See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/17738683

Independent Functioning of Verbal Memory Stores: A

Neuropsychological Study

Article in Quarterly Journal of Experimental Psychology · June 1970

DOI: 10.1080/00335557043000203 · Source: PubMed

CITATIONS READS

658 16,187

2 authors, including:

Tim Shallice
Scuola Internazionale Superiore di Studi Avanzati di Trieste
320 PUBLICATIONS 55,860 CITATIONS

SEE PROFILE

All content following this page was uploaded by Tim Shallice on 16 February 2016.

The user has requested enhancement of the downloaded file.

 Q. J1 Rxp. Psychol. (1970)22, 261-273

INDEPENDENT FUNCTIONING OF
VERBAL MEMORY STORES :
A NEUROPSYCHOLOGICAL STUDY
T. SHALLICE? AND ELIZABETH K. WARRINGTON
The National Hospital, Queen Square, London

Five experiments are described concerning verbal short-term memory performance

Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

of a patient who has a very markedly reduced verbal span. The results of the first
three, free recall, the Peterson procedure and an investigation of proactive interfe-
rence, indicate that he has a greatly reduced short-term memory capacity, while the
last two, probe recognition and missing scan, show that this cannot be attributed
to a retrieval failure. Since his performance on long-term memory tasks is normal,
it is difficult to explain these results with theories of normal functioning in which
verbal STM and L T M use the same structures in different ways. They also make
the serial model of the relation between STM and L T M less plausible and support
a model in which verbal S T M and L T M have parallel inputs.

Introduction
Clinical studies have made a substantial contribution to the understanding of
memory, despite the small amount of theoretically oriented work in this field.
Observations on the completely irretrievable few seconds or minutes of the retro-
grade amnesic period helped to give consolidation theory a certain plausibility.
Investigation of the amnesic syndrome, particularly the quantitative studies of
Milner (1966), provided strong support for the distinction between long and short-
term memory by demonstrating a marked defect of new learning in patients whose
short-term memory is intact. I n fact, Atkinson and ShifTrin (1968) considered
this the most convincing demonstration of the dichotomy.
The strongest objection to the use of clinical material has been made by Gregory
(1961). Using engineering analogies, he was primarily objecting to the inference
of a brain centre for a particular function when a lesion damaged that function.
Weiskrantz (1968) has pointed out that, in general, Gregory’s assessment of abla-
tion studies is much too pessimistic, and this seems particularly the case for “flow
diagram” models. Engineers often produce the analogue of a lesion when testing
whether a particular model is appropriate for some physical system (Milsum, 1966).
It therefore seems appropriate to use clinical data when generating or evaluating
such models.
Occasionally individual patients show a particularly striking behavioural deficit,
specific and limited to one functional system or process. The mere occurrence of
t Present address: Psychology Department, University College, London.
26 I
 262 T. SHALLICE AND E. K. WARRINGTON

such a dissociation of function provides positive evidence of cortical organisation

which needs to be explained by theories of normal function. What has not so far
been reported is a case where the long-term memory system (LTM), including
trace formation and retrieval, appears normal, even though the short-term memory
system is damaged. Such a case, besides further supporting the distinction between
the two systems, could provide information about how each of them operates and,
in particular, how the two systems are related. A case we have been investigating
appears to be of this type. The patient has a profound repetition defect (in the
neurological literature this deficit is the cardinal feature of “conduction aphasia”),
his digit and letter span being reduced to two items or less; yet on tests of long-
term memory his performance is normal. Warrington and Shallice (1969) reported
a clinical case study of this patient (K.F.), together with evidence to support the
view that the repetition defect cannot be accounted for in terms of motor defect,
auditory perceptual impairment or more generalized language or memory disturb-
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

