Neuropsychologia, 1968, Vol. 6, pp. 235 to 244. Pergamon Press. Printed in England
SPARING OF SHORT-TERM MEMORY
IN AN AMNESIC PATIENT: IMPLICATIONS
FOR STRENGTH THEORY OF MEMORY*
WAYNE A. WICKELGREN
Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, U.S.A.
(Received 13 September1967)
Abstract--Short-term recognition memory for single-digitnumbers, three-digit numbers, and
the pitch of pure tones was studied in a subject (H.M.) who appears to possess normal shortterm
memory, but to be almost completely lacking in the ability to form new long-term
memory traces. The primary purpose was to test the short-term memory component of a dualtrace
strength theory developed for normal subjects. To a first approximation, H.M.'s tracestrength
decay curves are exponential, as postulated by strength theory for short-term memory.
Furthermore, the rate of decay of this trace for H.M. is well within the range of normal subjects.
These findings agree with previous findings on H.M. in support of a dual-trace theory of
memory. More particularly, the present findings support the strength theory of memory
proposed by WICKELGRENand NORMAN[15].
As A CONSEQUENCEof bilateral removal of the mesial parts of the temporal lobes, H.M.
has an almost complete inability to form new long-term memory traces for both verbal
and non-verbal stimuli. What is perhaps most remarkable about H.M.'s deficit is its
purity. On the basis of present evidence, it appears that H.M. has no obvious personality
change, no deficit in performance tests that involve only previously acquired knowledge
(such as a standard intelligence test), very little remaining retrograde amnesia, no deficit
in long-term motor-skill learning, and no deficit in immediate memory span. Naturally,
it is still possible that some more sophisticated test would show a deficit in one of these areas
where previous tests have not, but there is considerable support for the assertion that H.M.'s
primary deficit is in the formation of new long-term memory traces for stimuli. For a
complete review of findings concerning H.M., see MILNER [-5] and other papers in this issue.
H.M.'s apparently normal short-term memory, accompanied by virtually complete
inability to form new long-term memory traces, provides a remarkable opportunity to
test theories of short-term memory under conditions where one can be quite sure that
little or no long-term memory is contaminating the results. The present paper tests whether
H.M.'s performance in certain verbal and non-verbal short-term recognition memory
tasks can be adequately described by the short-term memory component of a particular
theory, namely, strength theory [-12, 14, 15].
The major assumptions of strength theory as applied to short-term recognition memory
are as follows: (a) The short-term memory trace for an item can be described by a single
* This work was supported primarily by grant, MH 00890-03, from the National Institute of Mental
Health, U.S. Public Health Service. Further aid was received from a National Aeronautics and Space
Administration grant NsG 496, to Hans-Lukas Teuber.
235
236 WAYNE A. WICKELGREN
real-valued random variable (the "strength" of the item); (b) The mean of this random
variable is incremented by presentation of the item and decays (exponentially) as a function
of the duration of the filled (rehearsal-preventing) delay interpolated between presentation
and the test of the item; (c) The decision rule used to determine whether or not the test
item has been presented previously is the criterion decision rule of signal-detection theory
(decide "yes", if and only if the strength of the test item in memory exceeds a criterion).
There are two subsidiary assumptions which also seem to be approximately valid for
normal subjects, namely, that the noise in trace strength is normally distributed, and that
the variance is independent ofthe mean. These subsidiary assumptions simplify computation,
but are in no sense crucial for strength theory.
Strength theory has previously provided a rather good description of the performance
of normal subjects in the same short-term memory tasks to be used in the present study
with H.M., namely, recognition memory for three-digit numbers [15] and the pitch of
pure tones [14]. However, in the latter case, but not the former case, it was necessary, for
normal subjects, to assume some sort of intermediate-term memory component which
summed with the short-term memory component to produce the total memory trace.
If strength theory is correct, then it should not be necessary to assume an intermediateterm
component of tile memory trace in either case for H.M., and H.M.'s data for both
numbers and tones should be well fit by a single exponentially decaying short-term memory
trace with approximately the same acquisition and decay parameters as normal subjects.
