Neuron

Article

Mesmerizing Memories: Brain Substrates

of Episodic Memory Suppression
in Posthypnotic Amnesia
Avi Mendelsohn,1,3 Yossi Chalamish,1,3 Alexander Solomonovich,2 and Yadin Dudai1,*
1Department of Neurobiology, The Weizmann Institute of Science, Rehovot, Israel
2Hypnosis Unit, Wolfson Medical Center, Holon, Israel
3These authors contributed equally to this work.

*Correspondence: yadin.dudai@weizmann.ac.il
DOI 10.1016/j.neuron.2007.11.022

SUMMARY et al., 1968; Rosen and Engle, 1998; Levy and Anderson, 2002;
Racsmany and Conway, 2006). Another involves investigation
Two groups of participants, one susceptible to post- of pathological conditions in which normal memory suppression
hypnotic amnesia (PHA) and the other not, viewed occurs by definition, such as psychogenic or functional amnesia
a movie. A week later, they underwent hypnosis in (Markowitsch, 1999), or is postulated to occur, such as sponta-
the fMRI scanner and received a suggestion to forget neous confabulation (Schnider, 2003). Still another approach
the movie details after hypnosis until receiving a re- bridges the worlds of cognitive research and the clinic. It ad-
dresses certain memory deficits that occur with aging (Hasher
versal cue. The participants were tested twice for
and Zacks, 1988; Gazzaley et al., 2005) or following posthypnotic
memory for the movie and for the context in which
suggestion (Kihlstrom, 1997).
it was shown, under the posthypnotic suggestion The present work uses hypnosis as a tool to tap into memory
and after its reversal, while their brain was scanned. suppression in the brain. Hypnosis was known to healers and
The PHA group showed reduced memory for movie their clients since the dawn of history and was harnessed into
but not for context while under suggestion. Activity the service of western medicine in the past 200 years, following
in occipital, temporal, and prefrontal areas differed the observations of Franz Mesmer, James Braid, and their
among the groups, and, in the PHA group, between followers (Braid, 1845; Gauld, 1995). It is considered in folk
suggestion and reversal conditions. We propose psychology as an altered state of consciousness. The majority
that whereas some of these regions subserve re- of scientific treatments do not refute this intuition, but differ on
trieval of long-term episodic memory, others are the type of alteration, its manifestations in nonhypnotic states,
and the conceptual framework and semantics used to define
involved in inhibiting retrieval, possibly already in
it. Formally, the phenomenon refers to a psychosocial situation,
a preretrieval monitoring stage. Similar mechanisms
mental state, mental or neuronal process, and behavioral proce-
may also underlie other forms of functional amnesia. dure (Hilgard, 1975; Kihlstrom, 1997; Kirsch, 1998; Wagstaff,
1998). The psychosocial situation is of a person, the hypnotized
INTRODUCTION subject, who acts on suggestion from another, the hypnotist. In
self-hypnosis, both roles are played by the same brain. The
For items in memory to be retrieved and guide behavior properly, state, as noted above, is that of altered consciousness, com-
suppression of some memory representations seems to be as monly described as dissociative. The latter notion has evolved
important as the expression of others (Hasher and Zacks, over the years to encompass different mental faculties, which
1988; Levy and Anderson, 2002; Schnider, 2003; Racsmany might also become dissociated in the absence of hypnosis
and Conway, 2006; Gilboa et al., 2006; Bjork, 2007). Indeed, (Hilgard, 1975; Kirsch and Lynn, 1995; Wagstaff, 1998). The pro-
when memory suppression fails, mnemonic-guided behavioral cess is that in which cognition and its brain substrates culminate
interactions with ongoing reality fail as well (Schnider, 2003; in the aforementioned mental state. And the behavioral proce-
Gazzaley et al., 2005). However, despite intriguing data on pos- dure is that in which the hypnotist invokes the aforementioned
tulated processes and manifestations of memory suppression process.
that emerged in recent years from laboratories and clinics alike Individuals vary in their susceptibility to hypnosis (Weitzen-
(Conway and Fthenaki, 2003; Schnider, 2003; Anderson et al., hoffer and Hilgard, 1962; Stern et al., 1979; Lichtenberg et al.,
2004), relatively little is known of the brain mechanisms that sub- 2004). Most pertinent to the topic of the present study is the
serve such suppression. well-established observation that high-hypnotizable individuals
Three major types of experimental approaches reign in the dis- can be induced during the hypnotic state into a situation in
cipline of memory suppression. One involves manipulation of which, on termination of hypnosis, they are unable to recall
learned material in healthy individuals, so that items to be re- information acquired either in the hypnotic session or before
called are either incidentally or intentionally blocked (Bjork it, until presented with a prearranged reversibility cue. This

Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc. 159
 Neuron
Brain Correlates of Posthypnotic Amnesia

Figure 1. Experimental Design

(A) Snapshots from the 45 min documentary movie
presented in the STUDY session.
(B) In the TEST session, performed a week later, all
of the participants underwent hypnosis, during
which they received a suggestion to forget upon
termination of hypnosis the movie details seen in
the STUDY, until they received a reversal cue
that cancelled the suggestion. After termination
of hypnosis, while under the posthypnotic sugges-
tion (Test 1) and following cancellation of sugges-
tion (Test 2), the participants were scanned while
performing a computerized retrieval test that
taxed memory for both movie details (Movie) and
for the contextual details of the study session
(Context). For further details, see Experimental
Procedures.

posthypnotic suggestion state is termed ‘‘posthypnotic amne- RESULTS

sia’’ (PHA; Kihlstrom, 1997). PHA is hence a retrieval rather
than storage deficit and resembles psychogenic or functional Behavioral Performance
amnesia, for which it has been proposed to serve as a model Memory Performance
(Kihlstrom, 1997; Barnier, 2002). PHA is believed to affect mostly Under the influence of the FORGET suggestion, the PHA group
information that is taxed in explicit memory tests (Kihlstrom, exhibited markedly reduced memory performance on Movie
1997). questions compared to the Non-PHA group (Figure 2A)
That PHA can be induced and relieved under controlled con-
ditions in a laboratory setting renders it an appealing model for
investigating brain mechanisms of memory suppression, which
are expected to control the transient retrieval block in functional
amnesia. In this study, we subject high-hypnotic-susceptibility
and low-hypnotic-susceptibility individuals to a controlled
situation that permits them to encode real-life-like episodic
memory. This is done by the presentation of a narrative docu-
mentary movie (Furman et al., 2007). A week later, we place
the participants in the fMRI scanner, hypnotize them, and induce
PHA. This is followed by testing the memory for details in the
movie or details in the context in which the movie was shown,
while brain fMRI signals are acquired (Figure 1). Memory perfor-
mance is tested twice: once when the posthypnotic suggestion
is active and once after it has been relieved by the reversibility
cue. This allows acquisition of brain activity maps in and after
memory suppression and comparison of brain activations in re-
call of target and context items in high-hypnotic-susceptibility
individuals and in their low-susceptibility controls. Our study
identifies large-scale neural circuits that are suppressed com-
pared to baseline activity during suppression of memory perfor-
mance. In addition, we show that left occipital and temporal cor-
tices are suppressed preferentially, whereas the left rostrolateral Figure 2. Memory Performance in the TEST Session
prefrontal cortex is activated preferentially when the memory (A) Performance of PHA (black) and Non-PHA (gray) groups on Movie (left bars)
performance is suppressed. We also demonstrate that in the and Context (right bars) during Test 1. A mixed-model ANOVA analysis, using
high-susceptibility subjects, a network of brain regions shows memory type as a within-subject factor and group as a between-subject factor
revealed a significant interaction, with reduced performance for PHA subjects
recovery from suppression following the reversal of the posthyp-
in Movie but not in Context (F1,22 = 20.38, p < 0.0005).
notic suggestion. We propose that, whereas some of the regions
(B) Performance of PHA (black) and Non-PHA (gray) groups, on Movie (left
identified in our study play a role in retrieval of long-term bars) and Context (right bars) during Test 2. No effects were revealed in
episodic memory, others are involved in inhibiting retrieval, pos- a mixed-model ANOVA. Dashed line indicates chance level performance.
sibly in a preretrieval monitoring stage. Error bars are SEM.

