INTACT PRIMING OF PATTERNS DESPITE IMPAIRED
MEMORY*
JOHN D. E. GABRIELI,~~ WILLIAM MILBERG,~ MARGARET M. KEANEt and
SUZANNE CORKING
tDepartment of Brain and Cognitive Sciences and Clinical Research Center, Massachusetts Institute of
Technology, Cambridge. MA 02139.U.S.A.; ZDepartment of Psychology. Northwestern University. Evanston.
IL, U.S.A.: and $GRECC. West Roxbury VA Center. West Roxbury, MA 02132, U.S.A.
(Received 24 Auyust 1989; uccepred 15 November 1989)
Abstract-The priming of patterns was examined in normal control subjects (NCS) and in the
amnesic patient H.M., whose anterograde amnesia followed bilateral medial temporal-lobe excision
33 years earlier. Despite H.M.‘s impaired recognition of the patterns. he demonstrated a pattern
priming effect equivalent in magnitude to that of the NCS. The results demonstrated that intact
priming with novel. nonverbal material can occur despite severe amnesia. Pattern priming may reflect
adaptive visual processes involved in learning novel perceptual organizations of known visual codes.
such as letters, words. shapes, and objects.
INTRODUCTION
PRIMING refers to the influence of prior processing of stimuli upon later performance with
those stimuli. For example, exposure to a word in a study list increases the likelihood that
normal and amnesic subjects will complete a 3-letter stem to that word [lo, 11, 20, 451,
will produce that word when listing members of a category [15, 18, 223, and will identify
that word when it is presented very briefly [2,4, 131 or when it emerges from visual noise
[ 11,453. Remarkably, the magnitude of these priming effects is normal (intact) in amnesic
patients despite their impaired recall or recognition of the study list words. The presence
of intact priming in amnesia indicates that priming can be mediated by memory systems
that are separable from those that subserve recall and recognition: a memory system is
defined as the minimal neural network required to record, retain, and retrieve a form of
knowledge.
Priming can be demonstrated on a broad range of tasks and is probably not a unitary
phenomenon. Nevertheless, most demonstrations of intact priming in global amnesia have
used verbal tasks, such as reading, spelling, or completing letters into words. There are,
however, a few examples of priming in amnesia that do not involve words as stimuli. After
amnesic patients identified incomplete line drawings of animals and common objects, they
demonstrated considerable improvement when reidentifying the incomplete drawings an
hour or more later [26,44]. A similar finding was reported in eight amnesic patients (seven of
*Presented in part at the Society for Neuroscience meeting in Washington, D.C., November 1986, and in the first
author’s doctoral thesis submitted to the Massachusetts Institute of Technology.
Korrespondence to be addressed to John D. E. Gabrieli, 102 Swift Hall, Department of Psychology,
Northwestern University, Evanston, IL 60208. U.S.A.
417
418 J. D. E. GABRIELI et al
whom had closed-head injuries) who reduced greatly their times to detect hidden figures in
picture puzzles that they had studied a day earlier [9]. Three amnesic patients also improved
their ability to find incongruous features in complex cartoon-type drawings taken from the
McGill anomalies task [46]; the improvements were maintained for as long as 4 days.
These demonstrations of nonverbal priming in amnesia are intriguing, but their meaning
remains elusive for several reasons. First, in no case was the priming demonstrated as being
intact in the amnesic patients (i.e. as being equal to that of control subjects). Priming in the
amnesic patients was either less than that seen in the control subjects [26, 441 or not
compared to that of control subjects [9, 461. Second, no tests of explicit recall and
recognition of the nonverbal stimuli were given in those studies, precluding a rigorous
dissociation between nonverbal priming and recall/recognition memory. It is possible that
given a multiple-choice or yes/no recognition test, the amnesic subjects in the earlier studies
would have shown a recognition impairment comparable to their priming impairment.
MILNER et al. [26] hypothesized that the small deficit in priming shown by amnesic patients
was due to their being restricted to using perceptual memory while normal subjects could
consult perceptual memory and explicit verbal memory of the prior testing session. Although
Milner’s hypothesis is plausible, it remains unknown whether nonverbal priming can be fully
intact in amnesia and can be dissociated rigorously from recall and recognition memory.
