Early onset of neural synchronization in the contextual
Kestutis Kveragaa,b,1, Avniel Singh Ghumanc, Karim S. Kassamd, Elissa A. Aminoffe, Matti S. Hämäläinena,b,
Maximilien Chaumona,f, and Moshe Bara,b,g
aAthinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA 02129;bDepartment of Radiology, Harvard Medical
School, Boston, MA 02115;cLaboratory for Brain and Cognition, National Institutes of Mental Health, Bethesda, MD 20892;dDepartment of Social and
Decision Sciences, Carnegie Mellon University, Pittsburgh, PA 15213;eDepartment of Psychology, University of California, Santa Barbara, CA 93106;
fDepartment of Psychology, Boston College, Chestnut Hill, MA 02459; andgDepartment of Psychiatry, Harvard Medical School, Boston, MA 02115
Edited by Robert Desimone, Massachusetts Institute of Technology, Cambridge, MA, and approved January 12, 2011 (received for review September 17, 2010)
Objects are more easily recognized in their typical context. How-
ever, is contextual information activated early enough to facilitate
the perception of individual objects, or is contextual facilitation
caused by postperceptual mechanisms? To elucidate this issue, we
first need to study the temporal dynamics and neural interactions
contextual network consists of the parahippocampal, retrosplenial,
and medial prefrontal cortices. We used functional MRI, magneto-
encephalography, and phase synchrony analyses to compare the
neural response to stimuli with strong or weak contextual associ-
ations. The context network was activated in functional MRI and
preferentially synchronized in magnetoencephalography (MEG)
for stimuli with strong contextual associations. Phase synchrony
increased early (150–250 ms) only when it involved the parahippo-
campal cortex, whereas retrosplenial–medial prefrontal cortices syn-
chrony was enhanced later (300–400 ms). These results describe the
neural dynamics of context processing and suggest that context is
activated early during object perception.
phase locking|oscillations|beta|visual cognition
popcorn in a movie theater would not surprise anyone familiar
with American movie theaters. However, at the opera, cham-
pagne and chocolate might be the more common sight. Whether
and how context can facilitate visual recognition has been the
subject of much research and vigorous debate. Behavioral re-
search studying the effects of contextual information in visual
perception has shown that information about context can facili-
tate recognition of visual scenes and objects (1–5). However,
Hollingworth and Henderson (6) have argued that at least some
of this contextual facilitation is attributable to a response bias
and that any contextual influence may be the result of later
postidentification processes rather than early activation of con-
textual information during recognition (6, 7). Here, we address
the central question of when in the recognition process contex-
tual information is activated by examining the temporal dy-
namics of neural regions involved in processing contextual
information. Furthermore, we use this information regarding the
temporal dynamics of the areas involved in contextual processing
to help elucidate the functions of the individual members of the
In the past decade, neuroimaging studies using stimuli with
strong and weak contextual associations have identified the
components and basic properties of the network that mediates
context-based associations; studies (8–18) have reported a core
set of regions (Fig. 1A) that are activated by contextual associ-
ations: the parahippocampal cortex (PHC), the retrosplenial
complex (RSC), and in some cases, the medial prefrontal cortex
(MPFC). Characterizing the individual and combined roles of
these regions in contextual association processing has been
challenging, because the PHC, RSC, and MPFC historically have
e know from experience that specific contexts are associ-
ated with specific objects. Seeing a jumbo-sized bucket of
been implicated in numerous cognitive processes. The PHC is
strongly activated during scene or place recognition and spatial
navigation (19), and it has been labeled the parahippocampal
place area (PPA), because it is preferentially activated when one
views pictures of places (20, 21). In addition, PHC has been
shown to play an important role in episodic memory (22–25).
These seemingly disparate roles of PHC can be reconciled if its
core function is taken more globally as mediating contextual
associations (8, 10, 26, 27), with spatial contextual associations
processed in the PPA and nonspatial contextual associations
processed in an abutting, slightly anterior region of the PHC (8,
10). In support of this view, studies have shown that the PHC
response is modulated based on the degree of contextual in-
formation contained in single objects (8, 17, 27).