ances. However, the interpretation of the defect as one of auditory verbal short-
term memory (STM) cannot be conclusive on the basis of negative evidence.
The terms L T M and STM were used loosely in the previous paragraph as the
lack of agreement among theorists has resulted in there being no generally
accepted terminology. Indeed there are likely to be more systems than these two.
A strong case can be made for one or more relatively peripheral auditory stores
(Neisser, 1967; Crowder and Morton, 1969) and also for one or more visual stores
(Sperling, 1967; Posner, Boies, Eichelman and Taylor, 1969). Yet it is widely
agreed that auditory verbal span performance is primarily mediated by a single
system. This is the primary memory of Waugh and Norman (1965), the auditory
information storage of Sperling (1967) and the auditory-verbal-linguistic STM of
Atkiison and Shiffrin (1968). This we will call STM. The secondary memory of
Waugh and Norman (1965) we will call LTM.
This paper reports K.F.’s performance on five experiments. The aim of the
first experiments is to provide “positive” evidence that K.F. has a defect of STM.
An account is given of his performance on so-called “two-component” tasks,
where, according to present theoretical views on memory, retrieval from both
STM and L T M can be inferred from different aspects of the results. Theoretically
these tasks should identlfy the effects of circumscribed impairment of STM: free
recall and the Peterson procedure are the examples of this type of experiment.
Evidence on the nature of pro-active interference operating in K.F.’s performance
is also presented, as this provides another source of “positive” evidence for an
STM defect.
Given that K.F. suffers from an impairment of STM, it is important to determine
which stage in the processing of STM traces is defective and particularly whether
K.F. suffers only from a retrieval defect, since on various L T M tasks his perform-
ance is approximately normal (Warrington and Shallice, 1969). T o test this, his
performance on alternative methods of retrieval is contrasted with his performance
on recall. If his defect is one of retrieval by recall, his performance should be in
the normal range on these tasks. If his defect is not one of retrieval, then his
performance on these tasks should be comparable to his performance on recall
tasks.
 NEUROPSYCHOLOCY OF VERBAL MEMORY STORES 263
Experiment I-Free Recall
A widely used two-component task is that of immediate free recall on lists of
words longer than the span. For a normal subject the serial position curve resulting
from this task is thought to have two components (see Fig. I). These are the rising
portion of the curve over the last few serial positions-the recency effect-and the
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

"I 4 n-5 n
Serial position

FIGURE I. A Schematic serial position curve for a list of length n. Curve A is shown by the
solid line. Curve B is identical to curve A except for the last s serial positions where it is shown
by the dotted line. The shaded area can be used to obtain an estimate of S.T.M. capacity.

earlier part of the curve, which can itself be divided into a primacy effect and a
flat section. If the subject counts backwards for 30 sec. before recall, the recency
rise disappears (Postman and Phillips, 1965; Glanzer and Cunitz, 1966). This is
usually explained by assuming that, of the two components of curve A, one results
from retrieval from L T M and the other from retrieval from STM, while curve B
has only the LTM component.
Method
Strings of ten high frequency words each of 4 or 5 letters were read to K.F. at a rate of I
word/z sec. Immediate recall was tested. K.F. was instructed to report as many words as
he could and was given I min. for recall. In all, 30 strings were presented over three testing
sessions.
Results
The number of words correctly recalled (out of 30) at each serial position is
shown in Table I. K.F.'s curve is almost flat over the middle eight points, but no
simple statistical procedure is available to substantiate this qualitative statement.
However, by using Kendall's S statistic, no trend, significant at the 5 per cent level,

TABLE I
Free recall scores for each serial position

Serial Position 1 23 4 5 6 7 8 9 1 0
No. correct I5 107 7 7 10 12 9 9 25
Per cent correct 50 33 23 23 23 33 40 30 30 83
 264 T. SHALLICB AND E. K. WARRINGTON

can be found over any set of these eight points; but if either of the extreme points
is added, significant trends are found. This means that the recency effect is limited
to one serial position (10th) instead of the normal value of about five or six (Murd-
ock, 1962; Postman and Phillips, 1965), while the primacy effect is also limited to
one or possibly two serial positions instead of the normal value of about three or
four. The reduction in the recency effect clearly supports the hypothesis of a
reduction in STM capacity. The reduction in the primacy effect can be accounted
for by a reduced rehearsal capacity; this is compatible with Atkinson and ShiRrin’s
(1968) views about the effect.