The status of nonverbal priming in amnesia is important for theories about the nature of
preserved priming capacities. In attempting to elucidate the processes that support verbal
priming, investigators have explored the boundary conditions for intact priming effects in
amnesia [33, 34, 391. Evidence exists that intact priming can occur across modalities and
without physical similarities between the primed word and the stimulus that evokes access of
the primed word [19,41]. Amnesic patients, however, have failed to show normal priming
for nonwords on some tasks [4, 10, but see 2,11, 131. One explanation for this finding is that
the nonwords cannot be known premorbidly by the patients, and that their anterograde
memory impairments prevent them from gaining a lexical or semantic representation of
newly seen nonwords. If priming depends upon the activation of an established
representation (e.g. [27]), it may be hypothesized that amnesic patients can show intact
priming only when they have premorbid knowledge of the primed response, i.e. when they
possess a normal initial representation of that stimulus. It is easy to identify knowledge of
words as being the basis of verbal priming. but is less clear what could constitute a knowledge
base for nonverbal priming.
The present study sought to discover whether an amnesic patient could show normal
priming with stimuli that were patterns and that lacked premorbidly established mental
representations of the sort that appear necessary for verbal priming. Dot-pattern stimuli
were used that could be organized or parsed perceptually to suggest a variety of figures
(Fig. I ). We postulated that the perceptual interpretation of a dot pattern depended upon
flexible visual processes that could be influenced by prior experience with a particular figural
organization of the dot pattern. If such priming effects were to occur in normal control
subjects (NCS), it would be possible to compare their priming effects to those of an amnesic
patient who, unlike the NCS, would be unable to remember the figures that were biasing the
later perceptual organization of the dot patterns.
Priming Study 1 examined whether experience with a figure primed (above a baseline or
control condition) the perceptual organization of a dot pattern in NCS and in the amnesic
patient H.M. Priming Study 2 was an extension of Study 1 that employed a different method
of establishing a baseline. Also, a recognition memory study was conducted in order to
NONVERBAL PRIMING IN AMNESIA 419
establish whether H.M. could remember the figures (i.e. the source ofthe priming) at the time
that the status of the priming effect was measured.
METHOD
Subjects
The amnesic patient H.M. (aged 61 or 62 years in the several testing sessions, with 12 years ofeducation) and 15
NCS participated in this study. H.M. underwent bilateral medial temporal-lobe resection in 1953. at age 27. in an
attempt to alleviate his otherwise intractable epilepsy 1371. The surgical removal was estimated to extend 8 cm back
from the tips of the temporal lobes, including the prepyriform gyrus, uncus, amygdala. hippocampus. and
parahippocampal gyrus. Temporal neocortex was reported to have been spared. Since the resection. H.M. has had a
severe and pervasive anterograde amnesia for verbal and nonverbal materials in all modalities [8. 381. H.M. took
part in 4 testing series that were at least 2 months apart: Priming Study 1 in April, 1986: the recognition-memory
tests in February, 1987; a replication of Priming Study 1in April. 1987: and Priming Study 2 in November. 1987.
In Priming Study 1, five men and six women constituted the NCS group, which had a mean age of 56.8 years
(range 49-65) and a mean educational level of 13.9 years (range 1I-18). In Priming Study 2. two men and one
woman constituted the NCS group: they had a mean age of 59.7 years (range 5CL74) and a mean educational level of
15.7 (range 14-17). For the test of recognition memory. two women and one man constituted the NCS group: they
had a mean age 66.3 years (range 57-82) and a mean educational level of 13.7 (range 12-17).
Materials
The stimuli, taken from GARNER [ 161. were derived from six dot patterns. each consisting offive dots from the nine
possible dots in a 3 x 3 square matrix (Fig. I). For each dot pattern. three target figures were constructed by
connecting the five dots with straight lines. Three 6-item priming test forms were made by randomly selecting one
target figure for each dot pattern. Explicit recognition memory for the figures was measured using 4-choice
recognition tests. There were three recognition memory tests, each ofwhich was paired to one of the priming tests in
that the same target figures were used for the corresponding priming and recognition tests, For each recognition
item. a dot pattern was shown at the top of the page and four figures that connected the dots were shown at the
bottom. One figure was the target and three were foils.
Procedure
For Priming Study 1, there were four test sessions. In Session I, subjects were asked to draw any figure onto each
dot pattern that connected all the dots with straight lines. The figures that subjects drew constituted their baseline
figures. In Session 2, which always occurred at least 6 hr later, subjects were asked to copy six target figures from
one form onto corresponding dot patterns (each figure was above the corresponding pattern on an 8; x 11 in. piece
of paper). Then, the materials were removed. and subjects were asked to write down the names of as many famous
entertainers from the 20th~century as they could: this interference task was done for 3 min. Next. subjects were given
the dot patterns (without any lines) and asked to draw any figure that connected the dots in each pattern with
straight lines. In Session 3. the same procedure was followed except that a second. different set of target figures were
copied onto the same dot patterns. and subjects were asked to write down the names of as many famous political
figures from the ZOth-century as they could for the interference task. In Session 4, the third set of target ligures were
copied onto the dot patterns. and subjects were asked to write down the names of as many athletes from the ZOthcentury
as they could for the 3-min interference task. Unlike Sessions 2 and 3, subjects were then asked to draw the
same figures they had just copied onto the dot patterns. Thus, Session 4 provided a measure of cued recall (the dot
patterns being the cues) of incidentally encoded figures. The three test forms were administered in the same order for
all subjects, with a day between Sessions 2. 3 and 4. H.M. participated in all four sessions on each of two separate
occasions, a year apart.
