Prefrontal Regions Orchestrate Suppression of Emotional Memories

via a Two-Phase Process

Brendan E. Depue et al.
Science 317, 215 (2007);
DOI: 10.1126/science.1139560

This copy is for your personal, non-commercial use only.

If you wish to distribute this article to others, you can order high-quality copies for your
colleagues, clients, or customers by clicking here.

Downloaded from www.sciencemag.org on July 1, 2014

Permission to republish or repurpose articles or portions of articles can be obtained by
following the guidelines here.

The following resources related to this article are available online at

www.sciencemag.org (this information is current as of July 1, 2014 ):

Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/content/317/5835/215.full.html
Supporting Online Material can be found at:
http://www.sciencemag.org/content/suppl/2007/07/10/317.5835.215.DC1.html
A list of selected additional articles on the Science Web sites related to this article can be
found at:
http://www.sciencemag.org/content/317/5835/215.full.html#related
This article cites 29 articles, 12 of which can be accessed free:
http://www.sciencemag.org/content/317/5835/215.full.html#ref-list-1
This article has been cited by 43 article(s) on the ISI Web of Science
This article has been cited by 25 articles hosted by HighWire Press; see:
http://www.sciencemag.org/content/317/5835/215.full.html#related-urls
This article appears in the following subject collections:
Neuroscience
http://www.sciencemag.org/cgi/collection/neuroscience

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2007 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.
Geisa Carolinne Ferreira Mendes - [email protected] - CPF: 106.772.336-63
 RESEARCH ARTICLE
using faces as cues and pictures as targets, suggests
that the impact of exerting control is larger for
Prefrontal Regions Orchestrate emotional than for nonemotional information
(6). Because of these findings, as well as the

Suppression of Emotional Memories clinical relevance of intrusive negative images in

posttraumatic stress disorder (PTSD) and obsessive-
compulsive disorder (OCD) (7–9), we used nega-
via a Two-Phase Process tive pictures known to activate the visual cortex as
memory targets (10). In both monkeys (11) and
Brendan E. Depue,1,2* Tim Curran,1,2,3 Marie T. Banich1,2,3,4 humans (10), retrieval of memories involves the
same regions of the brain that are involved in
Whether memories can be suppressed has been a controversial issue in psychology and cognitive encoding a memory (12, 13). Thus, we used the
neuroscience for decades. We found evidence that emotional memories are suppressed via two degree of activation in the hippocampus, amygda-
time-differentiated neural mechanisms: (i) an initial suppression by the right inferior frontal gyrus la, and visual processing areas that are known to be
over regions supporting sensory components of the memory representation (visual cortex, involved in the creation and retrieval of emotional
thalamus), followed by (ii) right medial frontal gyrus control over regions supporting multimodal memories (14, 15)—to determine whether mem-
and emotional components of the memory representation (hippocampus, amygdala), both of which ory suppression involves inhibition of the neural
are influenced by fronto-polar regions. These results indicate that memory suppression does machinery underlying memory processing.
occur and, at least in nonpsychiatric populations, is under the control of prefrontal regions. Two aspects of our version of the T/NT
paradigm were important (Fig. 1A). First, the
ne of the most controversial issues in We hypothesized that evidence for memory experimental phase included fixation items that

O psychology over the past 100 years is

the degree to which memories can be
manipulated, both whether they can be falsely
suppression would be provided by activation
below a fixation baseline in brain regions process-
ing components of memory representation (5). We
acted as a low-level baseline of brain activity to

1
Department of Psychology, University of Colorado,
created (1) and whether they can be suppressed. used the Think/No-Think paradigm (T/NT) of Boulder, CO 80309, USA. 2Center for Neuroscience,
Although some researchers have provided ini- Anderson and colleagues, in which individuals University of Colorado, Boulder, CO 80309, USA. 3Institute
of Cognitive Science, University of Colorado, Boulder, CO
tial evidence for memory suppression (2, 3), attempt to elaborate a memory by repetitively 80309, USA. 4Department of Psychiatry, University of
others claim that memory repression or suppres- thinking of it (T condition) or to suppress a mem- Denver Health Sciences, Denver, CO 80208, USA.
sion is a clinical myth in search of scientific ory by repetitively not letting it enter conscious- *To whom correspondence should be addressed. E-mail:
support (4). ness (NT condition) (2, 3). Our recent T/NT work, [email protected]