Similarly to the PHC, the RSC is consistently activated in tasks
involving spatial navigation (28), semantic and episodic memory
(29), and scene and object perception (8, 15, 30). As with the
PHC, the RSC has been shown to be involved in many different
cognitive processes, which makes it difficult to identify its specific
function or functions (31). One possibility is that it translates
allocentric representations that it receives from the hippocampus
and PHC into egocentric representations (31); another possibility
is that it transforms a particular instance of a context into a more
generic representation of that context (8, 26, 32). The MPFC is
likewise implicated in a vast number of cognitive and emotion-
related functions, including semantic integration of current con-
text, self-relevant processing, inferring mental states of others,
and shifting one’s perspective (17, 33–35).
Developing a better understanding of the specific roles of the
PHC, RSC, and MPFC in context processing requires that we
learn how these components of the context network function and
interact with one another by examining the neural dynamics and
communication patterns within this network and beyond. Spe-
cifically, we wanted to know when and under what conditions
contextual associations (CA) modulate functional communica-
tion between the PHC, RSC, and MPFC as well as the visual
cortex and possibly, other visual object processing regions, such
as the fusiform and orbitofrontal cortex (36–44). Therefore, we
used magnetoencephalography (MEG) to characterize the spa-
tiotemporal dynamics and cross-cortical communication within
the contextual associations network as well as with other regions
involved in visual recognition. Neural oscillations play a critical
role in the timing of neural events and tend to be synchronized in
Author contributions: K.S.K., E.A.A., and M.B. designed research; K.S.K. and E.A.A. per-
formed research; K.K., A.S.G., and M.S.H. contributed new reagents/analytic tools; K.K.,
A.S.G., K.S.K., E.A.A., M.S.H., M.C., and M.B. analyzed data; and K.K., A.S.G., and M.B.
wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
| February 22, 2011
| vol. 108
| no. 8
regions that exchange information (45–48). We computed phase-
locking estimates to measure functional connectivity to assess
whether neuronal responses in two regions have a consistently
covarying phase relationship with one another (49). Unlike
power estimates, which reflect the extent of synchronized neural
firing, phase-locking estimates provide information about the
timing of firing and interconnectivity between neural regions
(50), and they are believed to be a more reliable measure of
higher-level functions (51).
The behavioral paradigm that we used to investigate the dy-
namics of contextual association processing had been tested and
used in previous functional MRI (fMRI) studies (8) in a blocked
design. In the present study, we used a combination of rapid
even-related design with the same stimulus set in both fMRI and
MEG and used the fMRI activations to inform our region of
interest (ROI) selections in MEG. We showed participants pic-
tures of single objects with strong and weak CA (Fig. 1B) and
asked them to respond through a button box when they recog-
nized the object. The participants were not told that some
objects had strong CA, whereas others had weak CA. Our hy-
pothesis was that strong CA objects, because of their more ex-
tensive and powerful associations, should elicit increased phase
locking between the core contextual and visual regions compared
with weak CA objects. Finally, the timing of phase locking in the
context network directly tests whether contextual information is
extracted early enough to facilitate recognition, where such fa-
cilitation is useful (6).
Behavioral Results.The mean response time (RT) to recognize the
objects in strong CA (SCA) and weak CA (WCA) conditions was
very similar (SCA = 778 ± 504 ms and WCA = 773 ± 477 ms;
paired samples P > 0.82). Each subject’s RTs were screened for
outliers (2 SD above the mean) within condition. This resulted in
4.1% of trials being eliminated from the grand mean computa-
tion (3.9% for SCA and 4.3% for WCA). One subject’s RT
means were a group outlier (2.47 and 2.54 SD above the group
mean for SCA and WCA, respectively); however, excluding this
participant’s RT did not alter the relationship between SCA and
WCA RT (623 ± 158 ms and 622 ± 205 ms, respectively). Par-
ticipants were only asked to respond when they recognized the
object, and no verification or feedback as to the correctness of
the response was provided; the task was not designed to test
contextual facilitation. However, these results are comparable
with the RTs obtained with a different cohort of subjects using
a largely overlapping set of stimuli and the same task, where the
SCA and WCA RTs were likewise statistically equivalent (SCA =
702 ms and WCA = 684 ms, P = 0.76) (8). Therefore, any dif-
ferences in the neural activity between SCA and WCA conditions
could not be ascribed to RT and task performance differences.