160 Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc.
Neuron
Brain Correlates of Posthypnotic Amnesia

(46.6% ± 4.2% and 74.8% ± 3.3%, respectively, p < 0.00005). lation analysis across groups, as well as inter- and intragroup
No such difference was shown for Context (74.2% ± 4.2% and analysis of BOLD signal (for the flowchart of analysis, see
82.9% ± 2.3%, respectively, p = 0.21; interaction effect: F1,22 = Figure S1 available online).
20.38, p < 0.0005). In contrast, after cancellation of FORGET, Overall Task-Correlated Brain Activity
memory performance was similar in both groups and on both Overall brain activity during Test 1 Movie compared to fixation
question types (Movie: Non-PHA = 80.42% ± 1.71%, PHA = baseline was obtained in each group separately in order to
79.6% ± 2.44%; Context: Non-PHA = 82.9% ± 2.34%, PHA = identify brain areas that participated in task processing. The
78.75% ± 3.9%; interaction effect: F1,22 = 0.66, p = 0.42). Non-PHA group exhibited a vast network of activated regions
Thus, the memory block induced by FORGET was specific to correlated with answering the questionnaire for Movie questions
movie details and reversible. (Figure 3A, top panel; Table S1). These included mainly visual
In order to examine whether the decreased memory perfor- processing regions, bilateral thalamus, basal ganglia, bilateral
mance in the PHA group was a result of demand characteristics superior frontal gyrus (SFG), and cerebellum. In contrast, the
(i.e., deliberately withholding the correct responses for movie PHA group exhibited activity only in a minor subset of these
details to comply with perceived test demands), the SHAM regions, namely bilateral occipital lobes, right SFG, cerebellum,
group replicated the experiment. Briefly, SHAM went through and insula (Figure 3A, bottom panel; Table S1). The reduced
the study and test session in the same manner as did the other overall activity in the PHA group suggests a general reduction
groups; however, prior to hypnosis, they received instructions in neural activity compared to Non-PHA while answering Test 1
to answer the questions during memory Test 1 (i.e., under active Movie questions, i.e., under FORGET condition. To examine
posthypnotic suggestion) as if they were affected by the post- whether the reduced activity in the PHA group represented
hypnotic suggestion. They were not, however, instructed in any a generalized suppression phenomenon throughout the experi-
way what strategy to use in order to mimic the amnesic effect. ment, activity was also determined while answering Context
Memory performance during FORGET for both Movie and questions versus baseline for each group. In contrast to Movie
Context in SHAM was lower than the PHA and Non-PHA (Movie: questions, the overall activity during Context questions versus
33.06% ± 5.1%, Context: 59.4% ± 5.9%; between-subject main baseline revealed in both groups several overlapping networks
effect: F2,30 = 14.8, p < 0.00005). Complementary, Scheffe post of activity, including visual sensory and perceptual regions, cer-
hoc comparisons of the group factor across question types ebellum, parietal lobes, SFG, and IFG (Figure 3B; Table S2). The
revealed significant differences among all groups, demonstrat- fact that both groups showed activity in these regions indicates
ing a general reduced memory performance in the SHAM group that the overall neural suppression in the PHA group was selec-
compared to both Non-PHA (p < 0.00005) and PHA groups (p < tive for the Movie information. We complemented this analysis by
0.05). Upon cancellation of suggestion, memory performance performing conjunction analyses between PHA and Non-PHA
was found to be similar to the other groups in both Movie and groups for Movie and for Context questions during FORGET. In
Context conditions (81.6% ± 2.1% and 76.1% ± 2.7%, respec- line with the aforementioned results, smaller overlap of activation
tively, F2,30 = 1.2, p = 0.31). Hence, SHAM showed significantly was found in Movie compared to that in Context (Figure S2).
reduced memory performance in Test 1 compared to the PHA Whole-Brain Correlation between Brain Activity
group; whereas the PHA group performed at a chance level and Memory Performance
(46.6% ± 4.2%), memory performance in the SHAM group We correlated memory performance scores for Movie and
dropped well below the chance level (33.1% ± 5.1%), suggesting beta values of the all-participant GLM during Test 1. Using
deliberate withholding of information. a voxel-by-voxel whole-brain correlation analysis of memory
Reaction Times performance and beta values of movie in Test 1 in all subjects
In Test 1, the PHA group exhibited increased reaction times on (r > 0.55, p < 0.01, uncorrected), we revealed activity in several
Movie questions compared to the Non-PHA group (4473 ± 257 regions (Figure 4; Table S3). The highest correlations were
versus 3879 ± 152 ms, respectively). Increased reaction times found in left middle temporal gyrus (x, y, z peak activity location
in the PHA group were observed for Context as well (3768 ± 55, 7, 16, BA 21, r = 0.64, p = 0.001), left superior temporal
212 versus 3260 ± 131 ms), resulting in a main effect for group gyrus ( 54, 14, 8, BA 38, r = 0.62, p = 0.002), and left middle
across question types (F1,22 = 4.7, p = 0.04). In contrast, after occipital gyrus ( 45, 76, 8, BA 19, r = 0.65, p = 0.001). Activity
reversal of FORGET (Test 2), reaction times did not differ for patterns exhibited a left occipito-temporal hemisphere network
both groups in Movie (Non-PHA = 2962 ± 145 ms, PHA = that was activated proportionally to the retrieval success of
2803 ± 183 ms) and Context (Non-PHA = 2567 ± 122 ms, Movie. Direct correlation between the mean beta values of these
PHA = 2416 ± 166 ms). In both Test phases, main effects were regions and memory performance were plotted (Figure 4B).
found for question type, exhibiting longer latencies for Movie Thus, it seems that the ROIs delineated by this analysis specifi-
questions than Context questions (question type main effects: cally show an activity gradient that is proportionate to retrieval
Test 1: F1,22 = 40.4, p = 0.000002; Test 2: F1,22 = 26.9, p = success.
0.00003). Between-Group Comparison
We compared brain activity between PHA and Non-PHA
Brain Activity subjects during retrieval of movie details in Test 1 using a GLM
We set out to identify the neural correlates of suppressed consisting of all participants. As depicted in Figure 5A (see
memory performance that is postulated to be guided by the also Table 1), Non-PHA had higher activity compared to PHA
posthypnotic FORGET suggestion, by using whole-brain corre- in several regions, including right fusiform area (54, 22, 23,

Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc. 161
 Neuron
Brain Correlates of Posthypnotic Amnesia

Figure 3. Brain Activity on Movie Questions

and Context Questions in Each Group
(A) BOLD response during Movie, Test 1 in Non-
PHA (top panel) and PHA (bottom panel) groups.
Statistical maps (radiological orientation) are
shown for Movie > baseline and are overlaid on
axial slices of the average anatomical scan of all
subjects (z coordinates indicated for each image).
Maps here and in (B) below were obtained with
a threshold of t > 6, p < 0.0001, cluster size >
150 mm3. Activity in the Non-PHA group is shown
in multiple regions, including bilateral cerebellum,
occipital lobes (BA 18), insula/inferior frontal gyrus
(IFG) (BA 13/45), medial superior frontal gyrus (BA
6), and precentral gyrus (BA 4). The PHA group
shows reduced activation; activity is in cerebel-
lum, bilateral occipital lobes (BA 18), left insula/
IFG (BA 13/45), and medial superior frontal gyrus
(BA 6).
(B) BOLD response during Test 1, Context > base-
line for Non-PHA (top panel) and PHA (bottom
panel) groups.

