spatial skills, including the occipital lobe, which is the visual lobe
of the brain. Finally, the central executive is centered in prefrontal
regions of the brain, which are also active in planning, monitoring,
and other executive cognitive functions.

Applications of Working Memory

Reading Fluency
When we read, we use our working memory. In fact, your working
memory—at this very moment—is holding the words (phonological
loop) and ideas (episodic buffer) in this sentence so that you can
understand this sentence. Consider how difficult it must have been
for the patient K. F. to read, given that he could only keep two words
in his working memory at any particular time. It is likely that his
reading was slow and laborious. Given how important working
memory is to reading, it is likely that there may be some connection
between working memory capacity and reading ability (Peng et al.,
2018). Indeed, in children just becoming fully fluent readers (age 12
and under), there is a clear correlation between the capacity of
working memory and reading fluency (Daneman & Carpenter,
1980). For Daneman and Carpenter, working memory is composed
of a short-term memory storage system and a processing system,
much like the systems envisioned in Baddeley’s working memory
model (2018). In a recent meta-analysis, Peng et al. (2018) found a
correlation between working memory ability and reading skill in
children up to the fourth grade. After the fourth grade, it was verbal
aspects of working memory that most correlated with reading ability.
Thus, if the child can maintain more information in their short-term
memory storage, they can understand the material quicker with less
“looking back” (that is, the processing will be more efficient). Stated
another way, the better a child’s working memory is, the better their
reading ability will be.

Reading fluency: The ability to read at speeds sufficient to process and

understand written material.

Daneman and Carpenter (1980) asked young participants to read

sentences and process them for meaning. They were then asked
questions about the meaning of the passage and asked to retrieve
as many words as they could from the end of the last sentence. They
found that participants who could retrieve more of the last few words
also scored higher on comprehension. This supports the idea that
there is a relation between good working memory ability and reading
ability. This advantage continues into college. Daneman and Hannon
(2001) found that young adults with high working memory capacity
did better on their SAT tests than those with lower working memory
capacity. Shipstead, Harrison, and Engle (2015) also found
correlations between measures of working memory, both visual and
auditory, and fluid intelligence. They argue that part of this
advantage comes from the central-executive component. When one
is better able to direct attention, one will have more working space to
devote to problem-solving. On the other hand, this group has also
shown that training working memory to be better does not
necessarily result in better abilities on other tasks (Shipstead,
Redick, & Engle, 2012). Although with practice, we can improve our
ability to remember digit spans and do better on serial position curve
tests and other measures of working memory, improving on these
tasks does not automatically translate to better reading
comprehension, verbal fluency, or multitasking.

Verbal Fluency
Engle (2002) argued that working memory capacity was also related
to verbal fluency—that is, the ability to speak fluently without
pausing. In other words, those of us who intrude frequent “uhs” or
“hmms” in our speech are likely to have less efficient working
memory than those who are less likely to make these errors. Engle
argued that those with larger working memory capacities were also
those who spoke with fewer pauses. To test this, Engle and his
colleagues divided students into groups with high working memory
scores and those with lower working memory scores. The students
with the higher working memory scores were able to generate more
examples in a given category (e.g., tools) during a particular time
period than those with lower working memory scores.

Verbal fluency: The ability to talk without pausing or stopping.

Summary

Working memory refers to the neural structures and cognitive processes

that maintain the accessibility of information for short periods of time in
an active conscious state. Working memory holds a small amount of
information at any one time, which can be maintained by rehearsal.
Rehearsal is the active repetition of information in conscious awareness.
Working memory is the current term for this kind of memory, while
previous generations referred to it by the name short-term memory.
Working memory capacity can be measured by the digit span task or by
the recency effect in serial position curves. Most estimates of the
capacity of working memory indicate that we can maintain about seven
items. However, this is modulated by the length of the words we are
maintaining in working memory. Words that take less time to pronounce
are easier to maintain in working memory. In free recall of serial lists, the
primacy effect is associated with long-term memory, whereas the
recency effect is associated with working memory. Most recently,
working memory is thought to consist of several subsystems, including
the visuospatial sketchpad, the phonological loop, the episodic buffer,
and the central executive. Each subsystem is responsible for one aspect
of the working memory process, with the central executive coordinating
among them. Areas in the prefrontal and medial temporal lobes appear
to be the neural regions mediating working memory. Working memory is
linked to reading fluency and verbal fluency.
Key Terms
articulatory suppression 82
capacity 67
central executive 80
chunk 68
concurrent task 81
digit span task 67
duration of information in working memory 73
echoic memory 65
elaborative rehearsal 72
episodic buffer 80
iconic memory 65
interference 74
irrelevant speech effect 83
maintenance rehearsal 72
memory span task 69
n-back task 88
phonological loop 80
primacy effect 76
primary memory 65
pronunciation time 70
reading fluency 91
recency effect 76
rehearsal 72
rehearsal prevention task 73
sensory memory 65
serial position curve 76
short-term memory 65
verbal fluency 92
visuospatial sketchpad 80
word length effect 70
working memory 63