fMRI Results. A set of ROIs (PHC, RSC, and MPFC) that we had
identified a priori based on activations of contextual association
processing in previous studies (8, 10, 27) showed significantly
greater activation for the SCA > WCA contrast in this experi-
ment, replicating the previous studies (Fig. S1). Given our
a priori selection of the ROIs, we used the statistical threshold of
P < 0.001, uncorrected, with 5-voxel extent threshold and whole-
brain t1,14= 3.79, which is slightly more conservative than the
thresholding parameters that optimally balance Type I and Type
II errors in fMRI activations (P < 0.005, uncorrected) with 10-
voxel extent threshold (52). The following foci were activated
above the threshold and are reported here as the group statistical
peak of activation in average Montreal Neurological Institute
(MNI) space coordinates: PHC (−30, −30, and −26), RSC (−10,
−46, and 14; −8, −58, and 26), and MPFC (−14, 58, and −10).
Other major activation clusters included the right hippocampus
(34, −30, and −18), right RSC (16, −56, and 18), left fusiform
cortex (−36, −42, and −14), left middle occipital gyrus (−28,
−82, and −2), left (−40, 14, and −2) and right (40, 20, and 2)
insular cortex, and left middle temporal gyrus (−68, −14, and 2).
The context network ROIs in the left hemisphere were used to
anatomically guide the selection of the MEG context network
ROIs used for phase-locking analysis, as described in ROI Se-
lection and Extraction in Methods. The fMRI methods are de-
scribed in SI Methods.
MEG Phase-Locking Results. There was stronger phase locking in
the lower β-band for the core context regions for the SCA vs.
WCA conditions (Fig. 2). Specifically, there was significantly
greater phase locking in the SCA condition, occurring in the
lower β-band between early visual regions in the occipital cortex
and PHC (P = 0.02; all MEG phase-locking P values corrected
for multiple time frequency point comparisons as described in SI
Methods). This phase locking between the occipital cortex and
PHC began early, peaking between 150 and 200 ms at ∼15 Hz.
Another small phase-locking cluster higher in the β-band (∼22
Hz) was significant (P = 0.04) from 200 to 220 ms. Beginning
slightly later and overlapping with the occipital cortex–PHC
phase locking, PHC and RSC showed significantly greater (P =
0.03) phase locking in the SCA condition in the lower β-band
(13–15 Hz), peaking between 170 and 240 ms. Phase locking
between the occipital cortex and RSC was enhanced in the SCA
condition in the later period, peaking between 310 and 360 ms
(P = 0.03) at ∼15 Hz. Phase locking of the third region in the
context network, MPFC, with these regions was much weaker.
PHC–MPFC did not exhibit significant phase locking (P = 0.62),
whereas RSC and MPFC showed significantly greater phase
locking in the upper β-range only, centered at ∼22 Hz, which
occurred in the later period, between 370 and 400 ms (P = 0.05).
Last, the occipital cortex and MPFC did not show significantly
different phase locking (P = 0.27). We also computed the phase
angles for the significant time frequency periods in our phase-
locking ROI comparisons for each subject and found that, in
The Context Network
typically activated by SCA displayed on the left hemisphere of the inflated
brain surface. RSC, retrosplenial cortex; PHC, parahippocampal cortex; MPFC,
medial prefrontal cortex. (B) Examples of stimuli in the SCA and WCA con-
ditions. The SCA objects were rated as the most typical item of a particular
context (e.g., construction or kitchen) by a separate survey of 35 participants.
The WCA objects were rated as weakly associated with many different
contexts but not strongly or consistently associated with any one context by
another survey of 18 participants. The spatial frequency content of the
stimuli in these two conditions did not differ (Fig. S2).
Context network and stimuli. (A) The network of cortical regions
| www.pnas.org/cgi/doi/10.1073/pnas.1013760108 Kveraga et al.
virtually all cases, they were significantly different from 0° or 180°,
as were the mean group phase angles. These results are reported
in detail in SI Methods.