Experiment II-the Peterson Procedure

The technique of presenting verbal items and then attempting to prevent the
subject from rehearsing for a number of seconds (Brown, 1959; Peterson and
Peterson, 1959) has become widely used. The simplest explanation of the decline
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

in recall over seconds is that it results from a decrease in the STM trace strength
(e.g. Waugh & Norman, 1965; Peterson, 1966; Baddeley, 1968). While this view
has been challenged (Keppel and Underwood, 1962; Melton, 1963), it has recently
received support from the work of Baddeley and Warrington (1970) on amnesics,
in which they find that amnesics who have a defective LTM system show the typical
Peterson decay curve.
Method
Modificationswere made to the normal Peterson procedure (Peterson and Peterson, 1959).
K.F. had performed better with visual input than auditory input (Warrington and Shallice,
1969); therefore to prevent K.F. from using his relatively well-preserved visual STM,
three three-letter words were spoken to him at a rate of x/sec. In addition, instead of
counting backwards, which he could not do efficiently, he was asked to count forwards in
ones from I as fast as he could. The end of each counting interval was signalled by the
experimenter saying “now” and a new trial began as soon as K.F. indicated he could recall
no more. Ten sets of trials were carried out. Each set consisted of five initial trials with a
10-sec. counting interval, followed by 16 trials in which four each of 0, 5, 10 and 15 sec.
counting intervale occurred in varying random orders. The five initial trials of each set
were discarded so that only the “steady state” results were analysed.
Results
Figure 2 gives the proportion of trials in which all three items were correctly
recalled and also the proportion of items correctly recalled. Normal Peterson-
type curves show a decline over an interval of 15 sec. from 95 per cent (with zero
delay) to under 25 per cent (at 15-sec. delay) in number of trials all correct, with
the curves still declining after 15 sec. (Murdock, 1961). When assessed by the
criterion of all items correct, K.F. clearly showed no comparable decline. It
should be noted that K.F’s failure to achieve nearly 100 per cent correct with zero
delay is consistent with his inability to repeat three words. In experiment I of
Warrington and Shallice (1969) he was able to repeat three-word strings correctly
on only 3 out of 20 trials.
With the items-correct measure there is a significant trend for longer intervals to
produce lower scores (Kendall’s S statistic P < 0.01). If the zero delay condition
is omitted, there is no longer a significant trend (Kendall’s S statistic 2 = 0.8).
 NBUROPSYCHOLOGY OF VERBAL MEMORY STORES 265

The slight trend that occurs indicates that K.F.'s STM capacity is much reduced,
though not quite zero; this is consistent with the interpretation of the first experi-
ment.
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

Counting interval (sac.)

FIGURE2. Percentage of item and trials correct as a function of the counting interval. 0 ,
Trials correct; 0, Itema correct.

Experiment II-Proactive Interference Effect

These two experiments support the view of Warrington and Shallice (1969) that
K.F. has an almost complete inability to recall from STM. Yet, that study showed
that K.F.'s performance on long-term memory tasks is normal, indicating that his
r e d from LTM is unaffected. This raises the possibility that, even in tasks
normally involving short-term memory, K.F. would show characteristics typical
of retrieval from LTM rather than from STM.
If a series of digit spans is presented to a normal subject, his performance typi-
cally shows a practice effect by improving over the series (Melton, 1963). By
contrast, in analogous tasks using LTM, performance deteriorates over the series
due to the build-up of P.I. (Underwood, 1957). Thus K.F.'s performance over a
series of span tasks is of interest, since it is possible that he would show an LTM-
like decline, instead of the normal improvement.
Method
Ten series, each containing ten pairs of letters, were spoken to K.F. The two letters of a
pair were separated by one second and there was a maximurn of 5 sec. in which to recall the
two letters before the next pair WBB presented. There was a I-min. gap between each series.
Results
A serial position curve can be obtained by summing scores for each serial
position, shown in Table 11. The declining trend over the 10 serial positions is
TABLE I1
Number of &@s correctly recalled for each p o s i t h in a set

Position in the set 1 3 3 4 5 6 7 8 9 1 0

No. items correct
(max. 20) 15 12 12 11 12 g 12 6 7 7
 266 T. SHALLICE AND E. K. WARRINGTON

significant at the o*ooz level (Kenda.Uk S statistic). One possible interpretation is

that this decline results from fatigue. However, this is highly implausible, as
K.F.’s performance had been virtually perfect in a demanding identification task
in which he had to categorize 40 words spoken at a rate of 11sec. (Warrington and
Shallice, 1969). A second possible explanation of why K.F. shows this abnormal
characteristic is that, in STM, P.I. operates only on the last two or three items
(Wickelgren, 1965) and this makes it relatively unimportant for digit spans of the
normal size but important for one of two items. However, in this case the serial
position curve should reach an asymptote after the first or second position, which
it does not. This finding indicates that, in digit span situations, a considerable
portion of K.F.’s retrieval is from LTM.
Experiment IV-Recognition using a Probe Technique
Given that K.F.’s ability to recall from his STM is drastically reduced, it be-
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