(51, 32, 28, BA 46). For the aforemen-

tioned ROIs, beta values of Movie from
both groups were analyzed in an ANOVA
that included group (PHA, Non-PHA) and
test (Test 1, Test 2) as factors. Interaction
effects were found in all ROIs, stemming
from elevated activation in the PHA group
in Test 2 compared to Test 1, whereas
Non-PHA estimates were unchanged
BA 20), left middle occipital gyrus ( 21, 85, 5, BA 18), and left between the scans (interaction effects of ROIs: F1,20, p = 5.9,
anterior superior temporal gyrus ( 48, 11, 5, BA 22). Higher ac- 0.025; 4.6, 0.04; 16.5, 0.0005, respectively; Figure 6B, right
tivity in PHA was observed in one location only, the left rostrolat- panels). It is noteworthy that with the threshold used, no clusters
eral PFC (30, 56, 6, middle frontal gyrus, BA 10). This is in line with were found to show higher activity in Test 1 compared to Test 2.
the whole-brain correlation unveiling differential activation in the Apparently, although PHA subjects were engaged in the same
left occipito-temporal hemisphere (see above). ROI analysis of retrieval task for the second time, they showed exclusively higher
correlations between cluster-average beta values from Movie, activity patterns during the second retrieval, i.e., following allevi-
Test 1 and memory performance for all participants during ation of the amnesic suggestion.
Movie, Test 1 was performed, revealing the following correla- In the Non-PHA group, the comparison between Test 1 and
tions (Figure 5C): right fusiform gyrus, r = 0.48 (p = 0.02); left mid- Test 2 revealed higher activity for Test 1 in left parahippocampal
dle occipital gyrus, r = 0.37 (p = 0.09); left inferior frontal gyrus, r = gyrus ( 24, 12, 14), left superior frontal gyrus in two locations
0.53 (p = 0.01); left rostrolateral PFC, r = 0.39 (p = 0.07). ( 3, 26, 49, BA 8; 9, 8, 61, BA 6), and left medial frontal gyrus
Intra-Group Comparisons ( 9, 50, 16, BA 10). Beta score ROI analysis of the delineated
To examine the neural dynamics in BOLD signal between Test 1 regions revealed interaction effects, resulting from decreased ac-
and Test 2 in each group, we compared Movie in Test 1 versus tivity for the Non-PHA group during Test 2 compared to Test 1,
Test 2, and Test 2 versus Test 1 for each group separately. We whereas no such decrease was revealed in the PHA group
hypothesized that suppression of memory observed for Movie (interaction effects of ROIs: F1,20, p = 7.8, 0.01; 8.8, 0.007; 5.8,
questions during Test 1 would be accompanied by reduced ac- 0.025, respectively; Figure 6A and Table 2). The opposite activity
tivity in the PHA group, as compared to the activity following al- pattern (i.e., Test 2 > Test 1) was revealed as well in several
leviation of amnesic suggestion. Indeed, in the PHA group, regions (Table 2), although not in the same areas as in the PHA
higher activation patterns were observed only for Test 2 com- group.
pared to Test 1, while no activity was revealed for Test 1 com-
pared to Test 2 (Table 2 and Figure 6). The clusters that showed DISCUSSION
the highest correlations with memory performance in a subse-
quent ROI analysis are delineated in Figure 6B and are found We used posthypnotic amnesia (PHA) to investigate brain corre-
around the right fusiform area (27, 75, 11, BA 19), left middle lates of episodic memory suppression. In brief, our results show
occipital gyrus (33, 82, 4, BA 18), and left middle frontal gyrus that (1) PHA of long-term, real-life-like memories is evident in

162 Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc.
Neuron
Brain Correlates of Posthypnotic Amnesia

Figure 4. Correlation of Memory Perfor-

mance and BOLD Signal
(A) Correlation maps overlaid on an average
anatomical brain for all subject (n = 22) between
memory performance (percentage of correct an-
swers) and beta values for Movie during Test 1.
Clusters are shown in axial slices, circling regions
of interest, from top to bottom: Left middle tempo-
ral gyrus, L MTG (x, y, z = 55, 7, 16), left supe-
rior temporal gyrus, L STG ( 54, 14, 8), and
middle occipital gyrus, L MOG ( 45, 76, 8).
Effects are significant at r > 0.55, p < 0.01, uncor-
rected, cluster size > 150 mm3.
(B) Correlation plots between memory perfor-
mance and beta values of Test 1, Movie.

suggests a strategy different from that

used by the PHA group, who showed
chance-level retrieval performance.
Moreover, SHAM revealed a reduction
in nontargeted memory items as well, im-
plying a generalization of the simulated
memory drop. These exaggerated and
generalized effects are congruent with
PHA-simulator results in previous studies
(Williamsen et al., 1965; Kihlstrom, 1985),
suggesting that the PHA cannot be attrib-
uted merely to demand characteristics
(but see Wagstaff et al., 2001).
susceptible individuals in a controlled fMRI environment. The de- The brain regions that display above-baseline activity in the
crease in memory performance affects FORGET-targeted items Non-PHA group in Test 1 correspond to regions that were previ-
while sparing contextual memory. (2) PHA is correlated with re- ously reported to subserve declarative retrieval and attention
duced activity in multiple brain areas, particularly in the left ex- (Cabeza and Nyberg, 2000; Naghavi and Nyberg, 2005). In the
trastriate occipital lobe and the left temporal pole. In contrast, in- same test, only a small subset of regions was activated in the
creased activation is noticed in left rostrolateral prefrontal PHA group on Movie questions. These regions might represent
cortex. (3) Following reversal of the FORGET suggestion and re- a minimal sensory, cognitive, and motor network required to per-
covery of normal memory performance, increased activity is ob- form the behavioral task in the scanner. The elevated activity in
served in multiple areas, including occipital, parietal, and dorso- the brain of the PHA participants in comparison to baseline activ-
lateral frontal regions. ity on the Context questions under the same conditions only
That the PHA group exhibited reversible reduction of memory highlights the specificity of suppression of performance on the
performance under the control of the posthypnotic FORGET FORGET-oriented memory items. It is noteworthy that hippo-
suggestion is in line with previous reports of reversible retrieval campus and certain related limbic structures, known to subserve
block in PHA. The memoranda targeted to be forgotten in previ- declarative memory encoding and retrieval, did not display
ous studies were typically the hypnosis session itself (Evans, above-baseline activation in either of the groups in our analysis.
1988; Kihlstrom, 1997), word lists (Barnier et al., 2001; Bryant We considered the possibility that this is because these circuits
et al., 1999; David et al., 2000), or autobiographical events (Bar- were more active during rest compared with task periods (Stark
nier, 2002; Cox and Barnier, 2003). To the best of our knowledge, and Squire, 2001; Svoboda et al., 2006). However, we didn’t
this is the first PHA study to use controlled, extended real-life- observe higher hippocampal activation during baseline in com-
like memoranda, encoded well before the hypnosis session. paring baseline to Movie (unpublished data). Further analyses
A potential drawback of hypnosis studies in general and PHA using less stringent statistical thresholds and focusing on prese-
paradigms in particular is the risk of demand characteristics lected anatomical ROIs might be required to further determine
(Hilgard, 1975). It has been argued that the effect observed in the role of hippocampus and related limbic circuits, as well as
PHA merely expresses subjects’ wish to comply with the per- additional brain circuits, in our paradigm.
ceived task demands by intentionally withholding information Correlation of brain activity with memory performance in all the
(Coe et al., 1989). We approached this issue by examining a group participants, as well as the PHA-NonPHA groups comparison,
of low-suggestibility participants, SHAM, who were instructed revealed regions associated with the FORGET suggestion. Activ-
before the hypnosis to simulate PHA. The fact that SHAM ity in the left middle occipital gyrus was significantly reduced
displayed an exaggerated decrease in memory performance during FORGET in the PHA group. Furthermore, activity in that

Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc. 163
 Neuron
Brain Correlates of Posthypnotic Amnesia

Figure 5. Between-Group Comparisons on Movie Questions during Test 1

(A) Between-group statistical maps for Movie, Test 1 (t > 3.2, p < 0.005, uncorrected, cluster size > 150 mm3). BOLD activity is shown in axial slices. Encircled are
the right fusiform gyrus, R FFG (54, 22, 23), left middle occipital gyrus, L MOG ( 21, 85, 5), left superior temporal gyrus, L STG ( 48, 11, 5), and left middle
frontal gyrus, L MFG ( 30, 56, 6; left rostrolateral PFC).
(B) Plot of mean beta values for PHA (black) and Non-PHA (gray) for the ROIs depicted in (A). Values of t and p, from left to right, respectively: 3.6, 0.001; 3.8, 0.001;
3.9, 0.0007; 3.7, 0.001. Error bars are SEM.
(C) Beta values for Test 1, Movie for the respective ROIs correlated with memory performance for all subjects. Values of r and p are, from left to right, respectively:
0.48, 0.02; 0.37, 0.09; 0.53, 0.01; 0.39, 0.07.