In contrast to the phase-locking data, local power did not
differ between conditions in either the early or late time fre-
quency bands of interest (tfBOI) in any of the context regions, as
reported in detail in SI Methods. In addition, a trial shuffle
analysis (SI Methods) showed that a significant residual phase
locking remained after accounting for any potential coincidental
phase locking to stimulus onset in the ROIs (49). Specifically,
these analyses showed significant residual induced phase locking
in the SCA condition between the occipital cortex and PHC
from ∼150 to 200 ms (P = 0.014) at ∼15 Hz and from ∼200 to 220
ms at ∼22 Hz (P = 0.046), between PHC and RSC from ∼170
to 240 ms (P = 0.003), and between the occipital cortex and RSC
from ∼310 to 360 ms (P = 0.001).
Finally, to examine whether this phase locking was specific to
CA processing and the core context regions and not any object
recognition-related processes and regions, we also tested phase
locking between regions that we have previously observed to be
activated in both fMRI and MEG during object recognition (41,
43, 44). Specifically, we additionally examined phase locking
between the fusiform and orbitofrontal cortex and between these
regions and the occipital cortex, PHC, RSC, and MPFC. Criti-
cally, although these pair-wise comparisons showed increased
early phase locking relative to a prestimulus baseline, this phase
locking did not differ between SCA and WCA conditions. This is
consistent with the more general role that the fusiform and
orbitofrontal cortices play in achieving and facilitating object
recognition rather than the more specialized task of activating
and processing the contextual associations of visual stimuli,
which is attributed to the PHC, RSC, and MPFC.
The primary goal of this study was to examine the temporal
dynamics of the contextual association network and the com-
munication among its components (PHC, RSC, and MPFC). We
also examined the interaction of the context network regions
with other regions involved in visual recognition (occipital, fu-
siform, and orbitofrontal cortex). We had four main findings. (i)
We found significantly greater phase locking for objects with
SCA than those with WCA between the PHC, RSC, and occip-
ital cortex. (ii) This significant increase in phase locking occurred
primarily in the lower β-band, beginning and peaking early
(∼150–220 ms from stimulus onset) between the occipital cortex
and PHC in the lower (∼15 Hz) and higher (∼22 Hz) β-band
followed by and overlapping with PHC–RSC phase locking that
occurred between 170 and 240 ms and ∼13 Hz. (iii) Greater
phase locking between the occipital cortex and RSC started later,
peaking between 310 and 360 ms also in the lower β-band (∼15
Hz). (iv) The medial prefrontal cortex showed phase locking only
with the RSC, which was weaker, relatively late (∼370–400 ms)
and in the higher β-band (∼22 Hz).
The pattern of phase-locking enhancement that we found for
SCA stimuli compared with WCA stimuli suggests that contex-
tual information is extracted quite early during the course of
object recognition. Indeed, the timing of the interactions be-
tween the visual cortex and PHC and between the PHC and RSC
(Fig. 2A) coincides with the time period when object recognition
is believed to occur (43, 53–55). This early contextual activation
is at odds with the proposal of a recent study that asserted that
context-related activity in the PHC reflects explicit post-
recognition imagining of place information (56). Instead, fMRI
studies of the respective roles of the PHC and RSC suggest that
the PHC is involved in extracting contextual associations from
specific instances of stimuli, whereas the RSC invokes a more
abstract, prototypical representation of that particular context (8,
26, 32). This period of phase locking between the PHC and RSC,
thus, may reflect the translation of stimulus-specific contextual
associations into a more generic context representation. Although
our paradigm did not explicitly test or require contextual facili-
tation, the early context-based activation that we have observed is
in agreement with the idea that contextual information is extrac-
ted and processed early enough for facilitating object recognition.