comes relevant to ascertain whether this is due to a retrieval failure. This is of

particular theoretical importance in view of his normal L T M performance.
Warrington and Shallice (1969) showed that K.F.’s ability to match two strings of
digits was much impaired; the probe technique allows this disability to be more
accurately studied. Alternative retrieval conditions were used in order to test not
only his recognition ability but also his rehearsal capacity and the effects of response
interference.
Method
The experimenter spoke a string of 5 letters to K.F. at a one per second rate, a tap marking
the end of the series. Three retrieval conditions were used. In condition A, the experimen-
ter presented a probe letter to K.F., who had to say whether or not it had appeared in the
string. In condition B, K.F. had to repeat the last letter of the string before being presented
with the probe letter. In condition C there was a 20-sec. interval between the end of the
string and the presentation of the probe letter. Sets of 8 strings were used, 4 in which the
probe letter had not occurred and 4 in which each of the first 4 serial positions were probed.
Before each set K.F. was told which retrieval condition was to be tested. Twenty-seven sets
of each of the three conditions were presented to K.F. in a randomized design.
Results
The number of “yes” and “no” responses made in each condition and position
are given in Table 111. The false positive rates differ significantlybetween the three
TABLE
I11
Number of yes and no respottses fm each condition and serial position

Condition Response Position of test letter in string Not

I 2 3 4 present
~~

A Yes I2 22 16 26 43
No 1.5 5 11 I 65
B Yes I8 9 I9 23 28
No 9 18 8 4 80
C Yes 13 I1 I5 20 24
No I4 16 I2 7 84
 NEUROPSYCHOLOGY OF VERBAL MEMORY STORES 267
conditions (x2 = 8.5 P < 0.02 at 2 d.f.). Consideration of the serial positions
reveals no consistent trends in the proportion correct over positions 1-3. Even if
the somewhat dubious procedure of summing the scores of all conditions for each
serial position is used, there is no significant difference between any of the three
positions: m a . M2, xt3, xi3) 7 = 1.6N.S. (I d.f.); d23= 1-9N.S. (2 df.).
Therefore scores for serial positions 1-3 were combined for each condition, for
subsequent analysis. In all three conditions there is a significant difference between
scores on serial position 4 and on the other 3 serial positions combined xi = I 1.6,
P < 0.001 ( I d.f.): 2 = 7.1, P < 0.01 ( I d.f.); x: = 5.5, P < 0.02 ( I d.f.).
In order to allow for the significant difference in false positive rate, it seems most
appropriate to transform the scores to a d' measure, (see Table IV). Three
conclusions can be drawn from these findings.
(i) The recency effect is limited to two positions (the fourth and fifth) unlike
that for normal subjects where either all positions or at least 6 positions,
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

depending on list length, show a recency effect with 3-figure digits

(Wickelgren and Norman, 1967). It is assumed that in condition A, the
item in the fifth position is retained even though not tested since, in
condition B where it is tested, a recency effect also occurs in position 4.
TABLE IV
Values of D for each condition

Condition Position 1-3 Position 4

(summed)

A 0.56 2'00
B 0.82 1-68
C 0.72 1.41

(ii) Since for condition C, in which there was a 20-sec. delay before the probe,
d'l-3 is significantly less than C4, K.F. appears to be able to rehearse
material within his short-term span. As no check was made on position
5, it is not possible to say whether he attempted to rehearse one or two
items.
(iii) For normal subjects, similar values of d' are obtained with recognition
and recall measures (Norman, 1966). If K.F. has a storage deficit a
similar finding would be expected for him. As no probe digit experiment
similar to that used by Waugh and Norman (1965)has been tried with
K.F., only a rough comparison can be made between recall and recogni-
tion measures of d'. The results of position 4 in condition A and of
position 2 in a 2-letter recall task (Warrington and Shallice, 1969)when
transformed, give similar values of d'. This comparison is speculative
because it is based on an assumption of the equal effects of stimulus and
response interference, and because the d' estimate obtained from condi-
tion A is based on only one false positive response. However, it does
reinforce the conclusion drawn from finding (i) that K.F.3 impairment
is not limited to retrieval by recall.
 268 T. SHALLICE AND E. K. WARRINGTON