area was significantly correlated with memory performance. representations that were active in encoding (e.g., Morris et al.,
Occipital activation is commonly detected in retrieval of nonver- 1977; Tulving, 1983; Polyn et al., 2005; Johnson and Rugg,
bal material (Cabeza and Nyberg, 2000). Theory and data both 2007). For example, Johnson and Rugg (2007) report that recol-
point to reactivation or reconstruction in retrieval of types of lection of scenes but not verbal information activates occipital
regions that were activated in encoding of that specific stimuli
type. Similarly, Vaidya et al. (2002) show that the middle occipital
Table 1. Regions Showing Differences between Non-PHA gyrus is activated in recognition of words that served as cues for
and PHA in Test 1 encoded pictures but not for other words. It is therefore plausible
Non-PHA > PHA to assume that reduced activity in middle occipital gyrus during
Region x y z mm3 t Value p Value FORGET represents suppressed reinstatement of memory
L middle 21 85 5 245 4.32 0.0003 scene traces.
occipital gyrus (BA 18) The left temporal pole (BA 38 and anterior BA 22) showed
R fusifirm 54 22 23 382 5.48 0.00002 similar activity patterns to those of the occipital lobe, both in
gyrus (BA 20) correlations of brain activation with memory performance and
L superior 48 11 5 719 4.42 0.0002 in intergroup comparison of Movie questions during FORGET.
temporal gyrus (BA 22) The temporal pole is considered an association cortex based
L postcentral 39 22 52 512 3.9 0.0008
on its connectivity with multiple sensory systems and its activity
gyrus (BA 3) in response to both visual and auditory stimuli (Olson et al.,
R claustrum 33 14 4 342 4.02 0.0006
2007). It was implicated in emotional and social processing,
theory of mind, real-life memory, and formation of narratives
PHA > Non-PHA
from spoken sentences (Maguire et al., 1999; Maguire and
L middle 30 56 6 678 3.43 0.002
Mummery, 1999; Graham et al., 2003; Gallagher and Frith,
frontal gyrus (BA 10)a
a
2004; Olson et al., 2007). It fits hence to subserve retrieval of
This area is referred to in the text as L rostrolateral PFC.
the socially and narrative-embedded audiovisual information

164 Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc.
Neuron
Brain Correlates of Posthypnotic Amnesia

Table 2. Regions Showing Intragroup Differences between Tests attention between environmental stimuli and self-generated
representations. We suggest that the increased activation of ros-
Non-PHA: Test 1 > Test 2
trolateral PFC in the PHA group during FORGET reflects an early
Region x y z mm3 t Value p Value
implicit decision on whether or not to trigger further retrieval pro-
L parahippocampal gyrus 24 12 14 1077 5.36 0.0003
cesses, taken on the basis of the correspondence of the external
L middle 9 50 16 365 4.55 0.001 cue to the internal representation of the FORGET suggestion. We
frontal gyrus (BA 10) propose to dub the stage in which this early decision is taken as
L superior 3 26 49 346 4.74 0.0007 ‘‘preretrieval monitoring,’’ because the initiation of the retrieval
frontal gyrus (BA 8) cascade might be abated.
L superior 9 8 61 422 4.85 0.0006 The possibility could be raised that activation of rostrolateral
frontal gyrus (BA 6) PFC in memory suppression on Movie in PHA under FORGET
Non-PHA: Test 2 > Test 1 reflects increased retrieval effort. The identity of brain substrates
R inferior 30 91 14 188 4.93 0.0005 of retrieval effort has yet to be clarified (Rugg and Wilding, 2000),
occipital gyrus (BA 18) and though some studies did suggest BA 10 to be involved
R lingual 24 91 2 156 4.39 0.001 (Schacter et al., 1996), others specifically implicate other PFC
gyrus (BA 18) regions (Buckner et al., 1998; Heckers et al., 1998; Sohn et al.,
R precuneus (BA 7) 9 70 37 349 6.44 0.00007 2003). We have attempted to tap into potential substrates of
L precuneus (BA 7) 9 70 40 301 4.64 0.0009 retrieval effort in our protocol by postulating that in the control
R superior 36 53 31 254 4.74 0.0007 subjects, the longer the RT on a task, the more effortful the
frontal gyrus (BA 9) retrieval (Buckner et al., 1998). We hence contrasted brain activ-
L white matter 30 43 7 741 5.94 0.0001
ity for incorrect (longer RT) and correct (shorter RT) answers in
Non-PHA on Movie in Test 1, and identified activation in left
PHA: Test 2 > Test 1
superior frontal and right medial frontal gyri (BA 9), but not in
R middle 33 82 4 861 6.05 0.0001
rostrolateral PFC (Figure S3). Taken together, we therefore
occipital gyrus (BA 18)
deem less likely the possibility that rostrolateral PFC activation
L middle 27 82 7 316 4.95 0.0005
in our study reflects increased retrieval effort rather than prere-
occipital gyrus (BA 18)
trieval monitoring. Brain imaging methods with higher temporal
R fusiform 27 75 11 1848 3.67 0.004 resolution, i.e., EEG and MEG, might be useful in clarifying this
gyrus (BA 19)
issue further.
L cuneus (BA 23) 12 70 10 184 4.57 0.001 The differences in brain activity patterns between Test 1 (i.e.,
R inferior 33 58 40 877 5.95 0.0001 FORGET) and Test 2 (i.e., FORGET Reversed) were dissimilar for
parietal lobule (BA 39) each of the groups. Whereas Non-PHA participants showed
R precuneus (BA 7) 24 76 46 499 5.22 0.0003 both reduction and enhancement of activity following FORGET
R middle 51 32 28 432 5.12 0.0004 cancellation, PHA showed practically only enhancement follow-
frontal gyrus (BA 46) ing FORGET cancellation. This enhancement in Test 2 contrasts
L middle 33 1 46 263 5.45 0.0002 with the widely reported phenomenon of repetition suppression
frontal gyrus (BA 6) in subsequent tests (e.g., Henson and Rugg, 2003; Schacter
L superior 12 44 55 246 5.78 0.0001 and Buckner, 1998). That repetition suppression effects were
frontal gyrus (BA 8) not observed for the PHA group in Test 2 is in line with the sup-
R cerebellum 6 67 35 694 4.3 0.001 pression observed in this group during Test 1. The brain regions
L brainstem 3 28 5 316 5.79 0.0001 that were activated preferentially upon reversal of the FORGET
suggestion reveal a network of regions that has been docu-
mented in the literature in long-term memory retrieval (Svoboda
encoded during movie viewing. Indeed, in a recent study of sub- et al., 2006; Yancey and Phelps, 2001; Cabeza and Nyberg,
sequent memory for movie, activations were found in the right 2000). The areas in which recovery of activation was observed
temporal pole during encoding of subsequently remembered in Test 2 for PHA complement the areas in which activity was
items (Hasson et al., 2008). suppressed, in comparison with Non-PHA, in Test 1. The paral-
In contrast to the aforementioned regions, the left rostrolateral leled recovery of brain activity and memory performance
prefrontal cortex (PFC), displayed preferential activity during strongly suggests that suppression was exerted at early stages
suppression of memory performance. The engagement of PFC of the retrieval process, thus preventing the activation of regions
in retrieval of declarative long-term memory is proposed to be that are crucial for productive retrieval. This hence is congruent
associated with content-invariant retrieval mode rather than with our aforementioned proposal that PHA under FORGET
with content-specific ecphory (Lepage et al., 2000). The rostro- affects an executive preretrieval monitoring process, which
lateral PFC has been specifically implicated in meta-processes produces an early decision on whether to proceed or not on re-
and executive functions engaged in retrieval of episodic memory trieval, and in case of a Movie question, aborts the process.
(Nyberg et al., 2000; Gilbert et al., 2006; Moscovitch and Wino- Such preretrieval implicit pondering could be in line with, though
cur, 2002). Burgess et al. (2007) propose that rostral PFC is clearly not proven by, the prolonged reaction times on both
a ‘‘gateway’’ linking the outside and inside world, switching FORGET-targeted and untargeted items in the PHA group.

Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc. 165
 Neuron
Brain Correlates of Posthypnotic Amnesia

Figure 6. Within Group Comparison on Movie questions in Test 1 versus Test 2

(A) Statistical maps depicting voxels different between Test 1, Movie and Test 2, Movie in Non-PHA. Maps here and below were obtained with a threshold of t >
3.6, p < 0.005, uncorrected, cluster size > 150 mm3. Encircled are the left parahippocampal gyrus, L PHG ( 24, 12, 14), and left superior frontal gyrus, L SFG
(BA 8, 3, 26, 49, and BA 6, 9, 8, 61). The corresponding beta values for Test 1, Movie (left pair of bars) and Test 2 (right pair of bars) are plotted for Non-PHA
(gray) and PHA (black).
(B) Maps of voxels different between Test 1, Movie and Test 2, Movie in PHA. Encircled are the right fusiform area, R FFG (BA 19, 27, 75, 11), right middle
occipital gyrus, R MOG (BA 18, 33, 82, 4), and left middle frontal gyrus, L MFG (BA 6, 33, 1, 46). The corresponding beta values for Test 1, Movie (left
pair of bars) and Test 2 (right pair bars) are plotted for Non-PHA (gray) and PHA (black).
Error bars are SEM.