Later, the RSC begins phase locking with the occipital cortex
at around 300 ms from stimulus onset. Generally, neural activity
after about 200–300 ms has been associated predominantly with
feedback processes (57, 58). Therefore, this relatively late phase
locking between the RSC and occipital cortex may reflect feed-
back from the RSC to early visual areas. Such feedback could
be seen as predictions that bias processing in the early visual
regions to facilitate perception of subsequent stimuli from the
The relatively late (370–400 ms) RSC–MPFC phase locking
was relatively weak and therefore, should be interpreted with
caution. Nevertheless, the timing (around the N400) and pre-
frontal involvement are consistent with reports in the event-
related potential (ERP) literature ascribing such activations to
integration of semantic information into the current context (59–
62). In particular, the neural response 400 ms after a stimulus
context regions and early visual cortex. Left is the ROI pairs showing sig-
nificantly greater phase locking for SCA vs. WCA conditions. Vis, occipital
cortex; PHC, parahippocampal cortex; RSC, retrosplenial complex; MPFC,
medial prefrontal cortex. (Right) The color-map P values in the maps rep-
resent significance of univariate paired samples t tests between the phase-
locking values for SCA and WCA conditions across subjects. The maps are not
smoothed. The P values in green next to the largest clusters in each map
show the statistical significance of the cluster while correcting for multiple
comparisons. The correction for multiple comparisons was performed by
using a nonparametric cluster permutation test (84), which is described in
detail in SI Methods. (A) Significantly greater phase locking was seen in the
early (100–250 ms) period between Vis and the PHC and between the PHC
and RSC in the SCA condition relative to the WCA condition. (B) Significantly
greater late (250–400 ms) phase locking was seen between the Vis and RSC
and between the RSC and MPFC in the SCA condition relative to the WCA
condition. Fig. S3 also shows the reliability of MEG signal extraction from the
context network regions.
Significant phase locking for SCA vs. WCA conditions for the core
Kveraga et al. PNAS
| February 22, 2011
| vol. 108
| no. 8
steadily declines for consecutive objects or words that are con-
textually and semantically congruent (61). When a lexical or vi-
sual stimulus is semantically incongruent with the given context,
anterior brain activity is usually more negative andposterior brain
activity is more positive than for semantically congruent stimuli
(60–62). Furthermore, the MPFC in particular has been impli-
cated in the integration of semantic associations into a subjective
context (17). In that study, participants were asked during
a memorization task to indicate whether they remembered
a given stimulus by associating it with some subjective experience
(an association trial) or by noting some visual feature of the
stimulus (a feature trial). They found that the MPFC (along with
the PHC/hippocampus and RSC) showed greater activation for
association trials than for feature trials. The stimuli encoded by
forming an association were also better remembered (17). Thus,
we propose that the late phase locking between the RSC and
MPFC may represent the integration of the context evoked by
a particular object into the current context frame.
Importantly, the timing of the phase-locking enhancement and
the task in which the participants engaged (simple stimulus
recognition, which did not explicitly require the activation of its
context) strongly suggest that contextual associations are evoked
early and automatically during object recognition rather than
strictly as a postperceptual process. The timing of the significant
phase locking in our study and the frequency bands in which they
occurred are consistent with MEG and scalp or intracranial
electroencephalography (EEG) studies investigating object rec-
ognition processes (43, 55, 63–67). Bar et al. (43) found greater
phase locking between the occipital and orbitofrontal cortex and
later, between the orbitofrontal and fusiform cortex when stimuli
were successfully recognized rather than unrecognized. The
same pattern of phase-locking enhancement was observed when
the stimuli were low-pass filtered rather than high-pass filtered
(magnocellular-biased low-spatial frequency stimuli have been
shown to facilitate object recognition through the orbitofrontal
cortex) (41, 43, 44). Similarly, Ghuman et al. (55) recently
showed that facilitation of visual object recognition using repe-
tition resulted in greater phase locking between the dorsal pre-
frontal and fusiform cortex, occurring between 190 and 270 ms in
the low (∼14 Hz) β-frequency band. Additionally, the earlier that
this phase locking peaked, the more object recognition was fa-
cilitated. This is the same time frequency band in which SCA
objects showed greater visual–PHC and PHC–RSC phase lock-
ing in the current study.
Moreover, the anatomical regions between which MEG phase
locking was enhanced for the SCA vs. WCA objects are consistent
with a host of fMRI studies exploring the neural substrates of
contextual association processing (8, 10, 17, 27, 56). They also
overlap substantially with the default mode network, as discussed
in detail in Bar et al. (68). Note that no significant differences in
phase locking were found between those regions and general ob-
the latter regions have exhibited significant phase-locking differ-
ences when compared under different conditions in two separate
experiments (recognized vs. unrecognized and low vs. high spatial
frequency objects stimuli) (43). This shows that significant phase-
locking differences are task-, stimulus-, and region-specific.