On the other hand, if the good fit of strength theory for normal subjects was due to a
fortuitous interaction of short-term and intermediate-term or long-term memory in a
manner quite different from that postulated by strength theory, we might expect a subject
without intermediate-term or long-term memory to show marked deviation from the
single exponentially decaying trace postulated by strength theory for short-term memory.
At the same time, if H.M.'s data show close agreement with the short-term memory
component of strength theory with acquisition and decay parameters within the range
for normal subjects, and if his data do not require an intermediate-term or long-term
component, then the present study will provide further support for the hypothesis that
H.M. has a completely normal short-term memory accompanied by an inability to form
new long-term memory traces.
METHOD
All the experiments reported in the present paper use the method of "yes-no" recognition
memory for single items. The items in different experiments are single-digit numbers,
three-digit numbers, or pure tones, and the retention intervals vary from 0.25 sec to 8 sec.
The general procedure is to present the item or list of items to be remembered and then
after a delay to present a single test item about which H.M. must decide "yes" or "no"
("same" or "different") as to whether the test item had been presented previously on that
trial.
Exp. 1: Single digits
Procedure. On each trial H.M. got a ready signal, followed in about 1 sec by auditory
presentation of a list of 8 single-digit numbers presented at a rate of 3 digits per sec, followed
immediately by a single test digit, followed by H.M.'s decision as to whether the test (last)
digit had occurred in the previous list of 8 digits. H.M. usually gave his answer within
2 sec after the test letter and the next trial began about 3 see after this. In all the experiments
SPARING OF SHORT-TERM MEMORY IN AN AMNESIC PATIENT 237
reported in this paper, the trials themselves were recorded on tape, but the experimenter
started and stopped the tape recorder between trials to record H.M.'s responses.
H.M. had a card in front of him throughout the experiment which reminded him of
the decision he was to make. The instruction on the card was as follows: "Did the last
number appear earlier in the list?" From time to time the experimenter repeated this
instruction aloud to H.M. and indicated that he should keep looking at the card to remind
himself of the task. We must keep in mind that on part or all of some trials H.M. may not
have remembered what he was supposed to be doing, but as far as one can tell from his
speed and alertness in performing the task, this was not a major problem.
Design. There were 11 different conditions in the experiment. In 9 of the conditions
no digit was repeated in the list of 8 digits and the test digit was identical to the digit in
the first, second, third ..... eighth position in the list or else was not identical to any digit
in the list. The last condition occurred 3 times as often as the other conditions. In the other
two conditions, the second and fifth digits in the list were identical and the test item was
either that digit or a digit that did not occur in the list at all. Conditions were presented
in random order in blocks of 13 trials. There were 21 different examples of each condition,
some of which were presented twice and some three times to make a total of about 47
trials in each condition, except the false recognition condition after lists with no repeated
items, which had about 141 trials. The N for each condition is actually slightly smaller
than this because about 2 per cent of the trials were discarded due to failure of H.M. to
respond.
Exp. 2: Three-digit numbers
Procedure. On each trial H.M. got a ready signal followed by a list of 3-digit numbers
which was presented at the rate of one 3-digit number per sec, followed immediately by a
test 3-digit number, followed by H.M.'s decision as to whether the test (last) 3-digit number
had occurred earlier in the list. As in the previous experiment the subject had a card in
front of him that reminded him of the instructions as follows: "Did the last 3-digit number
appear earlier in the list ?"
Design. Lists of five and seven 3-digit numbers were presented (in different sessions)
and every serial position of each list length was tested approximately equally often (N
of about 40 for each position in length-5 lists and N of about 50 for each position in length-7
lists). New 3-digit numbers (not present in the preceding list) were tested 99 times for
length-5 lists and 186 times for length-7 lists. The same lists were used as in WICKELGREN
and NORMAN [15], and that article should be consulted for details of list construction and
randomization of conditions.
Exp. 3: Pure tones
Procedure. On each trial H.M. got a ready signal followed by a standard (S) tone
lasting 2 sec, followed by an interference (I) tone lasting a variable t~ sec, followed by a
comparison (C) tone lasting 1 see, followed by H.M.'s decision as to whether the S and C
tones were the same or different in pitch. H.M. always had a card in front of him that
said, "Did thefirst tone and the last tone sound the same or different in pitch?"