The protocol was modified for Priming Study 2, which was conducted 6 months after Study 1. As before, subjects
drew the first figure that came to mind on each of the dot patterns. These figures served as baseline ligures. Then, at
least 6 hr later, subjects copied target fgures (from the same set used in Priming Study I )that had been selected on
the basis of their not being drawn in the baseline session. This procedure ensured that none ofthe target figures were
an individual subject’s primary, or spontaneous, perceptual organization of the dot patterns. The rest of the test
(copy figures, the interference task, and drawing figures onto dot patterns) was administered as described above.
For the separate test of recognition memory. subjects copied one set of target figures onto dot patterns. Following
the same 3-min delay task used in the priming test, subjects were asked to indicate which one of the four figures on
the bottom of each page they had copied earlier onto the dot pattern displayed at the top of each page. Each of the
NCS did this task once with six target figures; H.M. did this task with IX target figures, 6 per session in each of
3 sessions.
TARGET
J. D. E. GABRIELI et ul.
SOME OTHER POSSIBLE COMPLETIONS
Fig. I. Dot patterns are shown in column one. Examples of target figures copied onto dot patterns (i.e.
primed figures) are shown in column two. Some other figures that could be drawn onto dot patterns
are shown In the remaining columns.
RESULTS
In order to determine whether the control subjects demonstrated priming in Study 1,
paired comparisons were made between the number of target figures drawn in the baseline
condition and the number drawn in the primed (post-copying) condition. In baseline
Session 1, the control subjects drew randomly a mean of 0.27 (SD, 0.47) target figures used
in Session 2 compared to mean of 1.27 (SD, 0.64) target figures following the copying in
Session 2. This difference was significant (r = 5.24, d.f. = 10, P~0.01. one-tailed), indicating
NONVERBAL PRIMING IN AMPU’ESIA JZI
that copying a target figure onto a dot pattern increased the probability of its subsequently
being drawn freely onto the same dot pattern. In baseline Session 1. the control subjects drew
randomly a mean of 1.18 (SD, 0.75) target figures used in Session 3 compared to a mean of
1.91 (SD, 1.04) target figures following the copying in Session 3. This difference also was
significant (r = 1.90, d.f. = 10, P<O.O4,one-tailed). Thus, control subjects showed significant
priming in both Sessions.
In order to compare the status of priming in H.M. to that of the control subjects. a strict
scoring method provided individual priming scores for Sessions 2 and 3. If a target figure was
drawn by a subject in the baseline condition, that figure was discounted for that subject.
Thus, each subject served as his or her own control. The priming score was the percentage of
remaining target figures drawn identically to the copied figures in Sessions 2 and 3. The
control group had a mean score of 20.8% (SD, 15.9). H.M. had, over four separate sessions. a
mean priming score of 28% (Fig. 2). Other scoring methods, including scoring by line
segments, produced the same finding of intact priming in H.M. In Priming Study 2, subjects
again served as their own controls because the target figures were selected to be different from
those drawn in the baseline condition. The priming score was the percentage of target figures
drawn identically to the copied figures. The control groups’s mean score was 27%: H.M.
scored 33 %.
Recognition Priming Study 1 Priming Study 2
Fig. 2. Results of the recognition memory test. Priming Study I, and Priming Study 2. showing mean
percentage correct above chance for normal control subjects (NCS) and H.M.
Inspection of H.M.‘s priming performance across the different sets of target figures and the
two different protocols (Priming Study I and Study 2) provides further evidence in favor of
his showing a reliable priming effect. H.M.‘s priming scores reflected 10 instances of his
primed drawing of seven different target figures. He primed once on five of the figures, two
times on another figure, and three times on another figure. For two of the dot patterns, he
drew different primed figures on the same dot pattern depending on the particular target
figure that he had copied in that session. He drew only one of the seven primed target figures
(a figure he primed on once) in the I8 baseline trials (3 trials per pattern) that were given at
the start of the three experimental series (Study I twice and Study 2). Restated in terms of
percentages, H.M. randomly drew 6% of the target figures that he subsequently primed on,
vs priming scores of 28 and 33% in Studies 1and 2, respectively. He demonstrated 90% of his
422 J. D. E. GAlWELl rr al.
priming on figures that he never drew spontaneously. Finally. H.M. showed priming in all
five separate testing sessions.