A B
100
Training Phase
90

80
+ x 40 faces
Percentage Recalled

70
Base
60
T
Experimental Phase 50
NT
Think 40

Think of 30
the previously m (Base) = 62.5%
associated picture 20
m (T) = 71.1%
10 m (NT) = 53.2%

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
OR x 12 per face
No-Think Number of Participants
Do not let the Fig. 1. (A) Experimental procedure. Individuals were first trained during
previously associated structural scanning to associate 40 cue-target pairs. During the experimental
picture enter
consciousness phase, brain activity was recorded using fMRI while individuals viewed only the
face (16 faces per condition, 12 repetitions per face; 3.5 s per face). On some
trials they were instructed to think of the previously learned picture; on other
trials they were instructed not to let the previously associated picture enter
consciousness. The presentation of only the cue (i.e., the face) ensures that
Testing Phase individuals manipulate the memory of the target picture. The additional faces (8
items) not shown during this phase acted as a behavioral baseline. During the
Give Short test phase, the individuals were shown the 40 faces and asked to describe the
+ Description x 40 pairs
previously associated picture. (B) Behavioral results: percentage recall for each
participant for T trials (green) and NT trials (red), with the dotted line indicating
baseline recall for items not viewed in the experimental phase. Recall differed for
T and NT items [t(15) = –4.29, P = 0.0006] because of a trend for greater recall in the T condition [t(15) = 1.49, P = 0.07] and a significant reduction
in recall in the NT condition [t(15) = –2.28, P = 0.02], calculated relative to baseline.

www.sciencemag.org
Geisa Carolinne Ferreira Mendes - [email protected] SCIENCE VOL 317
- CPF: 106.772.336-63 13 JULY 2007 215
 RESEARCH ARTICLE
determine whether activity in the regions of NT trials than for T trials—irrespective of recall also a negative signal change for NT trials and,
interest (ROIs) actually decreased (5). Although accuracy (5). Additionally, because we were conversely, a positive signal change for T trials,
Anderson and colleagues (2) demonstrated the specifically interested in effective control over relative to baseline. Brain areas underlying the
involvement of prefrontal regions in cognitive memory, we examined this same contrast only sensory representation of memory that showed
control over memory, they examined the contrast for NT trials in which the picture was forgotten such an effect were the visual cortex, including
between activation on NT and T trials without (NTf) and on T trials in which it was remem- bilateral BA17, BA18, and BA37 [fusiform gyrus
reporting fixation baseline contrasts, making it bered (Tr). Brain areas are discussed according (FG)], and the pulvinar nucleus of the thalamus
difficult to associate activity with a specific to their putative functional roles: (i) cognitive (Pul; Fig. 2B). Suppression of emotional memo-
condition (e.g., T or NT). control, (ii) sensory representation of memory, ries thus involved decreased activity in sensory
Second, in the experimental phase, each face and (iii) memory processes and emotional cortices that are normally active when memories
was shown 12 times with the presentation of the components of memory. are being retrieved, as well as in regions (i.e., Pul)
faces pseudo-randomly intermixed. Although in Sources and sites of cognitive control. Pre- that play a role in gating and modulating attention
our previous study (6) memory increased lin- frontal regions, right-sided and spanning BA 8, toward or away from visual stimuli (16, 17).
early with the number of times cognitive control 9/46, 47, and BA 10, exhibited NT > T contrast. Brain areas involved in memory processes
was exerted for T trials, this was not the case for A conjunction analysis indicated that these and emotional components of memory repre-
NT trials. Hence, we divided our data into quar- differences resulted from an increase in activity sentation that exhibited significant condition-
tiles (three repetitions per cue during each quar- for NT trials rather than a decrease in activation specific activity were the hippocampal complex
tile) to determine whether the neural mechanisms for T trials relative to baseline (Fig. 2A), which (Hip) and the amygdala (Amy; Fig. 2C), which
involved in memory suppression change with suggests that these regions are specifically are strongly involved in emotional learning and
repeated attempts. involved in controlling the suppression of emo- memory (18, 19), as would be required by the
The behavioral results shown in Fig. 1B tional memories. face-picture association in our paradigm. Hence,
indicated that individuals effectively suppressed Next, we sought to determine which regions the suppression of emotional memory also
memory. To investigate the neural basis of this support components of the memory representa- involved suppression of memory processes and
effect, we first analyzed the functional magnetic tion. These regions showed condition-specific emotional aspects of the memory representation.
resonance imaging (fMRI) data for instances of changes, which we defined as yielding not only a Two phases of memory suppression. We
NT > T—that is, significantly greater activity for significant difference for the NT > T contrast but next examined how brain activity changed with

