5 Robert H. Logie (from The Psychologist, 1314 (1999).

Abridged with permission of the

:' author.)
E
f There is fairly widespread, although not universal, agreement among researchers
that working memory is a useful concept, and interest in its characteristics has
grown dramatically in the last few years. Thei result is something of a debate as
to its characteristics (Miyake & Shah [I9991 . . . ), but I will focus on one par-
ticularly successful theory that has been developed largely in Britain and was inspired
by the work of Alan Baddeley and Graham Hitch (Baddeley, 1986). . . .
One version of this theory is illustrated in Figure 1 (adapted from Logie, 1995).
A key feature of the model shown in Figure 1 is that it contradicts the typical view,
in most contemporary introductory psychology textbooks, of short-term memory
or working memory as the gateway between perception and long-term memory.
The reasons for this change should become clear as the article progresses.
Also immediately apparent from the figure is that working memory comprises
a coherent collection of specialised cognitive functions. One group of functions
enables temporary storage of the visual appearance of objects and scenes (the visual
cache), a second group likewise offers temporary retention of verbal material
in terms of sounds or 'phonology' (the phonological loop), while a third offers a
co-ordinating executive function which enables the conscious manipulation of
 '1

EXTENSION

I
Robert I-
Logie Working Memory
~ 0 0 0 0 0 0 ~

Figure 7

information (central executive). All of the components draw on prior knowledge

(the 'knowledge base') and on the products of moment-to-moment perception
(interpreted via the knowledge base).
The model as a whole is derived from experimental work with normal adults
and children, from studies of individuals who have suffered brain damage, from
computational modelling, and from recent work using brain imaging techniques.
I will start with a description of what might be some basic requirements for a
respectable theory of working memory. . . .

Basics of working memory

Clearly, there are some basic features of a usable working memory. One such
feature is its limited capacity. Therefore, only some of our immediate past
experience is retained, with the selection based at least in part on task demands.
However, despite its limited capacity, working memory has to deal with the prod-
ucts of perception from vision, hearing, taste, smell and touch.
WORKING M E M O R Y : AN OVERVIEW

An additional feature is the temporary nature of working memory. This is obert H.

essential for updating moment to moment, and to avoid crowding our mind with
irrelevant information. Nevertheless, it is useful to be able to extend the reten-
tion of crucial information when necessary.
The use of the descriptor 'working' implies that this is not simply a passive,
temporary deposit box for the left luggage of perception (as was one suggested
role for the old concept of short-term memory). The contents of working mem-
ory can be combined with stored knowledge and manipulated, interpreted and
recombined to develop new knowledge, assist learning, form goals, and support
interaction with the physical environment.
The theory of working memory outlined in Figure 1 offers a conceptual tool
with which to understand the nature of the mental apparatus and of the activity
that might support at least some of its basic features.

Visual appearance and location

. . . [I]tis not difficult to describe, from memory, many of the core features
of a scene that we have glanced at a few moments before. Nor is it overtaxing
mentally to count the number of doors in your house by recalling a few details
of the layout of each room.
One way to account for this form of mental ability is to suggest that we have
a temporary memory that can hold information about where objects are and what
those objects are, thereby allowing us to guide hand and arm movements to the
correct location and to pick up the objects in the absence of vision.
This ability is highly adaptive, in that it allows us to recall the location of
objects in our immediate environment that are not currently in view (e.g. behind
us, occluded by larger objects, or in the dark), and to recall those locations when
we change our location or orientation. We can still remember the layout if we
turn around or leave the room.
This temporary memory also allows us to recall and describe the physical appear-
ance or layout of familiar environments, such as the inside of our living room or
the main square of our home town. The details are drawn from our knowledge
base and become available in the temporary memory for the period that the task
requires. . . . I. .