Curd recall and recognition memor?
The cued recall scores of incidentally encoded target figures from Session 4 (Study 1) were
calculated in the same way as the priming scores, i.e. a target figure was counted as being
recalled correctly only if it was completely correct. The control group had a mean score of
23.5% (SD, 21 .l). H.M., over two sessions, had a mean score of 33.3%. Because different
figures were used in Session 4 than in Sessions 2 and 3, caution must be taken in comparing
the cued recall scores to the priming scores. With this caveat in mind, the cued recall scores of
the control subjects were surprisingly low. Their scores were not significantly higher than
their priming scores (20.8%) and were lower than H.M.‘s cued recall score. A month or more
later, we retested five of the control subjects with the same materials used in Session 4 but
with different instructions. The subjects were told, as they copied the figures, that their
memory for the figures would be tested; otherwise the procedure was the same. These five
subjects, who originally recalled a mean of 15.6% of the figures following incidental
encoding, recalled 47.0% following intentional encoding.
The recognition score was the percentage of correct answers; chance performance was
25%. The control group had a mean recognition score of 78% (range 67.-100). H.M.‘s score
was 33% (Fig. 2). It remained possible, however, that H.M. could have recognized those
particular figures that had evoked priming in him. Therefore, his recognition score was
recalculated separately for the seven figures that he had primed on. H.M. recognized only
one of those figures corrrectly, a 14% recognition performance that was below chance (25%)
and below his overall recognition performance of 33%.
DISCUSSION
Copying the target figures influenced H.M.‘s subsequent free drawing onto the dot
patterns as much as copying influenced the free drawing of the NCS. H.M. had a normal
magnitude of priming on three different test forms, on three separate testing occasions, and
with two different protocols. The intact priming of patterns occurred despite H.M.‘s
impaired recognition of the target figures. The results demonstrate a novel form of pattern
priming in normal subjects, provide the first report ofintact nonverbal priming in H.M., and
show that intact priming of nonverbal material can be dissociated from impaired recognition
memory of the same material. Because the patterns did not relate directly to premorbidly
formed verbal representations, their similar influence on normal and amnesic subjects
extends the boundary conditions for the occurrence of intact priming in global amnesia.
The cued recall results are less straightforward. In accordance with prior analyses ofverbal
cued performance by amnesic patients [20,41], we interpret H.M.‘s apparently good cued
recall performance as another example of his intact priming capacity. That is, when asked to
recall and draw the previously copied target figures, H.M. simply drew the first figures that
came to mind as he had in the priming sessions. Thus, his cued recall score of33% was similar
to his priming score of 28%. The same reliance on the first figure that came to mind, however,
did not help on the multiple-choice recognition test. H.M. scored identically (33%) on the
cued recall and multiple-choice recognition tests, whereas control subjects did far better on
the recognition test (78%) than on the cued recall test (23.5%). H.M.‘s pattern of
performance on the two tests of explicit memory for the target figures is unlike that of the
NONVERBAL PRIMING Il.- AMNESIA 423
normal control subjects, and supports the idea that his cued recall responses were based upon
priming mechanisms and not upon the mechanisms that support raall and recognition
memory.
Unlike the results reported by GRAF et al. [20] and by SHIMAMURAand SQUIRE [413. the
NCS in the present experiment did not have a better cued recall than priming score,
suggesting that the control subjects did not have strong voluntary or explicit access to their
memory of the copied target figures. Like H.M., the control subjects may have drawn the first
figures that they could imagine. It may be that H.M. achieved normal levels of nonverbal
priming in this study because normal subjects could not consult explicit memory of any
target figures without the entire target figure being shown, as in the multiple-choice
recognition test. Additional support for the notion that pattern priming may depend upon
perceptual learning mechanisms that are dissociable from recognition memory comes from
two recent studies with normal subjects [29,35]. One study [29] involved stimuli that were
quite similar to the ones used in the present study, and also used a four-alternative, forcedchoice
measure of recognition memory. Unlike the present study, the measure of priming was
performance on a speeded perceptual task. Recognition memory and priming in normal
subjects were dissociable by the criteria of differential declines in memory performance over
time and of stochastic independence between recognition and priming performance. The
other study with normal subjects found a dissociation between priming and recognition
memory for line drawings of three-dimensional objects [35]. The results of those two studies
and the failure of the control subjects in the present experiment to score higher on cued recall
than on priming measures (unless subjects were forewarned of a memory test) indicate that
the preservation of pattern priming in H.M. reflects the separability of perceptual learning
and recognition processes in the course of normal cognition.