Fig. 2. Functional activa-

tion of brain areas involved
in (A) cognitive control, (B)
sensory representations of
memory, and (C) memory
processes and emotional
components of memory
(rSFG, right superior frontal
gyrus; rMFG, right middle
frontal gyrus; rIFG, right
inferior frontal gyrus; Pul,
pulvinar; FG, fusiform gyrus;
Hip, hippocampus; Amy,
amygdala). fMRI data were
analyzed using a hemody-
namic response function–
convolved epoch; statistical
parameter maps (SPMs)
were thresholded on a vox-
elwise basis at Z = 2.81, P =
0.005. To adjust for false
positive errors on an area-
of-activation basis, we set a
clusterwise threshold at P =
0.05 (cluster size of 120) as
determined by the Analysis
of Functional NeuroImages
program AlphaSim. Red in-
dicates greater activity for
NT trials than for T trials;
blue indicates the reverse.
Conjunction analyses re-
vealed that areas seen in
blue are the culmination of
increased activity for T trials
above baseline as well as
decreased activity of NT
trials below baseline.

216Carolinne Ferreira Mendes - [email protected]

Geisa 13 JULY 2007- CPF:
VOL106.772.336-63
317 SCIENCE www.sciencemag.org
 RESEARCH ARTICLE
increased attempts at cognitive control, over trials, In contrast to rIFG, rMFG activation in- BA10 and each of the prefrontal regions was
as well as covariation of prefrontal activation and creased later and remained active. Activity in observed in the quartile preceding the maximal
the two putative sets of posterior cortex (sensory Hip and Amy appeared to follow that of rMFG correlation between rIFG and rMFG and their
areas, Amy-Hip complex). This was ac- in reverse: In the first quartile, activity was respectively associated brain regions (Fig. 4, B
complished by dividing the 12 attempts at cog- significantly above baseline, but it decreased to D). This pattern suggests that activity in
nitive control into quartiles (three repetitions per steadily to reach significance below baseline by BA10 modulates activity in rIFG and rMFG,
quartile). We then plotted percent signal change the fourth quartile (Fig. 3B). Increased activity which in turn modulates activity in posterior
(DS) across the quartiles of each brain area that in rMFG did not predict activity in Hip and regions. This result is consistent with recent
showed differential activation for NT trials rela- Amy in the first or second quartiles, but did so research suggesting that activity in BA10
tive to T trials (5). significantly in the third and fourth quartiles increases with task complexity and is involved
Two patterns of temporal change in activation (Fig. 3C). This covariation pattern suggests that in other aspects of higher-order cognition, such as
were observed, each associated with different rIFG and rMFG suppress different aspects of monitoring of internal states, memory retrieval,
groupings of prefrontal and posterior brain areas. memory processing during different temporal emotional processes, coordination of higher-level
The two groupings were composed of (i) right phases. goals, and modulating other areas of PFC to attain
inferior frontal gyrus (rIFG), Pul, and FG, and (ii) To assess the linkage between Hip activation these goals (21, 22).
right middle frontal gyrus (rMFG), Hip, and and memory, we performed a subsequent mem- Conclusion. Our finding of below-baseline
Amy. Unique temporal patterns of prefrontal ory analysis (remembered versus forgotten) on activation in brain areas related to sensory
cortex (PFC) influence on posterior brain regions Hip activation during the fourth quartile, when representation (Pul, FG) and memory (Hip,
were observed for each group. rIFG showed early the effects of cognitive control were largest. We Amy) for emotional stimuli during NT trials is
activation in the time course of suppression, found a linear decrease of activity in Hip: Tr > consistent with the operation of an active process