Review Questions
 1. Why is working memory considered to be the active contents of
consciousness? How does the concept of working memory differ from
the concept of short-term memory?
2. Describe three main differences between working memory and long-
term memory.
3. How does the digit span task measure working memory? How is it
modified by the pronunciation time effect?
4. How did Naveh-Benjamin and his colleagues demonstrate the
importance of pronunciation time on the capacity of working memory?
5. How is the serial position curve measured?
6. Describe one variable that affects the primacy portion of the curve and
one variable that affects the recency portion of the curve.
7. What evidence supports the idea that the visuospatial sketchpad and
the phonological loop are separate subsystems in working memory?
8. What is the role of the central executive in working memory?
9. What neuropsychological evidence exists to support the notion that
working memory is a distinct memory system separate from long-term
memory?
10. How is working memory related to reading ability?

Online Resources

1. For animal working memory, see

http://www.psych.utoronto.ca/users/shettle/sararsch.html.
2. For more on K. Anders Ericsson’s work, go to
https://psy.fsu.edu/faculty/ericssonk/ericsson.hp.html.
3. For a demonstration of serial position, go to
http://cat.xula.edu/thinker/memory/working/serial/.
4. For information on Alan Baddeley, go to
https://www.york.ac.uk/psychology/staff/academicstaff/ab50/.
5. For a nice overview of working memory, go to
http://cat.xula.edu/thinker/memory/working/.

Descriptions of Images and Figures

Back to Figure

The horizontal axis shows labels of four languages – English, Spanish,

Hebrew, and Arabic. The vertical axis shows “Number of Words
Repeated” from 5 to 8, in increments of 0.5. The graph begins at 7.7 for
English, falls to 6.5 for Spanish, rises to 6.6 for Hebrew, and ends at 5.9
for Arabic. All values are approximate.

Back to Figure

The figure shows a set of three tasks. In each task, the three words that
a participant has to remember are shown in the first line. In the second
line, the numbers that the participant counts backward and the duration
are shown. The words and numbers in each set are given below.

First set:

Repeat: Apple, hammer, shell

9 seconds: 417, 414, 411, etcetera.

Second set:

Repeat: Honey, lumber, dragon

15 seconds: 674, 671, 668, etcetera.

Third set:

Repeat: Pigeon, lawyer, marker

3 seconds: 116, 113, 110, etcetera.

The third line in each set reads:

“Retrieve the three words.” A blank line is provided for the words.

Back to Figure

The horizontal axis shows “Seconds” from 0 to 18, in increments of 3.

The vertical axis shows “Percent Recalled” from 0 to 90, in increments of
10.

The graph begins at 80 percent at 1 second, falls to 50 at 3 seconds, to

45 at 6 seconds, to 30 at 9 seconds, to 18 at 12 seconds, and ends at
17 in 15 seconds. All values are approximate.

Back to Figure

The list shows the following words in three columns:

1. medal

2. paintbrush

3. typewriter

4. sofa

5. cushion

6. pasture

7. clock

8. dragon

9. captain

10. carbon

11. lawyer

12. bubble

13. lemon

14. fountain

15. mask

16. lunch

17. water

18. racket

19. market

20. folder

Back to Figure

The horizontal axis shows the labels “Item 1” through “Item 15.” The
vertical axis shows “Percentage of Words Recalled” from 0 to 100, in
increments of 20. The serial position curve begins at 100 percent for
Item 1, falls steeply to 60 percent by Item 6, flattens till Item 7, and again
dips to around 28 percent by Item 9. It remains at that level up to Item
11, rises to about 55 percent by Item 12, and then rebounds to around
93 percent for Item 15. The approximate percentages, by item position,
are listed below:

1: 100

2: 92

3: 89

4: 77

5: 68

6: 60

7: 60

8: 45

9: 28

10: 28

11: 28

12: 55

13: 61

14: 80

15: 93

Back to Figure

In the model, the “Visuospatial Sketchpad” and “Phonological Loop” that

are the visual working memory and auditory working memory are shown
at the same horizontal level. The “Episodic Buffer” is shown between
these two memory systems. At a level higher than the three systems is
the “Central Executive” that coordinates all three systems. The “Episodic
Buffer” is also linked to “Long-term Memory” shown below it. The
components of Long-term Memory include: “Episodic Memories”, “Visual
Semantics”, and “Language.”
Back to Figure

The horizontal axis shows the three groups- “Quiet”, “Instrument”, and
“Voice.” Participants in the groups were tested on maintaining
information in working memory while listening to: no sounds at all,
singing, or music from instruments, respectively. The vertical axis shows
“Percentage of Errors” from 0 to 60, in increments of 10. Each group
shows a bar whose height represents the percentage.

The approximate percentages of errors, by group, are as follows:

Quiet: 38
Instrument: 45
Voice: 55