The preservation of pattern priming in H.M. is consistent with dichotomies that have been
drawn between the different kinds oflearning that are impaired and intact in amnesia. &HEN
and SQUIRE[6] proposed a distinction between declarative and procedural kinds of memory
and GRAF and SCHACTER [IS] a distinction between explicit and implicit memory. In
accordance with many prior studies with H.M., he was impaired by the multiple-choice
recognition measure of declarative, explicit memory. He was unimpaired, however, by the
procedural, implicit priming measure. It should also be pointed out that the prior
dichotomies did not focus exclusively upon measures of verbal memory. There are examples
of skill learning that are preserved in amnesia, that are consistent with the declarative/procedural
and explicit/implicit distinctions, and that do not have an apparent verbal basis.
Mirror tracing [25], rotary pursuit [7], and serial reaction time [30] are three such skills. It is
difficult, however, to compare skill learning and priming studies directly, because the former
have always been given in protocols that involve many trials of the same kind in a single
session, while priming studies have tended to use a single presentation of a stimulus and a
single subsequent response.
The pattern priming results address theories about the basis of intact priming in amnesia.
SCHACTER and GRAF [36] proposed that there are two kinds of priming, activational and
episodic. Activational priming is viewed as being dependent upon the activation of long-term
memory representations and being intact in severely amnesic patients who established
premorbidly the required memory representations. Episodic priming is seen as being
dependent upon the formation of new memory representations, or new associations, that are
mediated, at least in part, by the same episodic memory processes that support recall and
recognition. Therefore, episodic priming may be preserved only in patients with relatively
424 J. D. E. GABRIELI et al.
mild memory impairments (e.g. [36,42, but see 21). It is difficult to describe pattern priming
as only the activation of long-term memory representations, because it seems unlikely that
either normal subjects or H.M. had pre-existing memory representations associating the
particular dot patterns with the particular target figures that were employed in the present
experiment. Pattern priming required the formation of new associations between dot
patterns and target figures that were established in the course of copying target figures onto
dot patterns. The new associations influenced subsequent processes of perceptual parsing
and organization that assigned a structure to the dot pattern and guided the NCS’ and
H.M.‘s primed drawings. The measure of new associations was the presence and degree of
priming. H.M.‘s severe amnesia, however, precludes the possibility that his intact priming
pattern reflects the operation of recall and recognition mechanisms [8, 381. Thus, H.M.‘s
intact pattern priming cannot be classified satisfactorily as an example of either activational
or episodic priming.
We propose a different interpretation of the basis of pattern priming, one that emphasizes
the visual information-processing demands of perceiving a figural organization emerging from
a pattern of unconnected dots. Through vision, we may see an infinite variety of stimuli as we
encounter novel stimuli or all the changes in appearance of known stimuli due to variation in
size, position, shading, and occlusion. Because there is an infinite number of potential visual
stimuli, a perceptual learning system cannot establish a limited vocabulary of potential stimuli,
as a system designed for lexical access may do. An adaptive perceptual system must have the
capacity to establish a new association between any visual stimulus perceived in a specific
instance (e.g. a dot pattern) and the outcome of visual processes that represent the stimulus in
meaningful visual codes, such as letters. words, objects, or, in the present case, target figures.
Such an adaptive perceptual system may also be activational in part in if it uses a limited longterm
vocabulary ofshapes (e.g. 1)or parsing routines (e.g. 21) to analyze the variety ofpercepts
that may be encountered in vision. Thus, it seems that perceptually based priming must involve
the formation of novel associations between percepts and structural descriptions of percepts.
As shown by H.M.‘s intact pattern priming and by prior reports of robust perceptual priming
in amnesic patients [9, 26, 44, 461. new associations supporting perceptual priming may be
established despite severe deficits in episodic memory.
The idea that novel perceptual associations may be established without declarative or
explicit memory is consistent with a number of studies that found intact perceptual priming
of nonwords in severely amnesic patients. Activational theories of priming predict that
amnesic patients cannot show priming for nonwords because amnesic patients do not have
pre-existing memory representations of nonwords and, unlike normal subjects. cannot
acquire such novel representations postmorbidly. Yet H.M. has shown intact priming for
nonwords on two perceptual tasks (1 1. 13: and see 2 for a similar finding with S.S.).
Moreover, the activational account predicts that severely amnesic patients will not prime for
new associations between premorbidly known, but premorbidly unassociated, words.