Memory span, counting and language acquisition

Thus far, I have said very little about the more verbally oriented functions of
working memory. At one level, these can be seen primarily as a means to store
phonological properties of words for brief periods. Yet it turns out that the
characteristics of this temporary memory for words - the phonological loop -
have sigdcant implications for a wide range of everyday activities. Any activity
that requires retention of a verbal sequence such as remembering a new telephone
number long enough to dial it, repeating a foreign word or counting objects would
1 rely on this aspect of working memory.
There is now a large literature on this topic suggesting that verbal temporary
I memory is closely linked with the speech system. So, for example, immediate
 EXTENSION

memory span for digits (such as a telephone number) is severely disrupted if the
experimental participant is required to repeat aloud an irrelevant word, such as
'the the the' - a technique known as articulatory suppression.
Moreover, sequences of words that take longer to say, such as 'hippopota-
mus, university, parliament', are more difficult to remember than are sequences
of short words, such as 'zebra, school, policy' (Baddeley et al., 1975). Closely
linked to this finding is the observation that people who can speak quickly tend
to have longer digit spans than do people who speak more slowly (Nicolson, 1981).
This link between spealung rate and memory span has important practical
implications. For example, Ellis and Hennelley (1980) observed that digit span in
Welsh-speaking children was poorer than in English-speaking children. This was
attributed entirely to the fact that the words for digits in Welsh take longer to
pronounce than do the digits in English.
Similar results have been reported showing shorter digit spans in Italian, where
again the words for digits take slightly longer to pronounce than in English (Della
Sala & Logie, 1993), and longer digit spans in Chinese where digit words are
very short (Stigler et al., 1986). In other words, digit span is language specific,
and we should be cautious when interpreting digit span scores across languages
and cultures.
In terms of Figure 1, overt speaking rate appears to act as a constraint on the
rehearsal rate of the phonological loop rehearsal component (indicated by the 'mouth'
and the looped arrow in the diagram). Longer words take longer to rehearse,
therefore the information in the phonological loop begins to decay before it can
be-reactivated by rehearsal of the items.
A second important application for working memory is in counting and
mental arithmetic. For example, when counting we have to keep track of where
we are in the counting sequence at any one time. If we are counting objects, such
as coins or number of events, we continually have to update our mental record
of where we are in the counting sequence. If we are counting an array of objects,
such as words on this page, then we have in addition to keep track of which
words have been counted and which have not. This is particularly important if the
items are scattered randomly in front of us.
Some years ago, Alan Baddeley and I (Logie & bad dele^, 1987) examined
whether the working memory theory could shed any light on how normal adults
accomplish such tasks. We asked participants to repeat aloud an irrelevant word
(articulatory suppression) while they were counting a series of flashes on a com-
puter screen.
Because we already knew that articulatory suppression affects verbal working
memory, we were interested to find out whether normal counting would also
- be affected. It would be affected if the phonological loop were crucial to the
counting process.
. Articulatory suppression had a dramatic effect on counting. Participants rarely
, achieved the correct total, although they could do so if they were not repeating
an irrelevant word. The disruption did not arise bemuse pa*.tiepnts had to do
two things at once, because asking these same peopdi: ta .a~p@*%dly tap their hand
I WORKING M E M O R Y A N OVERVIEW