which lasted through the second quartile of Tf > baseline > NTr > NTf (5), with only that of of suppression. Furthermore, our findings are less
repetitions, and then a decreased level of NT trials below baseline. Further analyses consistent with two sets of alternative interpreta-
activation during the last two quartiles (Fig. showed that activity of rMFG and Hip correlated tions of responses in behavior and brain activa-
3A). Activity of Pul and FG followed a similar with behavioral success at suppression (5). Taken tion that we observe on NT trials. One alternative
time course of DS, which lagged behind that for together with the previous data, these findings explanation is replacement, in which participants
rIFG. To express quantitatively the relationship in suggest that rMFG modulates activity in Hip, think of a new cue-target association or substitute
activation between these brain areas as a function which is directly related to behavioral success. an alternate memory to create distraction from the
of time, we plotted the correlations of activation These data replicated the finding of Anderson target memory (23). However, the generation of a
of rIFG with Pul and FG, respectively, as a and colleagues, so the relationship between pre- new associate would be associated with activa-
function of quartile (Fig. 3C). Greater activity in frontal and hippocampal regions may well be tion, not deactivation below baseline of sensory
rIFG in the second quartile was significantly involved in suppressing memory recall. and memory representations to represent and
associated with decreased activity in Pul and FG Higher-order orchestration of memory sup- encode the new association. In essence, replace-
during the second quartile, but showed no sig- pression. Only one prefrontal region, BA 10, ment would manifest functionally in brain
nificant relation for later quartiles (20). This cor- exhibited activity across the entire experimental activations as in T trials.
relation pattern is consistent with the temporal phase (Fig. 4A). Because this region receives A second alternative suggests that reduced
pattern of DS: The apparent rIFG inhibition of its minimal sensory input and is predominantly recall on NT trials involves disengagement with
associated brain regions begins early, in the first connected with other prefrontal areas, we in- the stimuli (e.g., momentary mind-wandering or
quartile, reaches maximum in the second quartile, spected its relationship with rIFG and rMFG for introspection), because in some studies BA10
and is greatly reduced thereafter. each quartile. The maximal correlation between activation has been correlated with internally fo-

Fig. 3. Time course of percentage A IFG Pul FG B MFG Hip Amy

signal change (DS) relative to base- 0.20 0.20
line for the two regional groupings.
(A) rIFG shows significantly in-
% Signal Change

% Signal Change

0.10 0.10
creased activity only for the first
and second quartiles [quartile 1,
t(15) = 3.65, P = 0.001; quartile 2, 0.00 0.00
t(15) = 2.66, P = 0.01], whereas the
activities of Pul and FG decrease to -0.10 -0.10
significantly below baseline during
quartile 4 [Pul, t(15) = –2.11, P = -0.20 -0.20
0.05; FG, t(15) = –1.96, P = 0.05]. 1 2
Quartile
3 4 1 2 3 4
Quartile
(B) rMFG displays significantly ele-
vated activity during the second, third, and fourth quartiles (Fig. 4B) [quartile 2, t(15) = –4.42,
P = 0.0005; quartile 3, t(15) = 2.10, P = 0.05; quartile 4, t(15) = 2.08, P = 0.05], whereas Hip
C IFG&Pul IFG&FG MFG&Hip MFG&Amy
0.20
and Amy initially exhibit increased activity in quartile 1 but show significantly decreased activity
by quartile 4 [Hip quartile 1, t(15) = 3.11, P = 0.007; Hip quartile 4, t(15) = –1.98, P = 0.05;
Correlation Coefficient