The postulated preretrieval monitoring is a top-down process. propose, on the bases of our data, that PHA abates a very early
Top-down mechanisms, which enable the allocation of attention stage. This probably differs from some other paradigms of
to relevant stimuli while ignoring irrelevant ones (Gazzaley et al., memory suppression.
2005), have been proposed to play a key role in behavioral Influential experimental paradigms have been developed to
manifestations of hypnosis that involve suppression or modula- investigate memory inhibition and suppression. In retrieval-
tion of sensory input (Raz et al., 2006). In the present paradigm, induced forgetting, retrieving exemplars from a set of learned
bottom-up sensory input is held constant in both Test 1 and 2 items in a category was shown to inhibit retrieval of other, non-
and only task demands are altered. Hence, even if only task practiced exemplars (Anderson et al., 1994). In the think/no think
constraints are taken as a guide, interpretation of the etiology paradigm, cueing to intentionally reject thinking about a paired
of the memory suppression in terms of top-down modulation is associate was shown to ultimately suppress retrieval of that
indeed reasonable. specific association (Anderson and Green, 2001, but see Bule-
How do our findings correspond to previous data on memory vich et al., 2006). Neuroimaging studies using the think/no think
suppression? It should be stated at the outset of this comparison paradigm implicate in memory suppression activation of regions
that the term ‘‘suppression’’ is used in the literature in different in the dorsolateral prefrontal cortex (DLPFC) and attenuation of
connotations, ranging from suppression that is assumed to oc- hippocampal activation (Anderson et al., 2004). Two procedural
cur during ongoing normal retrieval, to suppression of unwanted attributes of the think/no think paradigm should be particularly
memories as construed within the conceptual framework of psy- noted. First, participants are well trained, and second,
chiatry, to assumed suppression of proper retrieval in certain suppression is exerted on memory immediately after the study
mnemonic pathologies. Sometimes it is equated or paralleled phase. This should place high demands on working memory,
with the broad usage of ‘‘inhibition’’ in memory research hence the activation of PFC. In contrast, in our paradigm, com-
(Roediger et al., 2007). Hence, one should note the conceptual plex memory items are taxed a week after their encoding. This
framework that is explicitly or implicitly used in attempts to iden- is expected to tax working memory less.
tify brain substrates of memory suppression. Furthermore, par- Memory suppression has been also proposed to dominate
ticularly pertinent to comparison among studies of different certain pathologies in which the suppression mechanisms may
manifestations of suppression is the question at which time in not necessarily mimic or exacerbate suppression that occurs
the retrieval process memory is assumed to become sup- in normal retrieval. Such a pathology, by definition, is psycho-
pressed. Retrieval is a multistage process (Rugg and Wilding, genic or functional amnesia (Markowitsch, 1999). PHA has
2000; Sakai, 2003; Gardiner, 2007). As noted above, we been specifically suggested as an experimental model for

166 Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc.
Neuron
Brain Correlates of Posthypnotic Amnesia

functional amnesia (Barnier, 2002). Neuroimaging studies of In the individual PHA screening, each participant was instructed to read
functional amnesia are rare. PET studies have indicated both a short story thoroughly. Next, the participants were induced into a hypnotic
state, using relaxation techniques, in the same manner as in the group session.
reduction (Markowitsch, 2003) and enhancement (Yasuno
Approaching the dehypnotization stage of the hypnotic procedure (i.e., termi-
et al., 2000; Fink et al., 1996) in fronto-temporal regions when nation of hypnotic state), approximately 20 min after reading the story, a sug-
tested for recollection of apparently forgotten memory. In an gestion to forget the story details was conveyed by the hypnotist, along with
fMRI study of a person suffering from functional amnesia for a reversibility cue, designed to cancel in due time the forgetting suggestion
his native language and autobiographical memories, reduced (see Supplemental Data). Following complete dehypnotization of the hypnotic
frontal activity compared to controls was unveiled on working state, a short memory pen-and-paper questionnaire was administered, con-
taining 13 yes/no questions regarding the story. Upon completion of the ques-
memory and lexical tasks involving the native language (Glisky
tionnaire, the reversibility cue was provided, followed by administration of the
et al., 2004). Although our data point to altered activity in
same questionnaire again. The number of changed answers from the first to
fronto-temporal regions as well, additional combined neuropsy- the second test was summed up for each participant, and the median score
chological and functional neuroimaging research is needed to was then used to classify subjects as posthypnotic amnesia (PHA) subjects
delineate the role of identified brain circuits in functional amnesia or Non-PHA subjects.
that presents in the clinic. We postulate, however, that other
forms of functional amnesia may also be a consequence of Experimental Protocol
retrieval abortion at a preretrieval monitoring stage and, there- The protocol included a STUDY session and a TEST session. The STUDY was
performed out of the magnet and the TEST in the magnet. The manipulation of
fore, may indeed be modeled at least partially by PHA.
memory by hypnosis was performed in the TEST.
All in all, our data identify brain circuits that subserve suppres- STUDY Session
sion of retrieval of long-term memory of a real-life-like extended The study material was a 45 min movie, produced and filmed in-house specif-
episode in the course of posthypnotic FORGET suggestion. ically to serve in real-life-like memory studies. The movie was a documentary
Some of these regions are likely to play a role in normal retrieval. depicting a routine day in the life of a young Israeli woman. It depicted both
Others are likely to be engaged in dysfunctions that involve an mundane activities such as preparing meals and talking on the phone, along
with potentially more interesting events, such as rehearsing for a play, teaching
executive decision to abort subsequent retrieval.
a drama class, and riding rollerblades with friends (Figure 1A). The movie was
viewed in a quiet room on a standard 17 inch computer monitor, with sound
EXPERIMENTAL PROCEDURES delivered through a headphone set. Participants were given written instruc-
tions that they were about to watch a 45 min movie and that their only task
Participants is to try to concentrate throughout. They were not specifically instructed to
One hundred and thirty-seven volunteers were recruited from the Weizmann remember the movie details, and were not told they were going to be tested.
Institute of Science and the Faculty of Agriculture of the Hebrew University, TEST Session
Rehovot. The experimental protocol was approved by the Institutional Review Based on a prior study from our lab on long-term memory of cinematic material
Board (IRB) of the Sourasky Medical Center, Tel-Aviv, at which the fMRI scan- (Furman et al., 2007), showing high memory performance a week after learn-
ning was carried out, and approval of the use of hypnosis was given by the ing, retrieval was assessed 1 week after viewing the movie. Prior to scanning,
Division of Medical Professions, Ministry of Health, Jerusalem. All the partici- the participants signed informed consent and MRI safety forms.
pants were native Hebrew speakers. They were given the hypnosis suscepti- After entering the scanner, participants lay passively in the absence of scan-
bility test in groups (see below). Of these, 46 individuals who passed the ning and were induced into a hypnotic state through instructions conveyed by
predefined hypnotizability criterion were examined individually for their capac- the hypnotist via the magnet’s headset system. The induction into the hypnotic
ity to sustain posthypnotic amnesia (see below). On the basis of the posthyp- state lasted approximately 10 min and was performed by standard relaxation
notic test score, subjects were labeled as susceptible to posthypnotic techniques. Toward the dehypnotization of the subject from the hypnotic state,
amnesia (PHA) or not susceptible (Non-PHA). Ultimately, 25 individuals the hypnotist presented the FORGET suggestion, which conveyed the instruc-
(25.8 ± 2.3 years, 17 female, 12 PHA) proceeded to participate in the experi- tion to forget the movie viewed during the study session, and a reversibility
ment. Twenty-three performed the experiment in the MRI scanner (11 PHA) cue, intended to reverse FORGET at due time. Participants were then de-in-
and two (1 PHA) were not tested in the scanner because of metal teeth braces duced from the hypnotic state, followed by a memory test, administered while
and carried out the experiment outside the scanner. One subject from the Non- their brain was scanned (Figure 1B; Test 1, FORGET). Immediately following
PHA group was later excluded from the analysis due to reading disabilities. In Test 1, the hypnotist cancelled FORGET by administering the reversibility
addition, nine subjects from the original volunteer pool who did not pass the cue, followed by a second scanning of the same memory test consisting of
hypnotizability criterion served as a PHA SHAM group and performed the ex- the same questions (Test 2, FORGET Reversed). The procedure was
periment outside the scanner. performed for participants in both PHA and Non-PHA groups.
The memory test was a computerized questionnaire (delivered on Presenta-
tion software, Neurobehavioral Systems, San Francisco, CA, version 10.3). It
Screening for PHA Susceptibility consisted of 40 questions about items from the movie (Movie questions) and
Groups of 5 to 20 volunteers were presented with a 40 min lecture on the 20 questions about contextual details from the STUDY session (Context ques-
nature of hypnosis, given by a certified M.D., who later performed the hypnosis tions). Movie questions targeted details from the movie itself, e.g., The actress
procedure (Y.C.). Following the lecture, subjects underwent a 15 min hypnotic knocked on her neighbor’s door on the way home (YES/NO). Context questions
assessment procedure, using standard relaxation techniques for hypnosis were designed to serve as a control for memory items that were not suggested
induction followed by five hypnotic suggestions adopted from the Stanford to be forgotten posthypnotically, e.g., During the movie, the door to the room
Hypnotic Susceptibility scale (Weitzenhoffer and Hilgard, 1962; Lichtenberg was closed (YES/NO). Questions were constructed in the form of short senten-
et al., 2004) and the Hypnosis Induction Profile (HIP; Spiegel and Spiegel, ces, of which half were true and half false. They were presented for 6 s each,
2004). The hypnotic suggestions included arm levitation (item E, HIP), arm showing simultaneously the question and the YES/NO options on the screen.
immobilization (item 8, Stanford scale), somatoseneory mosquito hallucina- Once an answer was given, the relevant label on screen (i.e., YES or NO)
tion, auditory mosquito hallucination, and visual mosquito hallucination (based changed its color from orange to green, thus providing response feedback.
on item 3, Stanford scale). Volunteers who exhibited successful performance After answering the questions, the text remained on screen for the remainder
on a minimum of three suggestions proceeded to undergo individual PHA of the trial time (up to 6 s). The trial was completed by a 4–8 s blank event. The
screening. blank events included a fixating cross in the center of the screen. Answers and

Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc. 167
 Neuron
Brain Correlates of Posthypnotic Amnesia

reaction times were recorded onto a log file, enabling computation of correct each question type (Movie questions and Context questions) in each scan
response percentage and mean reaction time of each participant. (Test 1 and Test 2). GLM 1 included subjects from both groups and was
The experiments were presented on a PC in the magnet console room via an used for correlation analysis and intergroup comparisons, while GLM 2 and
LCD projector, which projected onto a screen behind the subjects. The ques- 3 consisted of separate-group models for testing intragroup effects between
tions were viewed through a mirror mounted on the head coil, and answers Test 1 and Test 2 (Figure S1). Time periods between questions were consid-
were executed by pressing a four-button digital response box. Each retrieval ered as baseline. Trial lengths were considered as the time between text onset
test lasted 12.5 min. Head pads were placed around the head to reduce and subjects’ response, while the remainder of the event (from answer until
head movements, and ear plugs were given to subjects for noise protection. completion of 6 s) was defined as a separate condition. The condition time
PHA SHAM Group vectors were convolved with a canonical hemodynamic response function
The SHAM group performed the STUDY in the same manner as the other (HRF). Six head motion parameters were inserted in the GLM as covariate
participants, while TEST was performed outside the scanner. In TEST, prior regressors (three translation and three rotation parameters).
to the beginning of the hypnosis procedure, they were given written instruc- Overall Task-Correlated Activity
tions (Supplemental Data) explaining that they were a control group in Activity patterns for Movie and Context during Test 1 were determined by
a PHA experiment. Specifically, they were told that they were about to receive producing statistical maps for each group separately and for each of the two
a suggestion to forget certain information for a limited time period, until a re- conditions against the fixation baseline throughout the scan. The statistical
versibility cue is provided. In addition, they were instructed to act as if the sug- maps were thresholded at p < 0.0001, with a minimum cluster size of
gestion affected them, even if that feeling was not genuine. Following the 150 mm3 . The maps for each group were overlaid on the average anatomical
instructions, participants were induced into a hypnotic state and were given brain of the 22 subjects, depicting activity on five axial slices for each group in
the FORGET suggestion as the other groups. Their task was to answer the each condition. The activation loci were collected for each group and were
memory questionnaire in Test 1 as if they were under suggestion, and in summarized in tables, providing the center of mass Talairach coordinates of
Test 2 as if the suggestion was reversed. The memory questionnaires were each cluster.
delivered via a standard PC on a 17 inch screen. Memory-Correlated Activity
In order to identify brain regions that are related to memory performance,
fMRI Acquisition a whole-brain voxel-by-voxel correlation between percentage of correct
Imaging was performed on a 3T GE Signa Horizon echo speed scanner answers and BOLD measurements was computed for all the subjects com-
(Milwaukee, WI) with a resonant gradient echoplanar imaging system. All bined. Memory performance score per subject in Movie questions in Test 1
images were acquired using a standard quadrature head coil. The scanning was treated as a covariate and was correlated with beta values from GLM 1
session included anatomical and functional imaging. 3D sequence spoiled (see above) for Movie > Baseline, Test 1. Thus, an activity map for all partici-
gradient (SPGR) echo sequence, with high-resolution 1 mm slice thickness pants was obtained, revealing voxels that were significantly correlated with
(FOV = 24 3 24, matrix = 256 3 256, TR/TE = 40/9 ms) was acquired for memory performance. The threshold used for obtaining the statistical map
each subject. This anatomical scan allowed for volume statistical analyses was set at r > 0.55, p < 0.01, uncorrected, with cluster size of a minimum of
of signal changes during the experiment. In addition, T2 and FLAIR weighted 150 mm3. For extracting ROIs that were particularly correlated with behavior,
scans were acquired as part of the clinical protocol of the imaging facility. for each cluster, the average beta value for each participant was extracted and
For the BOLD scanning, T2*-weighted images (TR/TE/Flip angle = 2000/40/ correlated with memory performance for the whole group. The three clusters
80, FOV = 20 3 20 cm2, matrix size = 64 3 64) were acquired (32 oblique slices, with the highest correlation values were plotted in a graph depicting the rela-
15 toward coronal plane from ACPC, thickness 4 mm, gap 0 mm, covering the tion between memory performance and beta values in those regions
whole cerebrum) in runs of 12,000 images (375 images per slice). (Figure 4B).
Between-Group Comparison
Behavioral Analysis Statistical maps were generated by performing a random-effects two-sample t
Memory performance was calculated separately for each condition type (i.e., test contrast, comparing Movie questions in Test 1 between the two groups
Movie and Context) for each of the two memory tests in the TEST session (i.e., (using GLM 1, see above). Significance was tested at p < 0.005, uncorrected,
Test 1 and Test 2), by calculating the percentage of correct responses for each and with cluster size of at least 150 mm3. Several of the resulting regions were
participant. Memory performance values were then transformed by using the selected for a region of interest analysis (see Results). For each cluster, the
arc-sine square-root transformation. The transformed scores were analyzed mean beta value across voxels for each subject was calculated for PHA and
with mixed-model analyses of variance (ANOVAs) for each test separately, Non-PHA groups separately and plotted. Separate t tests were performed
with question type (Movie, Context) as the within-subject factor and group on the mean beta values of each ROI between the groups for specific estima-
(PHA, Non-PHA) as the between-subject factor. Reaction times (RT) of answer tion of effect. In addition, mean beta values of Movie, Test 1 from the selected
latency throughout the memory test sessions were analyzed by calculating the ROIs were plotted against Memory performance in Movie, Test 1 for all
mean RT of individual subjects for each question type in each of the tests. subjects. r and p values of each correlation were reported.
Mean RTs were inserted into a mixed-model ANOVA for each test separately Test 1 versus Test 2
in the same manner as for the memory performance scores. Using the separate GLMs for each group (GLM 2 and 3) with the same condi-
tions as described above, random-effects analysis was carried out for each
fMRI Analysis group, comparing by a one-sample t test Movie questions between the two
Preprocessing and data analysis were performed using BrainVoyager QX 1.8 scans (Test 1 > Test 2 and vice versa). Statistical maps were obtained by
(Brain Innovation, Maastricht, Netherlands). Functional images were corrected this contrast, using a threshold of p < 0.005, uncorrected, with cluster size
for slice timing, head movements, and linear drifts. Low frequencies were of at least 150 mm3. ROIs were chosen from these maps on the basis of highest
filtered out from the data. Images were spatially smoothed using a 6 mm correlations between beta value of memory Test 1 minus memory Test 2
full-width at half-maximum (FWHM) Gaussian kernel. The first seven volumes (Movie condition) and memory performance of each participant. For those re-
(14 s) from the beginning of each scan were removed from the data set to allow gions, mean beta values of Movie questions for each group (Test 1, Test 2),
for signal equilibrium. Functional and anatomical scans were spatially normal- were analyzed with mixed-model ANOVAs with test (Test 1, Test 2) as the
ized by extrapolation into a 3D volume in Talairach space (Talairach and Tour- within-subject factor and group (PHA, Non-PHA) as the between-subject
noux, 1988). Functional scans were superimposed onto the 3D high-resolution factor.
SPGR volume set and were interpolated into the same resolution as the SPGR
anatomical scans (voxel size: 1 3 1 3 1 mm).
Preprocessed time series data for each individual scan were analyzed with Supplemental Data
multiple regression. Three General Linear Models (GLM’s) were specified to The Supplemental Data for this article can be found online at http://www.
investigate the conditions of interest, generating separate regressors for neuron.org/cgi/content/full/57/1/159/DC1/.