MOKOVITC-H [B], however. found priming of new associations in amnesic patients on two
tasks involving perceptually degraded verbal stimuli (the patients in that study were a mixed
group in terms of etiology and severity). In addition. COHEN et ~1. [S] reported that amnesic
patients showed substantial priming on a symmetry-detection task involving novel
checkerboard patterns. The activational:episodic distinction, therefore, may not apply to
perceptual tasks in which even severely amnesic patients seem to learn new associations.
That distinction, however. may still apply to priming tasks that do not seem to be primarily
perceptual in nature. such as word completion, fragment completion, category exemplar
NONVERBAL PRIMING IN AMNESIA 425
production, lexical decision, and sentence puzzles. Performance on these tasks is well
documented to be intact in severly amnesic patients when activation of pre-existing memory
representations can account for priming (e.g. 2, 3, 10, 11, 12. 13-15, 17-20, 23, 3941) but
impaired when new associations are needed to support priming (e.g. 3, 10, 13.22.24.36.42.
but see 2). SCHACTER [34], in a review of priming studies with normal subjects and of
neuropsychological studies with agnosic and dyslexic patients, has come to a similar
conclusion about “perceptual representations systems” that may mediate priming on tasks in
which word form or structural descriptions of objects play an important role.
We have suggested [ 13, 14,231 that extrastriate visual areas may be especially important
in perceptual priming. Extrastriate areas show selective activation in positron emission
tomography (PET) studies when normal subjects perform a task thought to be sensitive to
word form [32]. Furthermore, single-cell recordings in monkeys have shown that neurons in
extrastriate visual cortex (V2), but not primary striate visual cortex (Vl ), respond selectively
to stimuli that humans perceive as contours. or subjective continuities, despite no actual line
being present [31,43]. In the pattern priming task, one could think of the lines that subjects
draw connecting the dot patterns as expressions of subjective continuities, and the priming of
normal subjects and of H.M. as changes in the probability of perceiving one subjective
pattern of line continuities over other plausible subjective patterns. In that case, pattern
priming could reflect, at least in part, changes in extrastriate perceptual mechanisms that
occured when subjects copied target figures onto dot patterns.
Pattern priming and the other sorts of perceptually based priming discussed above
demonstrate the influence of prior perceptual analysis upon later perceptual organization of
flexibly interpretable stimuli. H.M.‘s intact pattern priming indicates that such perceptual
learning does not depend upon recall and recognition memory capacities. The poor cued
recall performance of the NCS demonstrated that even people with unimpaired recall and
recognition have poor conscious access to the newly acquired perceptual information.
Perhaps the perceptual associations that mediate pattern priming are established in early
stages of visual processing that are relatively inaccessible to awareness. It may be that visual
areas of the brain learn from perceptual experience by changing or redistributing the
probability of deriving a particular perceptual analysis of a given stimulus. In the case of
stimuli that have a correct interpretation, such as incomplete or hidden figures, experiencebased
change results in enhanced visual skill in item-specific object recognition. In the
present experiment, there were no objectively correct figural organizations of the dot
patterns, but the priming may be interpreted as an experience-based change in deriving one
particular figural analysis of a dot pattern. The vocabulary of figure knowledge is less well
specified than that of verbal or object knowledge, but some proposals have been made (e.g. I,
21). Pattern priming, then, may reflect important adaptive properties of early perceptual
processes that play a useful and ubiquitious role in vision.
Acktlowledgements-We thank Dr Brenda Milner for the opportunity to examine H.M. and arc grateful IO the soar
of the Clinical Research Center for their special care of H.M. Dan Schacter and an anonymous reviewer provided
very helpful comments. We also thank Eliizd McGrand and Sasha Nyary for their assistance in producing this
manuscript. This work was supported by grants MH24433, RR 00088, 2 T32 GM07478. and VA Merit Reveiw
097443765.
REFERENCES
I. BEIDERMAN, I. Human image understandmg: recent research and a theory. Compufrr Vision. C‘ruphic’s. und
Image Procr.kq 32. 29 73. 1985. .
426 .I. D. E. GABRIELI et ol.
2. CERMAK;,L. S.. BLACKFORD. S. P., O’COX~OR. M. and BLEICH, R. B. The implicit memory ability of a patient
with amnesia due to encephalitis. Brain Cognif. 7, 145-156, 1988.
3. CERMAK, L. S., BLEICH, R. P. and BLA~KFOR~XS. P. Deficits in the implicit retention of new associations by
alcoholic Korsakoff patients. Brain Co@. 7, 312. 323. 1988.
4. CERMAK. L. S.. TALBOT, N.. CHANDLER, K. and WOLBARST. L. R. The perceptual primmg phenomenon m
amnesia. Neuropsgcholoyia 23, 615-622, 1985.