Design. There were 7 values of ti, 0.25, 0.5, 0.75, 1, 2, 4, and 8 see, and 3 values of
frequency difference between the S and C tones, - 10 eps, 0 cps (appeared twice as often
238 WAYNEA. WICKELGREN
as the others), and 10 cps. The S tones were randomly assigned frequencies between 400
and 490 cps in 10 cps steps. The I tone was 930 cps. Conditions were randomized in blocks
of 7 ×4=28 trials, with 18 different blocks. Each block was tested about twice giving an
N of about 72 for 0 cps conditions and 36 each for 10 and -10 cps conditions.
RESULTS
Responseprobabilities
The probabilities of correct and false recognition for items from each serial position
(k) in lists of 8 single-digit numbers and lists of 5 and 7 three-digit numbers are shown in
Table 1. Conditions where the test item was not in the prior list are indicated by a *
Table 1. Correct and false recognition probabilities in H.M.'s short-term
recognition memory for verbal stimuli
Exp. 1 (Single digits) Exp. 2 (Digit triples)
L5 L7
k P(yes) k P(yes) k P(yes)
* 0.14 * 0.04 * 0.12
1 0.91 1 0.44 1 0.29
2 0.87 2 0.63 2 0.21
3 0.85 3 0.84 3 0.31
4 0.85 4 0.88 4 0.50
5 0.92 5 1.00 5 0.64
6 0.94 6 0.83
7 1.00 7 1.00
8 1.00
"2-1-5 0.09
2+5 1.00
The correct and false recognition conditions in the lists of 8 single-digit numbers where
the second and fifth items were identical are indicated by 2, 5 and *2, 5, respectively.
The results indicate a general decline in frequency of correct recognition of a test item
as the number of items intervening between presentation and test increases, but there
appears to be some primacy effect as well, in the single-digit lists and the length-7, threedigit
number lists.
The probabilities of correct and false recognition of comparison tones in the delayed
comparison task are shown in Table 2. Since only one item, the S tone, is to be remembered
and the different delay intervals ti, could and undoubtedly did produce different biases
to respond "same", it is necessary to assess both correct and false recognition rates for
each h. On the whole, Table 2 indicates a deterioration in the accuracy of distinguishing
SPARING OF SHORT-TERM MEMORY 1N AN AMNESIC PATIENT 239
C tones that are the same as or different from the S tone, with increasing (filled) delay
between the two.
Table 2. Correct and false recognition probabilities in H.M.'s short-term
recognition memory for pitch
Exp. 3 (Tones)
tl P(Same IS--C=0) P (Same IS-C= + 10)
0.25 0.69 0.15
0.50 0.72 0.17
0.75 0.78 0.19
1.0 0.85 0.32
2.0 0.88 0.39
4.0 0.60 0.44
8.0 0.32 0.31
Strength theory
The correct and false recognition probabilities shown in Tables 1 and 2 can be transformed
into measures of the difference between correct and false recognition conditions
in average strength of the memory trace for the test items. This transformation from pairs
of response probabilities into average strength-difference measures uses the criterion rule
of signal detection theory, which can be applied in a very direct and plausible manner to
recognition memory [1, 7, 15]. The essence of the theory is that correct and incorrect
test items can be characterized by a unidimensional measure of strength in memory, the
value of which, under constant experimental conditions, is approximately normally distributed
with a mean which is higher for correct than for incorrect test items.
In general, it is not necessary to assume that the trace strength distributions have equal
variance, but for the types of tasks reported in this paper, variances have always been
approximately equal. It was not possible to obtain confidence judgments from H.M. in
the present experiments. Thus, it was not possible to test the equal variance assumption
for H.M., and we must assume that the variances of correct and incorrect strength distributions
are approximately equal for H.M., just as they are for normal subjects. For the
same reason, we must assume without direct proof that the strength distributions for H.M.
are approximately normal, just as they are for normal subjects. Under these assumptions,
one can obtain measures of the difference in mean strength for correct test items vs. incorrect
test items. This strength difference is the d' value obtained by looking up the correct and
false recognition probabilities in the tables of d' of ELUOTT [2]. SO much for the decision
assumptions of strength theory.