0.00
Amy quartile 1, t(15) = 3.11, P = 0.007; Amy quartile 4, t(15) = –2.15, P = 0.04]. (C) -0.20
Correlation coefficients for rIFG and rMFG grouped with respective associated posterior brain
regions as a function of quartile. *P ≤ 0.05, +P ≤ 0.01, #P ≤ 0.001. -0.40

-0.60

-0.80

-1
1 2 3 4
Quartile

www.sciencemag.org
Geisa Carolinne Ferreira Mendes - [email protected] SCIENCE VOL 317
- CPF: 106.772.336-63 13 JULY 2007 217
 RESEARCH ARTICLE
cused attention. However, we found that BA10 suppression is most readily accomplished by pre- cating that Hip activity has been effectively
activity precedes and covaries with the activity of frontal modulation of activity in sensory cortices suppressed.
two separate prefrontal areas that are initiated that is likely to recreate the sensory percept. This At a broader level, our findings extend re-
at different temporal points in the process of process, however, does not predict subsequent search suggesting that prefrontal brain areas asso-
suppressing activation in two different sets of behavioral recall and may be a “late-correction” ciated with inhibitory mechanisms (BA 10 and
posterior brain regions. This time-phased, region- mechanism (26) to suppress prepotent activity in superior, inferior, and middle FG) are lateralized
specific activity pattern of BA10 is more complex sensory cortices generated as a result of viewing predominantly to the right hemisphere (27, 28).
than that expected for simple mind-wandering the cue. In contrast, activity of rMFG (over all We have shown the involvement of these areas in
during NT trials, and is more consistent with quartiles) and Hip (during the last quartile) pre- the suppression of emotional memories, which
BA10’s proposed role of orchestration of mul- dicts behavioral suppression (5). This suppres- replicates current literature suggesting that these
tiple prefrontal processes involved in guiding sion process may reduce access to the memory areas are active in the suppression of emotional
complex tasks (21, 22). itself, by blocking Hip and Amy activity needed reactivity (29, 30). Activity in these brain areas,
Our findings suggest that the suppression of for retrieval of the emotional memory. However, along with inhibition over Hip and Amy, sug-
emotional memory involves at least two path- this mechanism becomes effective only after re- gests that suppression of emotional memories may
ways with staggered phases of their modulatory peated attempts at control, as shown by the in- use mechanisms similar to those used in emotion
influence. The first pathway involves cognitive creased correlation in activity between rMFG-Hip regulation. Thus, various right-lateralized PFC
control by rIFG over sensory components of and rMFG-Amy over quartiles (Fig. 3C). Once areas may be involved in coordinating suppres-
memory representation, as evidenced by reduced successful suppression of hippocampal activity is sion processes across many behavioral domains,
activity in FG and Pul. This finding is consistent achieved, the need to invoke the “late-correction” including memory retrieval, motor processes, feel-
with computational models that posit that activa- process apparently decreases, as reflected in the ings of social rejection, self motives, and state
tion and inhibition of the thalamus is a critical decreased rIFG-Pul and rIFG-FG correlations. emotional reactivity (27–32).
means of gating working memory information Our data also provide an intriguing hint that, Our findings may have implications for ther-
(24). A second pathway involves cognitive con- as suggested in clinical practice, it is necessary apeutic approaches to disorders involving the
trol by rMFG over memory processes and emo- to “revisit” an emotionally distressing memory inability to suppress emotionally distressing mem-
tional components of memory representation via before it can be controlled. We found that early ories and thoughts, including PTSD, phobias,
modulation of Hip and Amy (2, 25). The overall in the course of suppression (during quartiles ruminative depression/anxiety, and OCD. They
timing of these suppression effects appears to be 1 and 2), greater Hip activity was observed for provide the possibility for approaches to control-
orchestrated by a modulatory influence of BA10, NTf trials than NTr trials, as was also found by ling memories by suppressing sensory aspects
first over rIFG, then over rMFG. Further research Anderson et al. in nontemporal analyses (2, 5). of memory and/or by strengthening cognitive
is needed to determine the extent to which the We speculate that memories that activate Hip to control over memory and emotional processes
two pathways act independently or interactively. a greater degree are more amenable to cognitive through repeated practice. Refinement of ther-
Our findings also provide insight into the control because they are more elaborated and apeutic procedures based on these distinct means
potential neurocognitive mechanisms by which may provide increased access. By the fourth of manipulating emotional memory might be an
emotional memories are suppressed. Our data quartile, activation in Hip for NTf trials (but not exciting and fruitful development in future
are consistent with the idea that initial memory NTr trials) is significantly below baseline, indi- clinical research.