Experiment V-Missing Scan

A technique which seems to test an alternative retrieval system-missing scan
(Bushke, 1963)-provides further information on whether K.F.’s difficulty is one
of retrieval.
Method
K.F.’s missing scan was tested for digit sets of size 3,4and 5. For example, for the set of
sue 3, two of the digits I, 2 and 3 were presented in either order and K.F. had to respond
with the digit of the set that was not presented. Two series of 10were presented for each of
the digit sets of size 3,4and 5. Digit strings of length 2,3 and 4,drawn randomly from sets
of size 3,4and 5, respectively,were also tested by recall. Each pair of series tested contained
one of recall and one of missing scan. The digits were spoken at a rate of I/sec. and X.F.
was allowed to respond at his own pace.

Results
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

The number of correct responses (out of 20) obtained for the missing scan and
the recall task for each digit set are shown in Table V. Relative to the maximum
possible scores K.F.’s performance varied between a 10and 34 per cent superiority

TABLE
V
Number of i t m recalled with missing scan and recall

Digit set sue 3 4 5

String length 2 3 4
Missing scan no. correct 15/20 12/20 9/20
Recall no. string correct 13/20 7/20 3/20

for missing scan. Bushke (1963), using various string lengths, has shown that
missing scan performance is more likely to be correct than that of ordinary digit
span. In his experiment the superiority of missing scan performance varied
between 25 and 45 per cent for comparable success rates. Thus K.F.’s missing
scan performance further supports the view that his defect is not limited to one
retrieval method.

Discussion
The results reported in this paper support the hypothesis suggested by Warring-
ton and Shallice (1969) that K.F.’s disability could be accounted for by a greatly
reduced STM capacity. In both the two-component tasks, the estimate of the
STM component obtained was close to zero. There was a great reduction in the
recency effect in free recall and a virtual lack of decline in performance in the
Peterson experiment. Furthermore, the marked and increasing effect of pro-active
interference over a series of span tasks is analogous to its effect on normal subjects
when LTM alone is used (Underwood, 1957). This contrasts with the pattern of
results produced by normal subjects in span experiments where improvement
occurs with practice (Melton, 1963), presumably because P.I. has much less effect
in STM.
 NEUROPSYCHOLOCY OF VERBAL MEMORY STORES 269

K.F.’s impairment on STM recall tasks can be explained by postulating a

defect either of a specific retrieval mechanism or of storage capacity. If his impair-
ment were one of retrieval by recall, tasks not requiring recall should not show a
comparable deficit, yet experiments 4 and 5 , using the probe and missing scan
techniques, give results far below normal and comparable with his performance on
recall. This finding would not be relevant if recognition and missing scan were
more complex than recall, that is, if they consisted of a recall operation followed
by a matching or inverting operation respectively. Introspectively this possibility
seems implausible for recognition, and rather more plausible for missing scan. If
recall were the least complex retrieval method, then its use in attempts to measure
STM capacity should presumably result in higher values for normal subjects than
those resulting from recognition or missing scan. As a rough indication, if the
d’ measure of STM is used for normal subjects, the value obtained from recall is
approximately equal to that obtained from both recognition (Norman, 1966) and
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