What implications do these findings have for the role of con-
text in visual object recognition? The long-running debate about
the role of context in vision has split along two main types of
models: (i) the so-called interactive models, which posit that
context exerts influence on and can facilitate object recognition
either by scene constraints on object processing (i.e., the per-
ceptual schema models) (3, 69) or priming memory representa-
tions of context-congruent objects (i.e., priming models) (1, 4,
70), and (ii) the functional isolation model (6, 71), which main-
tains that object recognition and context activation are func-
tionally separate, noninteracting processes. The conclusions that
we can draw regarding the contextual facilitation of object rec-
ognition are somewhat limited, because our paradigm was not
designed to test directly whether context (e.g., supplied by
a contextually congruent prime scene or word label) can facilitate
recognition of a subsequently shown object. Nevertheless, the
findings in this study indicate that neural communication in the
context network, particularly interactions that involved PHC,
differentiated between SCA and WCA objects early enough to
facilitate recognition and hundreds of milliseconds before their
recognition was reported through a behavioral response.
An alternative explanation for our findings could be that the
SCA and WCA stimuli were somehow physically different and
that these low-level differences accounted for the observed dif-
ferential phase locking in the context regions. However, this is
unlikely for several reasons. First, we analyzed the spatial fre-
quency content of the stimuli in the SCA and WCA conditions
and found no significant physical differences between the stimuli
in these conditions (Fig. S2). Moreover, if the physical differ-
ences had accounted for the phase-locking differences between
the SCA and WCA conditions, we would expect these differ-
ences to manifest themselves mostly in the low-level visual pro-
cessing areas and perhaps, low- and intermediate-level object
form regions in the ventral temporal lobe rather than in higher-
level association regions. However, in contrast, phase synchrony
differed significantly only when the relatively high-level regions
specifically implicated in contextual association processing (i.e.,
PHC, RSC, and MPFC) were involved, and we found no sig-
nificant phase-locking differences otherwise. This phase-locking
sequence preferentially occurs when the stimuli with SCA bind
this network together.
To conclude, the early phase-locking enhancement that we
found for objects with SCA indicates that contextual information
is activated early during object recognition rather than solely as
a late postperceptual process. Such rapid activation of contextual
if one thinks of the evolutionary pressures faced by most organ-
isms. For example, seeing a paw print or scat of a predator and
rapidly activating the context associated with this image could
afford the prey animal enough time for an escape, conferring it
a significant evolutionary advantage over time. The brain con-
stantly generates predictions about what it is about to perceive
next (72), and activating contextual associations early during
where it is most likely to yield biologically important information.
Participants. Fifteen subjects participated in the fMRI study (all right-handed;
eight males). Nine of fifteen subjects (five males) also participated in the MEG
experiment. No participants were excluded from the study. The age range
was 22–39 y, with the mean age of 25.5 y and SD of 5.5 y. All had normal or
corrected to normal vision. Their informed written consent was obtained
according to the procedures of the Massachusetts General Hospital In-
stitutional Review Board and approved under Human Studies Protocols
2000P-000949 and 2002P-002035.
Stimuli and Task. We used color pictures of everyday objects in the two
conditions of interest (SCA and WCA conditions). The object stimuli in the
conditions of interest (i.e., objects with SCA and WCA) were selected on the
basis of survey ratings and had been used previously (8). Briefly, the SCA and
WCA items were obtained by conducting separate surveys with a separate
population of subjects. All SCA objects were rated as the most typical objects
of a particular context (e.g., a stove for the kitchen context) by a group of 35
subjects. Ratings from another group of 18 subjects were used to compile
a list of objects that were not strongly associated with any particular context
but rather, weakly associated with many different contexts (examples in Fig.