Fig. 4. (A) Time course of percent A BA10 IFG MFG B BA10&IFG BA10&MFG
signal change for BA10, rIFG, and
0.20 0.8
rMFG. (B) Correlations of rBA10
with rIFG and rMFG as a function 0.6
Correlation Coefficient

of quartiles of suppression repeti- 0.15

% Signal Change

tions. (C) Correlations of BA10 0.4

with rIFG, rIFG with Pul, and rIFG 0.10 0.2

with FG as a function of quartiles
of suppression repetitions. (D) 0
Correlations of BA10 with rMFG, 0.05
-0.2
rMFG with Hip, and rMFG with Amy
as a function of suppression repe- 0.00 -0.4
titions. Arrows in (C) and (D) 1 2 3 4 1 2 3 4
Quartile Quartile
indicate a preceding orchestration
of control. For the quartile in which
the correlation was maximal, great- C BA10&IFG IFG&Pul IFG&FG D BA10&MFG MFG&Hip MFG&Amy
er activity in BA10 was associated 0.8 0.8
with significantly increased activ- 0.6 0.6
ity in rIFG (r = 0.56, P = 0.03) and
Correlation Coefficient

0.4
Correlation Coefficient

0.4
rMFG (r = 0.77, P = 0.001), but 0.2 0.2
not with other brain areas [Pul, r = 0 0
–0.16; FG, r = –0.16; Hip, r = –0.22;
-0.2 -0.2
Amy, r = –0.09; Ps > 0.40 (5)]. *P ≤
0.05, +P ≤ 0.01, #P ≤ 0.001. -0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
1 2 3 4 1 2 3 4
Quartile Quartile

218Carolinne Ferreira Mendes - [email protected]