missing scan (Kaminska, personal communication). However, if the STM traces

are poorly laid down or lose strength more rapidly than normal, then all three
methods of retrieval should show roughly comparable poor performance, which is
the case with K.F.
K.F.’s performance on three L T M tests, the Wechsler P.A. task, a 10-word
learning task (Stevenson, 1968) and the Warrington and Weiskrantz (1968) in-
complete-word learning task, was shown to be approximately normal (Warrington
and Shallice, 1969). Thus his L T M system appears to be unaffected. The free
recall results reported here also support this view, but such an interpretation would
be less convincing if based on these alone. The Peterson results are not relevant
to this question (see below).
Postman and Phillips (1965) found that the amount retrieved in free recall after
counting backwards for 30 sec. is the same as that retrieved after immediate recall
for the whole serial position curve except the last “recency effect” section (see
Fig. I); this amount can therefore be used to estimate L T M retrieval capacity.
The recency effect can be used to provide an approximate estimate of STM
retrieval capacity. [If one assumes retrieval from STM and L T M to operate
independently, greater accuracy is obtained by the use of a probability theory
correction (Waugh and Norman, 1965).] I n the free recall experiment the middle
8 points can be considered as the “flat” part of K.F.’s serial position curve. Values
of 3.2 and 0.5 respectively, were derived for K.F.’s L T M and STM capacity.
Most comparable data for normal subjects is based on student subjects. However,
Baddeley and Warrington (1970), using 6 hospital control subjects, have obtained
values of 3-55 (s.D.0.61) and 1-65 (s.D.0.71)for L T M and STM respectively with
a 3 sec. (i.e. longer) presentation rate.
One common interpretation of Peterson curves is that they arise from a declining
STM and a constant L T M factor (Waugh and Norman, 1965). As no significant
trend is found in K.F.’s performance on the 5-, 10- and 15-sec. intervals, these
results could be used to estimate the asymptotic level of his Peterson curve and thus
his L T M capacity. An 0-12probability of recalling 3 words correctly is the
estimate obtained. However, it is doubtful if this method of estimating L T M cap-
acity is valid. Whether normal subject’s curves are asymptotic is unclear; Baddeley
 270 T. SHALLICE AND E. I(. WARRINGTON

and Warrington (1970) showed that the decline continues at least up to 60 sec.
Also they found that even for the 60-sec. interval normal subjects produce the
same results as amnesics, who have a defective LTM. Even if this procedure were
a valid means of estimating L T M capacity, it would be diffcult to use the estimate
for comparison purposes as the inter-subject variance at long intervals is very large
(see Murdock, 1961). In addition, the results at any interval are markedly affected
by a number of factors, including inter-trial interval (Loess and Waugh, 1967),
coding possibilities and skill.
The findings from these experiments indicate that K.F. has a defective STM
system while his performance on L T M tasks is normal. By contrast, in the classic
amnesic syndrome, the STM system has been shown to be normal with the use
of very similar techniques (Wickelgren, 1968; Baddeley and Warrington, .1970).
This provides strong evidence for a double dissociation of function; that is, there
are at least two verbal memory stores, the STM and the LTM, which are able
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

to operate independently.
Auditory

inpull C
I I 1

I t
store

Speech
output
control
L Speech output

Semantic
analysis store

FIGURE 3. Suggested model for processing in auditory verbal memory experiments. The output
connections from the L.T.M.unit are left unspecified, since these experiments provide no relevant
evidence. Rehearsal is equivalent to the operation of circuit ABC, “the echo box”.
More important, the results present difficulties for those theories in which
STM and L T M are thought to use the same physical structures in different ways
(e.g. Hebb, 1961; Norman, 1968). They also indicate that the frequently used
flow-diagrams (Waugh and Norman, 1965 ;Atkinson and SMrin, 1968; Murdock,
1967; Bower, 1967) in which information must enter STM before reaching LTM,
may be inappropriate. On this model, if the STM system were greatly impaired,
one would expect impairment on L T M tasks, since the input to the L T M store
would be reduced. To counter this argument it could be assumed, in line with
one of Atkinson and Shiffrin’s models, that the rate of transfer from S T M to
L T M is inversely proportional to the capacity of the STM. Since K.F.’s STM
capacity is very low, the rate of transfer would be correspondingly high, which
would explain his normal long-term memory. This is, however, a completely
ad hoc explanation, and it is rather implausible to assume that one function of
the STM system should remain totally unaffected, whilst the other function is
seriously impaired.
There are two further arguments against this hypothesis. Firstly, the hypothesis
 NEUROPSYCHOLOCY OF VERBAL MEMORY STORES 271

provides no explanation of why STM should be subject to acoustic and L T M to

semantic interference (Baddeley, 1966). Secondly, one common way of making
more plausible the ad hoc assumption about rate of transfer is to assume that
the LTM trace is laid down by means of rehearsal. This cannot be a complete
explanation of the formation of L T M traces since, with techniques preventing
rehearsal, an L T M component has been demonstrated (Murray, 1968).
In the light of these findings, it is suggested that a model in which the inputs
to STM and L T M are in parallel rather than in series should be considered. The
flow diagram shown in Figure 3 is proposed. As a theory of the relation between
short-term and long-term memory, it has certain similarities to those theories
suggested by Broadbent (1958)and Wickelgren (1968),though neither author is
so explicit. Earlier Buhler (1908),on the basis of experiments on memory for
proverbs, suggested that semantic and “sensory” memory for verbal material are
mediated by independent processes. According to this model K.F. would have
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