The short-term memory acquisition-assumption is essentially that presentation of an
item boosts its mean strength by a certain amount and boosts the mean strength of an
equivalence class of similar items by a lesser amount. Let the difference between these two
initial acquisition strengths be e (where the unit of measurement is the standard deviation
240 WAYNE A. WICKELGREN
of the trace strength distributions). The acquisition parameter (~) represents the initial
discriminability of correct vs. one class of incorrect test items, at the time of presenting the
correct item.
The short-term memory decay assumption is that this difference in strength (discriminability)
of the correct and incorrect test items decays exponentially as a function of the
time delay interpolated between presentation and test, provided the time delay is filled
with intervening activity to prevent active rehearsal.
The prediction of strength theory for H.M.'s data is that the discriminability of correct
vs. incorrect test items (d'), in any task, will decay exponentially as a function of delay (t~).
Thus, d'(t~)= ae-Ptj, where the decay rate (/3) would be expected to remain constant over
different tasks, but the acquisition constant (~) might vary considerably depending on the
similarity of the correct vs. incorrect test items. For example, if one used incorrect C tones
20 cps from the S tone, one would certainly expect to find a larger value of the acquisition
parameter (~), but one would not expect the decay rate (fl) to change.
Goodness-of-fit of strength theory
According to the strength theory just described, log [d'(ti)] is a linear function of tL
with a y-intercept of ~ and a slope of -/3. The results of such semi-logarithmic plots of
the empirically obtained d'(h ) values for single digits, length-5 digit triples, length-7 digit
triples, and pure tones are shown in Fig. 1.
4.0L~ I
3-°L"-,,.. o
2.0
i • SINGLE DIGITS
~ L5 DIGIT TRIPLES
02 " u L7 DIGIT TRIPLES
x TONES
i
I 2 :3 4 5 6 7 8
t I = DELAY (sec)
FIQ. l.
In all four cases, d' has a marked tendency to decline with increasing delay. Furthermore,
the rate of decay (/3) seems to be approximately the same in all four cases. However,
the acquisition parameter (~) is different in each case. The theoretical lines shown in Fig. 1
are for fl= 0.29 and ~ = 3.3, 4.5, 2.8, and 2.0 for single digits, length-5 digit triples, length-7
triples, and pure tones, respectively.
SPARING OF SHORT-TERM MEMORY IN AN AMNESIC PATIENT 241
There are some interesting deviations from the theoretical lines shown in Fig. 1, namely,
a marked primacy effect for single digits, a slight primacy effect for length-7 digit triples,
and a peculiar fiat portion of the decay curve for pitch in the region of tt less than 2 sec.
If one estimates ~ and fl for length-7 digit triples and uses the false recognition rate
to estimate the criterion for a "yes" response, one can test the goodness-of-fit of the strength
theory (in predicting the correct recognition rate) by a chi square test on 7 - 2 = 5 d.f. For
the length-7 digit triples, 7~2=5.93 on 5 d.f., p>0.30, indicating an extremely good fit
of the theory to the data.
Now let us take the value of the short-term memory decay parameter (fl) estimated
from the data for length-7 digit-triples and use it along with separate estimates of the
acquisition and criterion parameters obtained from each other set of data, to obtain predictions
of the correct recognition rates for each of the conditions of the other experiments.
The chi square values for the fit of the theory to each of the other experiments were: X2=
11.65 on 7 d.f., p > 0.10, for single digits, X2= 6.56 on 4 d.f., p > 0.10 for length-5 digit triples,
and X2= 12.15 on 6 d.f., p>0.05 for pure tones. The goodness-of-fit tests indicate a reasonably
good fit in every case.
The acquisition parameters for H.M. are somewhat lower than those obtained in
the previous experiments using normal M.|.T. and Harvard students. This is hardly
surprising, and of no theoretical significance, especially since his immediate memory span
was somewhat below the memory span for normal M.I.T. and Harvard students, before
his operation, and was not changed by the operation. The acquisition parameter is subject
to differences in coding strategy, ability to maintain concentration, and the degree to which
the subject follows instructions. Any or all of these factors, though particularly the last
one, might account for H.M.'s slightly lower degree of original learning of each item.