II on the table or presenting them with random words or numbers had virtually no Robert H.
( effect on counting. It was the requirement to generate repeated speech that was Logie I
i crucial for the interference to appear.
' Therefore, mentally repeating the numbers is crucial for keeping track in a
counting task, and this mental repetition comprises the subvocal rehearsal com-
I ponent of the phonological loop.
!
Given that the phonological loop system had been shown to ~rovidean explana-
I
! tion for important aspects of verbal immediate recall, it appeared that the system
could also enable the process of keeping track moment to moment of where we
are in a counting sequence.
More recent work has shown that a similar account can be given for mental
+ + + +
arithmetic, p r t i d a ~ 1 yfor sums such as 5 3 6 8 2 = 2, in which a
cumulative total is repeatedly updated (Logie et aL, 1994). Each [time the] total
is updated, it is repeated subvocally, and this capitalises on the memory function
of the phonological loop.
Clearly, a large part of mental arithmetic relies on a knowledge base of known
solutions (McCloskey et al., 199I), but mental rehearsal within the phonological
loop appears to offer an account of how we keep track of intermediate totals as
we progress through a sum, as well as where we are in a counting sequence.
The phonological loop component of working memory has proved useful in
studies of language acquisition in young children. Susan Gathercole, Alan Baddeley
and others (e.g. Gathercole & Baddeley, 1989) have shown that young children's
ability to repeat a series of nonsense words at age 3 or 4 predicts their language
ability several years later.
This suggests that the ability to repeat an unfamiliar speech sound (a feature
of the phonological loop) is important for acquiring vocabulary and other language
skills, both in young children and in adults learning a second language.
The loop has also been found invaluable in interpreting the pattern of difficult-
ies encomtered by brain-damaged patients who show deficits of one component of
working memory while having other components intact (for a review, see Della Sala
& Logie, 1993).
Recent developments in neuroimaging techniques offer further converging evid-
ence for the concept of a phonological loop that has both a memory store and a
rehearsal system. Paulesu et a1. (1993) used positron emission tomography (PET)
to measure regional Cerebral Blood Flow (rCBF) during a series of tasks that sys-
tematically varied whether participants had to store a series of letters or mentally
repeat the letter sounds.
By contrasting rCBF observed during the performance of each task, Paulesu
et al. identified mental rehearsal of letter sounds with an area of the brain known
as Broca's area. This area is more commonly associated with aspects of speech
production, but appears also to be involved in 'mental speech' or the rehearsal
component of the phonologcal loop. The letter memory task was associated with
[a different] area known as the supramarginal gyrus. This pattern supported the
idea that the phonological loop comprises separate components for phonological
storage and for mental rehearsal.
Logie
Long-term working memory
EXTENSION

In the spirit of the examples given above, working memory should offer tempor-
1
ary memory for immediate past experience, plus some means to mentally rep-
resent the immediate environment and to manipulate and continually update the
contents of that mental representation. Also it should support the acquisition of
new knowledge, problem solving and decision making (for recent reviews, see
Logie & Gilhooly, 1998). However, none of this can happen in isolation from past
experience.
Knowledge accumulated over our lifetime is clearly available to each of us.
When we think of castles in Scotland, colleges in Cambridge, or television soap
operas, our accumulated knowledge - becomes readily available to us. . . .
Having activated such knowledge, we can then manipulate it and extract novel
information that had not been explicitly stored away, such as whether King's College
Chapel in Cambridge is larger or smaller than Crathes Castle in Aberdeenshire.
Similar questions about television soap operas would bring nothmg to mind for
some, or richly endowed memories of lots and characters for others.
Ericsson and colleagues (e.g. Ericsson & Delaney, 1998) distinguish between
a short-term working memory - on which I have focused thus far - and a long-
term working memory that accounts for the ease with which we can access highly
familiar stored knowledge. The greater our expertise in a particular domain, the
greater is our working memory capacity for information in that domain.
Thus, chess experts can retain details of chess games played simultaneously
even when blindfold (Saariluoma, 1995), and avid soccer supporters can remem-
ber scores from matches more accurately than can the more casual fan (Morris
et al., 1985).
Even experience in crime leads to a form of, albeit undesirable, expertise: !
experienced burglars can remember details in photographs of houses seen a few
moments before better than can samples of police officers or householders (Logie {
et a]., 1992). j
In each case, the expert knowledge allows very efficient coding and retrieval ,
of information within the area of expertise. These memory skills clearly do rely I
on the short-term working memory, but expertise greatly facilitates activation of
relevant information in the knowledge base, and this activated knowledge can offer
significant support for the more limited temporary working memory system.

Glossary
overt speaking rate: the rate of actual speech as compared with the rate at which speech
is rehearsed (repeated) in the mind.