Geisa 13 JULY 2007- CPF:
VOL106.772.336-63
317 SCIENCE www.sciencemag.org
 Our results suggest that effective voluntary 12. S. M. Kosslyn, W. L. Thompson, in The New Cognitive 28. A. R. Aron, T. W. Robbins, R. A. Poldrack, Trends Cognit.
suppression of emotional memory only develops Neurosciences, M. S. Gazzaniga, Ed. (MIT Press, Sci. 8, 170 (2004).
Cambridge, MA, ed. 2, 2000), pp. 975–985. 29. K. N. Ochsner et al., J. Cognit. Neurosci. 14, 1215
with repeated attempts to cognitively control pos- 13. C. L. Raye et al., Cortex 43, 135 (2007). (2002).
terior brain areas underlying instantiated memo- 14. A. D. Wagner et al., Neuroimage 14, 1337 (2001). 30. N. I. Eisenberger, M. D. Lieberman, K. D. Williams,
ries. In this sense, memory suppression may best 15. J. L. McGaugh, Science 287, 248 (2000). Science 302, 290 (2003).
be conceived as a dynamic process in which the 16. S. Kastner et al., J. Neurophysiol. 91, 438 (2004). 31. H. L. Urry et al., J. Neurosci. 26, 4415 (2006).
17. S. Kastner et al., Science 282, 108 (1998). 32. D. Knoch, A. Pascual-Leone, K. Meyer, V. Treyer, E. Fehr,
brain acquires multiple modulatory influences to 18. S. Hamann, Trends Cognit. Sci. 5, 394 (2001). Science 314, 829 (2006); published online 4 October
reduce the likelihood of retrieving unwanted 19. E. A. Phelps, Curr. Opin. Neurobiol. 14, 198 (2004). 2006 (10.1126/science.1129156).
memories. 20. Our activation of FG has been established to be different 33. I. Gautier et al., Nat. Neurosci. 3, 191 (2000).
(more posterior/medial) from the activation of the 34. We thank Y. Du, D. Singel, and R. Freedman for
fusiform face area (FFA) commonly observed in face assistance in functional imaging, and three anonymous
References and Notes perception studies (33). reviewers for their critical insights. Supported by the
1. E. Loftus, Neuroscience 4, 231 (2003). 21. K. Sakai, R. E. Passingham, J. Neurosci. 26, 1211 (2006). Graduate School, the Vice-Chancellor for Research, and
2. M. C. Anderson et al., Science 303, 232 (2004). 22. N. Ramnani, A. M. Owen, Nat. Rev. Neurosci. 5, 184 the Institute of Cognitive Science at the University of
3. M. C. Anderson, C. Green, Nature 410, 366 (2001). (2004). Colorado at Boulder (M.T.B.).
4. J. F. Kihlstrom, Trends Cognit. Sci. 6, 502 (2002). 23. P. T. Hertel, G. Calcaterra, Psychonom. Bull. Rev. 12, 484
5. See supporting data on Science Online. (2005).
6. B. E. Depue, M. T. Banich, T. Curran, Psychol. Sci. 17, 24. R. C. O’Reilly, Science 314, 91 (2006).
Supporting Online Material
441 (2006). 25. M. Beauregard, J. Lévesque, P. Bourgouin, J. Neurosci. www.sciencemag.org/cgi/content/full/317/5835/215/DC1
21, RC165 (2001). Materials and Methods
7. B. A. van der Kolk, J. A. Burbridge, J. Suzuki, Ann. N.Y.
Figs. S1 to S5
Acad. Sci. 821, 99 (1997). 26. L. L. Jacoby, C. M. Kelley, B. D. McElree, in Dual-Process
8. B. A. van der Kolk et al., Am. J. Psychiatry 153, 83 (1996). Theories in Social Psychology, S. Chaiken, Y. Trope, Eds. Tables S1 to S4
(Guilford, New York, 1999), pp. 383–400. References
9. C. Purdon, Behav. Res. Ther. 37, 1029 (1999).
10. P. J. Lang et al., Psychophysiology 35, 199 (1998). 27. H. Garavan, T. J. Ross, E. A. Stein, Proc. Natl. Acad. Sci. 4 January 2007; accepted 6 June 2007
11. E. T. Rolls, Hum. Neurobiol. 3, 209 (1984). U.S.A. 96, 8301 (1999). 10.1126/science.1139560