In any event, the nature of the memory trace is indicated not by the acquisition parameter,
but by the form of the decay function and its decay rate parameter(s). H.M.'s decay
functions appear to be reasonably well fit by a single exponentially decaying trace, with a
decay rate completely within the range of decay rates for the short-term memory traces
of normal subjects [15, 14].
The short-term memory component of strength theory provides a reasonably good
description of the performance of a subject who has virtually no ability to form long-term
memory traces. Thus, it cannot be argued that the good fit of strength theory for short-term
memory in normal subjects is due to a complex interaction of short-term and long-term
memory because a good fit is obtained for a subject who presumably has only short-term
memory.
Certainly, no evidence was found in the present study to indicate the necessity of estimating
any intermediate-term or long-term memory traces to fit the data of H.M. in any
of the present experiments. This is particularly interesting since it has proved necessary
to estimate an intermediate-term memory component to fit the pitch memory data in normal
subjects(WICKELGREN[13, 14]).
The primacy effects observed in two of three cases for H.M. are not surprising, since
primacy effects in short-term memory have also been observed for normal subjects [15].
More disturbing is the flat portion of the decay curve for pitch memory in the region
from 0.25 to 2 sec for H.M. This could be due to rehearsal by H.M. during the interference
tone, especially since it was impossible to instruct H.M. not to rehearse the standard tone
during the interference tone. However, an attempt to get similar results from one normal
subject who was instructed to rehearse the standard tone during the interference tone
242 WAYNEA. WICKELGREN
failed to show a comparably fiat decay curve in the region of 0.25 to 2 sec, though there
was a definite tendency in this direction. Perhaps H.M. is much more skilled at rehearsing
in the presence of distraction. This interpretation of the flat portion in H.M.'s decay curve
for tones must be considered completely tentative.
Probably the most disturbing feature of H.M.'s data is the vastly different degree of
learning, ~, for each triple in length-5 as opposed to length-7 lists. No such difference
was observed with normal subjects [-15]. One expects differences in degree of learning for
different types of items, but one does not, in general, expect differences for different list
lengths.
The most plausible explanation, in my opinion, is that H.M. had a greater tendency
to forget what he was supposed to be doing in the length-7 experiment than in the length-5
experiment. During acquisition, this could lead to a lower average degree of learning
(boosting of strength) for presented items. During retrieval, this could lead to an increase
in retrieval noise. Hence there would be a decrease in d'(ti) for either or both of two
reasons. First, the mean strength of old items is reduced, and/or second, the standard
deviation of the retrieval noise (which is the unit of measurement of strength) is increased.
The latter effect would be reflected entirely in a reduced value of the acquisition parameter
(~) with no effect on the decay parameter (/~) or the exponential shape of the decay
function. The former effect is more difficult to analyze, requiring us to assume that acquisition
(~) is a random variable, rather than a real variable. Any of the more plausible
theories of the probabilistic aspects of H.M.'s acquisition will predict a reduction in the
acquisition parameter. Unfortunately, they will also predict changes in the shape of the
decay function, depending on the amount of acquisition noise. This being the case, the
data suggest that the largest difference between length-5 and length-7 lists was an increase
in retrieval noise, with the level of acquisition noise being rather small in either ease.
DISCUSSION
The overall fit of strength theory to H.M.'s datais quite satisfactory. That this good fitwas
achieved using the same decay rate in all four eases is particularly convincing. Nevertheless,
there were some deviations from the theory that might have proven systematic in a larger
experiment. Thus, the support of strength theory in the present study must be considered
tentative.
However, what is overwhelmingly supported by the present study, in agreement with
all previous studies of H.M. and subjects with similar deficits [5], is that there are at least
two different memory traces in normal subjects, a short-term trace and a long-term trace.
In view of the fact that H.M.'s short-term memory seems to be normal, while his ability
to form new long-term memory traces is virtually completely absent, it seems rather unlikely
that short-term and long-term memory are based on a single trace with different
degrees of acquisition.