5. COHEN, N. J.. ABRAMS,I., HARLEY. W. S.,TABOR. L.. GORIXXQ.B. and SEJNO\S’SKI.T. J. Perceptual skill learning
and repetition priming for novel materials in amnesic patients. normal subjects. and neuron-like networh
models. Sac. Neurosci. Ahsrr. 12, 1162, 1986.
6. COHEN, N. J. and SQUIRE, L. R. Preservjed learning and retention of pattern-analyzmg skill in amnesia:
dissociation of knowing how and knowing that. Science 210, 207-210. 1980.
7. CORKIN. S. Acquistion of motor skill after bilateral medical temporal-lobe excision. :Vrurops~c/~(,/t,(lru 6,
255-265. 1968.
8. CORKIN, S. Lasting consequences of bilateral medial temporal lobectomy: clinical course and experimental
findings in H.M. Serninurs in Neurology 4, 252-262. 1984.
9. CROVITZ, H. F.. HARVEY,M. T. and MCCLANAHAY, S. Hidden memory: a rapid method for the study ofamnesia
using perceptual learning. Cortex 17, 273 278. 1981.
10. DIAMOND, R. and RozI~‘. P. Activation ofexistingmemories in the amnesic syndromes. J. Ahmmn. Ps~hol. 93.
988105, 1984.
I I. GABRIELI. J. D. E. Memory systems of the human brain: dissociations among learning capacities in amnesia
Unpublished doctoral dissertation, Massachusetts Institute of Technology, Cambridge. MA, 1986.
12. GABRIELI, J. D. E., COHEN. N. J. and CORKIN. S. The Impaired learning of semantic knowledge following
bilateral medial temporal-lobe resection. Brain Cognit. 7, 157-177. 1988.
13. GABRIELI, J. D. E. and KEANE. M. M. Prtming in the amnesic patient H.M.: new findings and a theory of intact
and impaired priming in patients with memory disorders. Sot. Neurosci. Ahstr. 14. 1290. 1988.
14. GABRIELI, J. D. E., KEANE, M. M.. STANGER. B. 2.. KJELGAARD. M. M., BANKS, K. S.. COKKII\;, S. and
GROWDON, J. H. Dissociations among perceptual-structural, lexical-semantic and event-fact memory
subsystems in global amnesia. Alzheimer’s disease. and normal memory (submitted).
15. GARDNER. H.. BOLLER. F.. MOREINES, J. and BUTTERS, N. Retrieving information from KorsakolI patients:
effects of categorical cues and reference to the task. Corte.x 9. 165-175. 1973.
16. GARNER, W. R. and CLEMENT,D. E. Goodness of pattern and pattern uncertainty. J. Vwh. Lrum. I’erh. Bd~m~.
2,446452, 1963.
17. GORDON, B. Preserved learning ofnovelinformation in amnesia: evidence for multiple memory systems. Brain
Cognit. 7, 257-282, 1988.
18. GKAF. P. and Schacter, D. L. Implicit and explicit memory for new associations in normal and amnesic subjects.
J. Eup. PsychoI.: Learn. Mem. Cognil. 11, 501 518. 1985.
19. GRAF, P., SHIMAMURA.A. P. and SQUIRE. L. R. Priming across modalities and priming across category levels:
Extending the domain of preserved function in amnesia. J. Ewp. Ps~chol.: Learn. Mem. Cognit. I I. 386 -396.
1985.
20. GRAF. P., SQUIRE,L. R. and MANULER, G. The information that amnesic patients do not forget. J. E.vp. Ps&ol.:
Learn. Mem. Cognit. 10, 164 178. 1984.
21. HOFFMAN. D. D. and RITHARUS. W. A. Parts of recognition. Coy&. 18, 65596, 1984.
22. KEANE, M. M.. GAHKIELI.J. D. E. and CORKIN, S. Multiple relations between fact-learning and priming in global
amnesia. Sm. Neurosri. Ahsrr, 13. 1454. 1987.
23. KEANE, M. M.. GABRIELI. J. D. E.. KJELC~AAR~.M. M., GR~WIX>X. J. H. and CORKIN, S. Dissociation between
two kinds of priming in global amnesia and Alzheimer’s disease. Sot. Neurosci. Ahstr. 14, 1290. 1988.
24. MCANI)REWS, M. P.. GLISKY. E. L. and SCHACTER, D. L. When priming persists: long-lasting implicit memory
for a single study episode in amnestc patients. .‘Veurops!,c/lo/ogiu 25. 487-506, 1987.
25. MILNER. B. Les troubles de la memoire accompagnant des lesions hippocdmpiques bilaterales. In Physiologic de
I’hippocumpr. Paris: Centrc National de la Richerche Scientifique. 1962.