168 Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc.
Neuron
Brain Correlates of Posthypnotic Amnesia

ACKNOWLEDGMENTS Fink, G.R., Markowitsch, H.J., Reinkemeier, M., Bruckbauer, T., Kessler, J.,
and Heiss, W.D. (1996). Cerebral representation of one’s own past: neural
We are grateful to O. Furman, E. Furst, R. Ludmer, M. Moscovitch, U. Nili, S. networks involved in autobiographical memory. J. Neurosci. 16, 4275–4282.
Rosenbaum, T. Sharot, and G. Winocur for discussions and P. Lichtenberg Furman, O., Dorfman, N., Hasson, U., Davachi, L., and Dudai, Y. (2007). They
for the Hebrew version of the Stanford Scale for Hypnotic Susceptibility. saw a movie: long-term memory for an extended audiovisual narrative. Learn.
This work was supported by grants to Y.D. from The Minerva Foundation Mem. 14, 457–467.
and The Nella and Leon Benoziyo Center for Neurological Diseases.
Gallagher, H.L., and Frith, C.D. (2004). Dissociable neural pathways for the
perception and recognition of expressive and instrumental gestures. Neuro-
Received: August 24, 2007
psychologia 42, 1725–1736.
Revised: October 26, 2007
Gardiner, J.M. (2007). Retrieval: on its essence and related concepts. In Mem-
Accepted: November 12, 2007
ory Concepts, H.L. Roediger, Y. Dudai, and S.M. Fitzpatrick, eds. (New York:
Published: January 9, 2008
Oxford University Press), pp. 221–224.
Gauld, A. (1995). A History of Hypnotism (Cambridge: Cambridge University
REFERENCES Press).
Gazzaley, A., Cooney, J.W., Rissman, J., and D’Esposito, M. (2005). Top-down
Anderson, M.C., and Green, C. (2001). Suppressing unwanted memories by
suppression deficit underlies working memory impairment in normal aging.
executive control. Nature 410, 366–369.
Nat. Neurosci. 10, 1298–1300.
Anderson, M.C., Bjork, R.A., and Bjork, E.L. (1994). Remembering can cause
Gilbert, S.J., Spengler, S., Simons, J.S., Steele, J.D., Lawrie, S.M., Frith, C.D.,
forgetting: retrieval dynamics in long-term memory. J. Exp. Psychol. Learn.
and Burgess, P.W. (2006). Functional specialization within rostral prefrontal
Mem. Cogn. 20, 1063–1087.
cortex (area10): a meta-analysis. J. Cogn. Neurosci. 18, 932–948.
Anderson, M.C., Ochsner, K.N., Kuhl, B., Cooper, J., Robertson, E., Gabrieli, Gilboa, A., Alain, C., Stuss, D.T., Melo, B., Miller, S., and Moscovitch, M.
S.W., Glover, G.H., and Gabrieli, J.D. (2004). Neural systems underlying the (2006). Mechanisms of spontaneous confabulations: a strategic retrieval
suppression of unwanted memories. Science 303, 232–235. account. Brain 129, 1399–1414.
Barnier, A.J. (2002). Posthypnotic amnesia for autobiographical episodes: Glisky, E.L., Ryan, L., Reminger, S., Hardt, O., Hayes, S.M., and Hupbach, A.
a laboratory model of functional amnesia? Psychol. Sci. 13, 232–237. (2004). A case of psychogenic fugue: I understand, aber ich verstehe nichts.
Barnier, A.J., Bryant, R.A., and Briscoe, S. (2001). Posthypnotic amnesia for Neuropsychologia 42, 1132–1147.
material learned before or during hypnosis: explicit and implicit memory Graham, K.S., Lee, A.C.H., Brett, M., and Patterson, K. (2003). The neural basis
effects. Int. J. Clin. Exp. Hypn. 49, 286–304. of autobiographical and semantic memory: New evidence from three PET
Bjork, R.A. (2007). Inhibition: an essential and contentious concept. In Memory studies. Cogn. Affect. Behav. Neurosci. 3, 234–254.
Concepts, H.L. Roediger, Y. Dudai, and S.M. Fitzpatrick, eds. (New York: Ox- Hasher, L., and Zacks, R.T. (1988). Working memory, comprehension, and ag-
ford University Press), pp. 307–313. ing: A review and a new view. In The Psychology of Learning and Motivation,
Bjork, R.A., LaBerge, D., and LeGrande, R. (1968). The modification of short- G.H. Bower, ed. (New York, NY: Academic Press), pp. 193–225.
term memory through instructions to forget. Psychon. Sci. 10, 55–56. Hasson, U., Furman, O., Clark, D., Dudai, Y., and Davachi, L. (2008). Enhanced
intersubject correlations during movie viewing correlate with successful
Braid, J. (1845). Mr. Braid on hypnotism. Lancet 1, 627–628.
episodic encoding. Neuron 57, in press.
Bryant, R.A., Barnier, A.J., Mallard, D., and Tibbits, R. (1999). Posthypnotic
Heckers, S., Rauch, S.L., Goff, D., Savage, C.R., Schacter, D.L., Fischman,
amnesia for material learned before hypnosis. Int. J. Clin. Exp. Hypn. 47,
A.J., and Alpert, N.M. (1998). Impaired recruitment of the hippocampus during
46–64.
conscious recollection in schizophrenia. Nat. Neurosci. 1, 318–323.
Buckner, R.L., Koutstaal, W., Schacter, D.L., Wagner, A.D., and Rosen, B.R. Henson, R.N., and Rugg, M.D. (2003). Neural response suppression, haemo-
(1998). Functional-anatomical study of episodic retrieval using fMRI. I. dynamic repetition effects, and behavioural priming. Neuropsychologia 41,
Retrieval effort versus retrieval success. Neuroimage 7, 151–162. 263–270.
Bulevich, J.B., Roediger, H.L., 3rd, Balota, D.A., and Butler, A.C. (2006). Hilgard, E.R. (1975). Hypnosis. Annu. Rev. Psychol. 26, 19–44.
Failures to find suppression of episodic memories in the think/no-think para-
Johnson, J.D., and Rugg, M.D. (2007). Recollection and the reinstatement of
digm. Mem. Cognit. 34, 1569–1577.
encoding-related cortical activity. Cereb. Cortex 17, 2507–2515.
Burgess, P.W., Dumontheil, I., and Gilbert, S.J. (2007). The gateway hypothe-
Kihlstrom, J.F. (1985). Hypnosis. Annu. Rev. Psychol. 36, 385–418.
sis of rostral prefrontal cortex (BA 10) function. Trends Cogn. Sci. 11, 290–298.
Kihlstrom, J.F. (1997). Hypnosis, memory and amnesia. Philos. Trans. R. Soc.
Cabeza, R., and Nyberg, L. (2000). Imaging cognition II: An empirical review of Lond. B Biol. Sci. 352, 1727–1732.
275 PET and fMRI studies. J. Cogn. Neurosci. 12, 1–47.
Kirsch, I. (1998). Deconstructing and reconstructing hypnosis. Cont. Hypnosis
Coe, W.C., Basden, B.H., Basden, D., Fikes, T., Gargano, G.J., and Webb, M. 15, 166–170.
(1989). Directed forgetting and posthypnotic amnesia: information processing
Kirsch, I., and Lynn, S.J. (1995). The altered state of hypnosis. Changes in the
and social contexts. J. Pers. Soc. Psychol. 56, 189–198.
theoretical landscape. Am. Psychol. 50, 846–858.
Conway, M.A., and Fthenaki, A. (2003). Disruption of inhibitory control of mem- Lepage, M., Ghaffar, O., Nyberg, L., and Tulving, E. (2000). Prefrontal cortex
ory following lesions to the frontal and temporal lobes. Cortex 39, 667–686. and episodic memory retrieval mode. Proc. Natl. Acad. Sci. USA 97, 506–511.
Cox, R.E., and Barnier, A.J. (2003). Posthypnotic amnesia for a first romantic Levy, B.J., and Anderson, M.C. (2002). Inhibitory processes and the control of
relationship: Forgetting the entire relationship versus forgetting selected memory retrieval. Trends Cogn. Sci. 6, 299–305.
events. Memory 11, 307–318.
Lichtenberg, P., Bachner-Melman, R., Ebstein, R.P., and Crawford, H.J.
David, D., Brown, R., Pojoga, C., and David, A. (2000). The impact of posthyp- (2004). Hypnotic susceptibility: multidimensional relationships with Clo-
notic amnesia and directed forgetting on implicit and explicit memory: new ninger’s Tridimensional Personality Questionnaire, COMT polymorphisms,
insights from a modified process dissociation procedure. Int. J. Clin. Exp. absorption, and attentional characteristics. Int. J. Clin. Exp. Hypn. 52, 47–72.
Hypn. 48, 267–289. Maguire, E.A., and Mummery, C.J. (1999). Differential modulation of a common
Evans, F.J. (1988). Posthypnotic Amnesia: Overview. In Hypnosis and Mem- memory retrieval network revealed by positron emission tomography. Hippo-
ory, H.M. Pettinati, ed. (New York: Guilford), pp. 157–192. campus 9, 54–61.

Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc. 169
 Neuron
Brain Correlates of Posthypnotic Amnesia

Maguire, E.R., Frith, C.D., and Morris, R.G.M. (1999). The functional neuro- evidence from positron emission tomography. Proc. Natl. Acad. Sci. USA
anatomy of comprehension and memory: the importance of prior knowledge. 93, 321–325.
Brain 122, 1839–1850. Schnider, A. (2003). Spontaneous confabulation and the adaptation of thought
Markowitsch, H.J. (1999). Functional neuroimaging correlates of functional to ongoing reality. Nat. Rev. Neurosci. 4, 662–671.
amnesia. Memory 7, 561–583.
Sohn, M.H., Goode, A., Stenger, V.A., Cameron, S.C., and Anderson, J.R.
Markowitsch, H.J. (2003). Psychogenic amnesia. Neuroimage 20 (Suppl 1), (2003). Competition and representation during memory retrieval: Roles of the
S132–S138. prefrontal cortex and the posterior parietal cortex. Proc. Natl. Acad. Sci.
Morris, C.D., Bransford, J.D., and Franks, J.J. (1977). Levels of processing ver- USA 100, 7412–7417.
sus transfer appropriate processing. J. Verb. Lang. Verb. Behav. 16, 519–533. Spiegel, H., and Spiegel, D. (2004). Trance and Treatment: Clinical Uses of
Moscovitch, M., and Winocur, G. (2002). The frontal cortex and working with Hypnosis. American Psychiatric Publishing.
memory. In The Frontal Lobes, D.T. Stuss and R.T. Knight, eds. (Oxford: Ox- Stark, C.E.L., and Squire, L.R. (2001). When zero is not zero: The problem of
ford University Press), pp. 188–209. ambiguous baseline condition in fMRI. Proc. Natl. Acad. Sci. USA 98,
Naghavi, H.R., and Nyberg, L. (2005). Common fronto-parietal activity in atten- 12760–12766.
tion, memory, and consciousness: shared demands in integration? Conscious.
Stern, D.B., Speigel, H., and Nee, J.C. (1979). The hypnotic induction profile:
Cogn. 14, 390–425.
normative observations, reliability, and validity. Am. J. Clin. Hypn. 21, 109–133.
Nyberg, L., Persson, J., Habib, R., Tulving, E., McIntosh, A.R., Cabeza, R., and
Svoboda, E., McKinnon, M.C., and Levine, B. (2006). The functinal neuroanat-
Houle, S. (2000). Large scale neurocognitive networks underlying episodic
omy of autobiographical memory: a meta-analysis. Neuropsychologia 44,
memory. J. Cogn. Neurosci. 12, 163–173.
2189–2208.
Olson, I.R., Plotzer, A., and Ezzyat, Y. (2007). The enigmatic temporal pole:
Talairach, J., and Tournoux, P. (1988). Co-Planar Stereotaxic Atlas of the
a review of findings on social and emotional processing. Brain 130,
Human Brain (New York: Thieme Medical Publishers, Inc.).
1718–1731.
Polyn, S.M., Natu, V.S., Cohen, J.D., and Norman, K.A. (2005). Category- Tulving, E. (1983). Elements of Episodic Memory (Oxford: Oxford University
specific cortical activity precedes retrieval during memory search. Science Press).
310, 1963–1966. Vaidya, C.J., Zhao, M., Desmond, J.E., and Gabrieli, J.D. (2002). Evidence for
Racsmany, M., and Conway, M.A. (2006). Episodic inhibition. J. Exp. Psychol. cortical encoding specificity in episodic memory: memory-induced re-activa-
Learn. Mem. Cogn. 32, 44–57. tion of picture processing areas. Neuropsychologia 40, 2136–2143.

Raz, A., Fan, J., and Posner, M.L. (2006). Neuroimaging and genetic associa- Wagstaff, G.F. (1998). The semantics and physiology of hypnosis as an altered
tions of attentional and hypnotic processes. J. Physiol. (Paris) 99, 483–491. state: towards a definition of hypnosis. Contemp. Hypn. 15, 149–165.
Roediger, H.L., III, Dudai, Y., and Fitzpatrick, S.M. (2007). Science of Memory: Wagstaff, G.F., Parkes, M., and Hanley, J.R. (2001). A comparison of posthyp-
Concepts (New York: Oxford University Press). notic amnesia and the simulation of amnesia through brain injury. Int.
Rosen, V.M., and Engle, R.W. (1998). Working memory capacity and suppres- J. Psychol. Psychol. Ther. 1, 67–78.
sion. J. Mem. Lang. 39, 418–436. Weitzenhoffer, A.M., and Hilgard, E.R. (1962). Stanford hypnotic susceptibility
Rugg, M.D., and Wilding, E.L. (2000). Retrieval processing and episodic scale, form C (Palo Alto, CA: Consulting Psychologists Press).
memory. Trends Cogn. Sci. 30, 108–115. Williamsen, J.A., Johnson, H.J., and Eriksen, C.W. (1965). Some characteris-
Sakai, K. (2003). Reactivation of memory: role of medial temporal lobe and tics of posthypnotic amnesia. J. Abnorm. Psychol. 70, 123–131.
prefrontal cortex. Rev. Neurosci. 14, 241–252. Yancey, S.W., and Phelps, E.A. (2001). Functional neuroimaging and episodic
Schacter, D.L., and Buckner, R.L. (1998). On the relations among priming, memory: a perspective. J. Clin. Exp. Neuropsychol. 23, 32–48.
conscious recollection, and intentional retrieval: evidence from neuroimaging Yasuno, F., Nishikawa, T., Nakagawa, Y., Ikejiri, Y., Tokunaga, H., Mizuta, I.,
research. Neurobiol. Learn. Mem. 70, 284–303. Shinozaki, K., Hashikawa, K., Sugita, Y., Nishimura, T., and Takeda, M.
Schacter, D.L., Alpert, N.M., Savage, C.R., Rauch, S.L., and Albert, M.S. (2000). Functional anatomical study of psychogenic amnesia. Psychiatry
(1996). Conscious recollection and the human hippocampal formation: Res. 99, 43–57.

170 Neuron 57, 159–170, January 10, 2008 ª2008 Elsevier Inc.