REPORTS
The data that we present were obtained from a re-
Scattering and Interference in gion identified as bilayer graphene (11). Scanning
tunneling microscopy (STM) measurements were
Epitaxial Graphene performed in a custom-built ultrahigh-vacuum,
low-temperature instrument. We measured the
scanning tunneling spectroscopy (STS) differential
G. M. Rutter,1 J. N. Crain,2 N. P. Guisinger,2 T. Li,1 P. N. First,1* J. A. Stroscio2*
conductance dI/dV (where I is current and V is
voltage) with lock-in detection by applying a small
A single sheet of carbon, graphene, exhibits unexpected electronic properties that arise from
modulation to the tunnel voltage at ≈500 Hz.
quantum state symmetries, which restrict the scattering of its charge carriers. Understanding the
Differential conductance maps were obtained by
role of defects in the transport properties of graphene is central to realizing future electronics
recording an STS spectrum at each spatial pixel
based on carbon. Scanning tunneling spectroscopy was used to measure quasiparticle interference
in the topographic measurement. All measure-
patterns in epitaxial graphene grown on SiC(0001). Energy-resolved maps of the local density of
ments reported here were taken at 4.3 K.
states reveal modulations on two different length scales, reflecting both intravalley and intervalley
STM topographic images (Fig. 1) show the
scattering. Although such scattering in graphene can be suppressed because of the symmetries of
atomic structure and different types of disorder
the Dirac quasiparticles, we show that, when its source is atomic-scale lattice defects, wave
for epitaxial graphene on SiC(0001). At the
functions of different symmetries can mix.
atomic scale, the graphene is imaged as a tri-
angular lattice (Fig. 1B), characteristic of imag-
uilt of a honeycomb of sp2-bonded car- effect (1, 2) and weak antilocalization (3, 4), that ing only one of the two graphene sublattices.

B bon atoms, graphene has a linear, neutrino-

like energy spectrum near the Fermi
energy, EF. This results from the intersection of
have spurred an intense scientific interest in
graphene (5). Bilayer graphene (5–7) is equally
distinctive: Quasiparticle states are chiral (6) with
Superimposed on this atomic structure is a mod-
ulation period of ≈2 nm caused by a reconstruc-
tion of the SiC interface beneath the graphene: a
electron and hole cones in the graphene band Berry’s phase 2p for the bilayer versus p for the SiC “6 × 6” superstructure (12). Survey images
structure at the Dirac energy, ED. The linear ener- monolayer (6). High carrier mobilities, chemical reveal two categories of defects. Type A defects,
gy dispersion and concomitant topological con- inertness, and the two-dimensional (2D) nature of such as mounds (red arrow in Fig. 1A), have an
straints give rise to massless Dirac quasiparticles graphene make it a promising candidate for fu- unperturbed graphene structure that is continu-
in graphene, with energy-independent propaga- ture electronic-device applications (1, 2, 5, 8, 9). ous, akin to a blanket. These defects are due to
tion speed vF ≈ 106 m/s (where vF is the Fermi In particular, graphene grown epitaxially on SiC irregularities in the interface layer between
velocity). Distinctive symmetries of the graphene substrates and patterned via standard lithographic graphene and the SiC bulk. In contrast, type B
wave functions lead to unusual quantum proper- procedures has been proposed as a platform for defects are atomic defects within the graphene
ties, such as an anomalous integer quantum Hall carbon-based nanoelectronics and molecular lattice itself (Fig. 1, A, C, and D) and are ac-
electronics (8, 9). companied by strong distortions in the local
1
School of Physics, Georgia Institute of Technology, Epitaxial graphene was grown on the silicon- lattice images. These distortions are of electronic
Atlanta, GA 30332, USA. 2Center for Nanoscale Science terminated (0001) face of high-purity semi- origin and are accompanied by large increases in
and Technology, National Institute of Standards and
Technology, Gaithersburg, MD 20899, USA. insulating 4H-SiC by thermal desorption of the local density of states (LDOS) at the defect
*To whom correspondence should be addressed. E-mail:
silicon at high temperatures (8, 10). This method site (13, 14). Quasiparticle scattering from type B
[email protected] (J.A.S.); [email protected] produces an electron-doped graphene system, defects gives rise to spectacular patterns in the
(P.N.F.) with the Fermi level 200 to 400 meV above ED. topographic images (Fig. 1, C and D) resulting

www.sciencemag.org
Geisa Carolinne Ferreira Mendes - [email protected] SCIENCE VOL 317
- CPF: 106.772.336-63 13 JULY 2007 219