26. MILNER. B.. CORKIX. S. and TEUBER, H.-L. Further analysis of the hippocampal amnesia syndrome: l4-year
follow-up study of H.M. Nuurops~clloloMia 6, 215.~234, 1968.
27. MORTON, J. The interaction of information in word recognition. Ps~c/iol. Rer. 76, 165-178, 1969.
28. M~S~OVIT~H, M.. WINOTUK. G. and MTLACHLAN. D. Memory as assessed by recognition and readmg time in
normal and memory-impaired people with Alzheimer’s disease and other neurological disorders. J. Eup.
f.~$w/.: Gm. 115, 331 347. 1986.
29. MUSEN. G. and TREISMAX.A. Implicit and explicit memory for visual patterns. J. Exp. Psyhol.: Learn Mm.
Cognit. 16. 127-137. 1990.
30. NISSEN. M. J. and BULLEMEK.P. Attentional requirements of learning: evidence from performance measures.
Cognit. Psycho/. 19, I-32. 19X7.
31. PETERHANS, E. and VON DEK HEYIIT. R. Mechanisms of contour perception in monkey visual cortex. II.
Contours bridging gaps. J. ,Yruro.wi. 9. 1749 1763. 1989.
NONVERBALPRIMINGIN AMNESIA 427
32. POWER. M. I.. PETERSEN:,S. E.. Fox. P. T. and RAICHLE. M. E. Localization of cognitive operations in the
human brain. Science 240, 1627-1631, 1988.
33. SCHACTER, D. L. Implicit memory: history and current status. J. Exp. Psycho/.: Learn. Mem. Cognit. 13,
501-518, 1987.
34. SCHATTER, D. L. Perceptual representation systems and implict memory: toward a resolution of the multiple
memory systems debate. In Development and Neural Bases ofHigher Cognitiw Function. .4nnals ofthe iVew Ihrk
Academy qf.Scrences, A. DIAMOND (Editor). in press.
35. SCHACTER.D. L., COOPER. L. A. and DELANEY.S. M. Implicit memory for unfamiliar objects depends on access
to structural descriptions. J. exp. Psychol.: Gen.. in press.
36. SCHACTER. D. L. and GRAF, P. Preserved learning in amnesic patients: perspectives from research on direct
priming. J. clin. exp. Neurops.vchol. 8. 727-743. 1986.
37. SCOVILLE, W. B.. DUNSMORE. R. H.. LIBERSON. W. T.. HENRY. C. E. and PEPE. A. Observations on medial
temporal lobotomy and uncotomy in the treatment of psychotic states. In Proceedings qfrhe Associarion /in
Research in Nervous and Mental Disease. pp. 347-369. Williams and Wilkins: Baltimore. Maryland, 1953.
38. SCOVILLE, W. B. and MILNER, B. Loss of recent memory after bilateral hippocampal lesions. J. .Yeuro/.
Neurosurg. Psychiar. 20, 1l-21, 1957.
39. SHIMAMURA.A. P. Priming effects in amnesia: evidence for a dissociable memory function. Q. J. e.~p. Psycho/.
38A, 619-644, 1986.
40. SHIMAMURA,A. P.. SALMON, D. P., SQUIRE. L. R. and BUTTERS.N. Memory dysfunction and word priming in
dementia and amnesia. Behar. Neurosci. 101, 347-351. 1987.
41. SHIMAMUKA, A. P. and SQUIRE. L. R. Paired-associate learning and priming effects in amnesia: a
neuropsychological study. J. e.up. Psycho/.: Gen. 113, 556570. 1984.
42. SHIMAMURA.A. P. and SQUIRE. L. R. Impaired priming of new associations in amnesia. J. esp. Ps~chol.: Learn.
Mem. Cognit. 15. 721-728. 1989.
43. VONDERHEYDT, R. and PETERHANS.E. Mechanism of contour perception in monkey visual cortex. I. Lines of
pattern discontinuity. J. Neurosci. 9, 1731-1748. 1989.
44. WARRINGTON. E. K. and WEISKRANTZ. L. A new method of testing long-term retention with special reference to
amnesic patients. Nature. Land. 217, 972-974, 1968.
45. WARRINGTON, E. K. and WEISRRANTZ. L. The amnesic syndrome: consolidation or retrieval? Nature. Lor$.
228, 628-663, 1970.
46. WARRINGTON, E. and WEISKRANTZ, L. An analysis of short-term and long-term memory defects in man. In The
Physiological Basis qf Memory,J. A. DEUTSCH. (Editor). pp. 365-396. Academic Press. New York. 1973.