sustained damage to the acoustic STM store or to the pathways linking the
phonemic analysis unit and short-term storage. Morton’s (1969)model, which is
also compatible with the results, differs by requiring the input to the STM
system to be dependent on modality-free word recognition.
The idea that semantic analysis is not necessary for STM storage is supported
by the observation of Geschwind (1965)that echolalia-the echoing of the doctor’s
speech by a patient-occurs in patients who appear not to understand the doctor.
More important, in contrast to the serial model, this model has the obvious
advantage of accounting for the difference in type of error found in STM and
LTM, namely, acoustic and semantic, respectively. Moreover, the disparity
between the anatomical correlates of the S TM and L T M (Warrington and Shallice,
1969)increases the likelihood of their operating in totally different ways.
On this model, the evolutionary function of STM is not that of a unit which
holds information while traces consolidate in LTM. In order to understand its
operation more fully, some assessment of why the human organism should contain
such a store seems appropriate. One possible function of the store is to assist in
the understanding of complex sentences, since the use of the store together with
rehearsal enables a person to attempt more than once to understand a sentence.
Another possibility is that the store is an integral part of the speech-production
system, possibly corresponding to the phonemic-representation level in Wickel-
gren’s (1969) theory of articulation. These possibilities will be discussed in a
later paper.
We are grateful to Dr. Denis Williams for permission to report our findings on this case.
We wish to thank Dr. David Legge for his helpful comments on the manuscript. Our
thanks are particularly due to Dr. R. T. C. Pratt for providing facilities which made this
investigation possible and for his continued encouragement. The trustees of H. J. Shorvon
Memorial Fund kindly gave a grant towards expenses.

References
ATKINSON, M. (1968). Human memory: a proposed system and its
R. C. and SHIFFRIN,
control processes. In SPFNCE,
K. W. and SPENCE, J. T. (Eds.), The Psychology of
Learning and Motivation: Advances in Research and Theory, Vol. 2. New York:
Academic Press.
 272 T. SHALLICE AND E. K. WARRINGTON

BADDELEY, A. D. (1966). The influence of acoustic and semantic similarity on long-term

memory for word sequences. Q. Jl exp. Psychol. 18,302-9.
BADDELEY, A. D. (1968). How does acoustic similarity influence short-term memory?
Q. Jl exp. Psychol. 20, 249-64.
BADDELEY,A. D. and WARRINGTON, E. K. (1970). Amnesia and the distinction between long
and short-term memory. J. verb. Learn. verb. Behav. (in press).
BOWER, G. (1967). A multicomponent theory of the memory trace. In SPENCE,K. W.,
SPWCB, J. T. a n d h ~ m oN.~ H., (Eds.), ThePsychologyof Learning and Motivation:
Advances in Research and Theory, Vol. I. New York: Academic Press.
BROADBENT, D. E. (1958). Perception and Communication. London: Pergamon Press.
BROWN,J. (1959). Information, redundancy and decay of the memory trace. In The
Mechanisation of Thought Processes. London : H.M.S.O.
BUHLER, K. (1908). Tatsachen und Probleme zu einer Psychologie der Denkvorgkge I11
uber Gedankenerinnerungen. Archiv. ges. Psychol. 12,24-92.
BUSHKE,H. (1963). Retention in immediate memory estimated without retrieval. Sc’ience,
N.Y.140, 5 6 1 .
CROWDER, R. G. and MORTON, J. (1969). Precategorical acoustic storage (PAS) Percep.
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