Of course, all that is definitely shown is that H.M. is lacking some vital component
in the consolidation process for long-term memory traces. It is logically possible that this
consolidation process simply boosts to a higher level the same trace established by external
stimulation, and that no amount of external stimulation can boost the trace to this high
level without the operation of the internal consolidation system. H.M.'s deficit is most
simply explained by assuming that a different trace is established by the consolidation
system. Nevertheless, it is possible to imagine a rather complex system in which the level
of a single trace is a complicated function of external stimulation up to some level and of
SPARING OF SHORT-TERM MEMORY IN AN AMNESIC PATIENT 243
the internal consolidation process beyond that level. However, a variable-acquisition singletrace
theory is contradicted by the vast difference in rate of decay observed in normal
subjects for traces with the same strength, but established in short-term vs. long-term
memory experiments.
MELTON [4] has suggested that we view memory as consisting of a single trace, whose
rate of decay is a variable that is a function of the number of repetitions, the nature of the
material being learned, the nature of the prior and subsequent material, etc. This variabledecay
single-trace theory is much harder to distinguish from a dual-trace theory than is a
variable-acquisition single-trace theory. However, H.M.'s data force the variable-decay
theory to assume that the reduction in decay rate with multiple presentation depends on
the operation of an internal consolidation process, while the initial establishment of the
rapidly decaying trace depends on a different internal consolidation process. Thus, either
type of single trace theory is required by H.M.'s data to postulate a dual consolidation
process.
But what of the evidence that short-term memory has functional properties very similar
to long-term memory [-3, 4, 6, 8-11, 13, 16]. The results of all of these studies can be
summarized as indicating that short-term memory is associative, just as is long-term
memory.* In verbal short-term memory at rapid rates of presentation, structural phonemic
coding is much more important than any coding by meaning, whereas when subjects have
more time to study the material to be remembered, they recode the material in a variety
of other ways. Nevertheless, the storage is by the strengthening of associations between
internal representatives in either case, only the nature of the internal representatives is
different. Furthermore, although there is greater replication in different subsystems of
the nervous system of the long-term traces than of the short-term traces, it is quite likely
that all subsystems can have both short-term and long-term memory traces formed between
their internal representatives.
Thus, all the evidence on the relationship between short-term and long-term memory
can be summarized in the following manner. Short-term memory is mediated by a rapidly
decaying facilitation of the association between (or within) internal representatives; longterm
memory is mediated by a much more slowly decaying facilitation of the association
between (or within) internal representatives. There are simply two different traces that can
facilitate the same connections between internal representatives.
* When one is studying item recognition memory, as in the experiments reported in the present paper,
one conceives of the facilitated associations to be those between the component parts of the internal representative
of an item.
REFERENCES
1. EGAN,J. P. Recognition memory and the operating characteristic. Indiana University Hearing and
Communication Laboratory. AFCRC-TN-58-51, AD-152650, 1958.
2. ELLIOTT,P. B. Table of d'. In Signal Detection and Recognition by Human Observers, J. A. SWETS
(Editor), pp. 651-684. Wiley,New York, 1964.
3. KEPPEL,G. and UND~nWOOI~,B. J. Proactive inhibition in short-term retention of single items. J. verb
Learn. verb Behav. 1, 153-161, 1962.
4. MELTON,A. W. Implications of short-term memory for a general theory of memory. J. verb Learn.
verb Behav. 2, 1-21, 1963.
5. MILNER,BRENDA.Amnesia following operation on the temporal lobes. In Amnesia, C.W.M. Wmra~"
and O. L. ZANGWILL(Editors), pp. 109-133. Butterworth, London, 1966.
6. MORDOCI<,B. B. The retention of individual items. J. exp. Psychol. 62, 618-625, 1961.
7. NORMAN,n. A. and WICKELGREN,W. A. Short-term recognition memory for single digits and pairs
of digits. J. exp. Psychol. 70, 479-489, 1965.
244 WAYNE A. WICKELGREN
8. W1CKELGREN,W. A. Short-term memory for repeated and non-repeated items. Quart. J. exp. Psycho/.
17, 14-25, 1965.