Psychophys. 5, 36573.
GESCHWIND, N. (1965). Disconnection syndromes in animals and man. Pt. 11, Brain,
88, 585-644.
GLANZER, M. and CUNITZ,A. R. (1966). Two storage mechanisms in free recall. J. verb.
Learn. verb. Behav. 5, 351-60.
GREGORY, R. L. (1961). The brain as an engineering problem. In THORPE,W. H. and
ZANGWILL, 0. L. (Eds.). Current Problems in Animal Behaviour. Cambridge:
Cambridge University Press.
HEBB,D. 0. (1961). Distinctive features of learning in the higher animal. In DELAFIUB-
NAYE, J. F. (Ed.). Brain Mechanisms and Learning. London: Oxford University Press.
KBPPEL, G. and UNDERWOOD, B. J. (1962). Pro-active inhibition in short-term retention of
single items. J. verb. Learn. verb. B e h . I, 153-61.
LOESS,H. and WAUGH, N. C. (1967). Short-term memory and intertrial interval. J. verb.
Learn. verb. Behav. 6,455-60.
MELTON, A. W. (1963). Implications of short-term memory for a general theory of memory.
J. verb. Learn. verb. Behav. 2, 1-21.
MILNER,B. (1966). Amnesia following operation on the temporal lobes. In WHITTY,
C. W. M. and ZANGWILL, 0. L. (Eds.), Amnesia. London: Butterworth.
MILSUM,J. H. (1966). Biological Control Systems Analysis. New York: McGraw-Hill.
MORTON, J. (1969). The interaction of information in word recognition. Psychol. Rev.
76, 165-178.
MURDOCK, B. B. (1961). The retention of individualitems. J. exp. Psychol. 62,618-25.
MURDOCK, B. B. (1962). The serial position effect of free recall. J. exp. Psychol.64,482-8.
MURDOCK, B. B. (1967). Recent developments in short-term memory. Br. J. Psychol.
58,421-33.
MURRAY,D. J. (1968). Articulation and acoustic confusability in short-term memory.
J. exp. Psychol. 78,679-84.
NEISSER,U. (1967). Cognitive Psychology. New York: Appleton Century Crofts.
NORMAN,D. A. (1966). Acquisition and retention in short-term memory. J. exp. Psychol.
72,36941.
NORMAN,D. A. (1968). Toward a theory of memory and attention. Psychol. Rev. 75,
522-36.
PETERSON, L. R. (1966). Short-term verbal memory and learning. Psychol. Rev. 73,
193-207.
-ON, L. R. and PETERSON, M. J. (1959). Short-term retention of individual verbal
items. J. exp. Psychol. 58, 193-8.
P o s m , M. I., BOIES,S. J., EICHELMAN, W. H. and TAYLOR, R. L. (1969). Retention of
visual and name codes of single letters. J. exp. Psychol. Monogr. 79,No. I, Pt. 2.
 NEUROPSYCHOLOGY OF VERBAL MEMORY STORES 273
POSTMAN, L. and PHILLIPS,L. W. (1965). Short-term temporal changes in free recall.
Q. Jl exp. Psychol. 17, 132-8.
SPERLINC, G. (1967). Successive approximations to a model for short-term memory. In
SANDERS, A. F. (Ed.), Attention and Performance. Amsterdam: North Holland.
STEVENSON,J. (1968). Some psychological effects of frontal lobe lesions in man. M.Sc.
Thesis (unpublished). University of Cambridge.
UNDERWOOD, B. J. (1957). Interference and forgetting. Psychol. Rew. 64, 49-60.
WARRINGTON, E. K. and SHALLICE, T. (1969). The selective impairment of auditory verbal
short-term memory. Brain, 9, 885-96.
WARRINGTON, E.K.and WEISKFUNTZ, L. (1968). New methods of testing long-term reten-
tion with specific reference to amnesic patients. Nature, Lond. 217, 972-4.
WAUGH,N. C. and NORMAN, D. A. (1965). Primary memory. Psychol. Rev. 72, 89-104.
WEISKRANTZ, L. (1968). Treatments, inferences and brain function. In WEISKRANTZ, L.
(Ed.), Analysis of Behauiourul Change. New York: Harper and Row.
WICKELGREN, W. A. (1965). Phonemic similarity and interference in'short-term memory for
single letters. J. exp. Psychol. 7 1 , 396-404.
WICKELGREN, W. A.(1968). Sparing of short -term memory in an amnesic patient: implica-
Downloaded By: [University of Cambridge] At: 11:35 17 February 2010

tions for strength theory of memory. Neuropsychologia, 6, 235-4.

WICKELGREN, W. A. (1969). Context-sensitive coding, associative memory, and serial order
in (speech) behaviour. Psychol. Rev. 76, 1-15.
WICKELCREN, W. A. and NORMAN,D. A. (1967). Strength models and serial position in
short-term recognition memory. J. math. Psychol. 3, 316-47.
Received July 19%

View publication stats