9. WICKELGREN,W. m. Acoustic similarity and retroactive interference in short-term memory. J. verb
Learn. verb Behav. 4, 53-61, 1965.
10. WICKELGREN,W. m. Short-term memory for phonemically similar lists. Amer. J. PsychoL 78, 567-574,
1965.
11. WICKELGREN, W. A. Phonemic similarity and interference in short-term memory for single letters.
J. exp. Psychol. 71, 396-404, 1966.
12. WICI(ELGREN,W. A. Consolidation and retroactive interference in short-term recognition memory for
pitch. 3. exp. Psychol. 72, 250-259, 1966.
13. WICKELGREN,W. A. Associative intrusions in short-term recall. J. exp. Psychol. 72, 853-858, 1966.
14. WICI(ELCREN,W. A. Associative strength theory of recognition memory for pitch. (In press).
15. WICKELGREN,W. A. and NORMAN,D. A. Strength models and serial position in short-term recognition
memory. J. math. Psychol. 3, 316-347, 1966.
16. WIC~ENS, D. D., BORN, D. G. and ALLEN, C. K. Proactive inhibition and item similarity in short-term
memory. J. verb Learn. verb Behav. 2, 440-445, 1963.
R~sum~--Les possibilit6s de reconnaissance mn6sique ~. court terme d'une part de nombres b.
un et gt trois chiffres et d'autre part de la hauteur de tons purs, ont 6t6 etudi6es chez un sujet
(H.M.) pr6sentant une amn6sie ant&ograde quasi-absolue sans d6ficit apparent de la r6tention
imm6diate. La raison premiere de cette 6rude 6tait d'6prouver une th6orie dynamique de la
m6moire d6velopp6e h partir d'observations faites chez le sujet normal, th6orie impliquant
l'existence de deux traces mn6siques de forces diff6rentes (dual-trace strength theory). A partir
de cette th6orie sur la m6moire ~tcourt terme, on 6tudie une courbe exponentielle de d6gradation
des engrammes dans le cas de H.M., courbe comparable ~tcelle que l'on trouve chez le sujet
normal. Ces r6sultats compl6tent les observations d6j~. faites chez H.M., observations qui
supportent la th6orie de la m6moire formul6e par WICKELGRENet NORMAN[15]
Zusammenfassung--Eine Versuchsperson (H.M.) wurde gepriift, bei der nicht nur ein normales
kurzfristiges Erinnerungsverm6gen zu bestehen scheint, sondern zugleich eine fast v611ige
Unf~ihigkeit, 1/ingerw~ihrende Ged/ichtnisspuren zu bilden. Die Versuche wurden anhand von
einstelligen und dreistelligen Zahlen, sowie auch anhand von Tonh6hen gewisser Grundt6ne
durchgeffihrt. Hauptzweck dieser Untersuchungen war es, den kurzfristigen Teil einer
doppelspurigen Ged~ichtnistheorie (verschiedene Prozesse fiJr kurz- und langfristige Erinnerungen)
zu pri.ifen. Diese Theorie, die wir fiir normale Versuchspersonen entwickelt hatten,
unterscheidet zwei verschiedene Prozesse, einen fiir kurzfristige, und einen weiteren ffir
langfristige Spurenbildungen, wobei beriJcksichtigt werden muss, dass die St~irke der jeweiligen
Spurenbildung sich mit der Zeit ver~indert (double-trace strength theory).
Die Kurven, die den Abfall in der St~irke der Spuren darstellen, sind bei H.M. ann~ihernd
exponentiell, ganz wie es in der St/irketheorie (strength theory) fiir kurzfristiges Erinnerungsverm6gen
postuliert ist. ~berdies unterscheidet sich der Grad dieses Abfalles bei H.M. nicht
wesentlich von dem bei normalen Versuchspersonen. Diese Ergebnisse stimmen mit frtiher an
H.M. getroffenen Beobachtungen iiberein, die gleichfalls eine Doppelspurtheorie des Ged/ichthisses
nahelegten. Insbesondere aber unterstiitzen die vorliegenden Beobachtungen die
St/irketheorie des Ged/ichtnisses, die [15] yon WICKELGRENund NORMAN aufgestellt wurde.