1B). Although one might wish for objects without any contextual associa-
tions as a comparison condition, in practice, real objects never appear in
isolation and thus, are not devoid of any contextual associations. Therefore,
the comparison in this study was between objects that should activate the
context with which they are strongly associated and objects that should not
| www.pnas.org/cgi/doi/10.1073/pnas.1013760108Kveraga et al.
strongly activate any particular context but may weakly activate many
contexts (8). In the third condition, which was included for other purposes
and not analyzed in this study, abstract 2D shapes were shown. All stimuli
were displayed on a gray background and spanned a maximum of 9.2° of
visual angle. Images were displayed using the stimulus presentation package
Psychtoolbox (73) running in Matlab (Mathworks) on a Macintosh Power
Mac G4. The conditions comprised 53 pictures each, which were repeated
three times, one in each of three sessions. Stimulus pictures appeared on the
screen for 1,700 ms with a 300-ms interstimulus interval. The stimulus onsets
were also randomly jittered by 0–300 ms. During MEG recording, the stimuli
were projected onto a translucent screen positioned in front of the seated
participant with the display resolution of 1,024 × 768 pixels and a refresh
rate of 75 Hz using an LP350 DLP projector (InFocus). Subjects were
instructed to report when they recognized the object (in the conditions of
interest) or report that the stimulus was an abstract shape. No feedback was
given after the response, and subjects were unaware of the purpose of
MEG Methods. The MEG was acquired with a 306-channel Neuromag Vec-
torview whole-head system (Elekta Neuromag Oy) comprising 204 orthog-
onally oriented planar gradiometers and 102 magnetometers at 102
locations. The system was housed in a three-layer magnetically shielded room
(ImedcoAG).Tocompute theheadpositioninside theMEGdewar,four head-
position indicator (HPI) electrodes were affixed to the subject’s head. The
positions of the HPI electrodes on the head as well as those of multiple
points on the scalp were entered with a magnetic digitizer (Polhemus
FastTrack 3D) in a head coordinate frame defined by anatomical landmarks,
which included the nasion and the left and right auricular points. Eye blinks
were monitored with four electrooculogram (EOG) sensors positioned above
and beside the subjects’ eyes. The MEG, EOG, and HPI sensors were sampled
at 600 Hz, band pass-filtered online in the range of 0.1–200 Hz, and stored
for offline analysis. Running online averages were computed to monitor and
note noisy or nonfunctioning channels.
MEG data were analyzed with the MNE software package (Hämäläinen).
Both the gradiometer and magnetometer data were included in the analysis.
First, MEG data were screened for eye blink artifacts and low pass-filtered at
40 Hz. Trials that exceeded particular thresholds and noisy or nonfunction-
ing (flat) channels were eliminated from further analyses. A high-resolution
structural MRI, acquired on a Siemens Allegra 3T scanner (Siemens Medical
Solutions), was used to construct a forward model and visualize MEG sources.
The MRI and MEG coordinates frames were coregistered using the HPI and
head surface points. A single-layer boundary element model (BEM) (74) was
constructed from the anatomical MRI, and it was used as a forward model to
constrain MEG source location to the cortex and compute the minimum
norm estimate (75) inverse solution. MEG sources were visualized on the
inflated surface of the MRI (76) as dynamic statistical parametric maps (77).
ROI Selection and Extraction. We selected our MEG ROIs with anatomical
guidance from our fMRI context network activation foci in this study, which
weresimilartothose foundinearlierfMRI studies ofcontextualprocessing(8,
27, 78). These regions included PHC, RSC, and MPFC in the left hemisphere
and were verified to be reliably detectable and separable using MEG (Fig.
S3). In addition, to investigate the communication between these contextual
association regions and early visual regions, we also extracted raw currents
from the occipital cortex site comprising ventral, medial, and polar occipital
cortex. Last, two additional ROIs that had exhibited phase locking in our
previous MEG experiments investigating top-down facilitation processes in
object recognition (43) in fusiform and orbitofrontal cortex were included as
control sites. In addition to anatomical guidance provided by the fMRI
activations of the context network in this study, we also used functional
constraints, which were applied to an independent dataset as follows. We
used our subjects’ second run of trials to display MEG activations, which
were noise-normalized dynamic statistical parametric maps (dSPMs), on each
individual’s inflated brain image using SCA vs. baseline comparison as our
selection contrast. We drew the ROI boundaries that encompassed the peak
context network activations in our fMRI experiment around MEG dSPM
activations that exceeded the uncorrected dSPM threshold of P < 0.05 at any
time between 100 and 450 ms. We then extracted raw, minimally filtered
(0.1–200 Hz) MEG currents from these ROIs in the first run of trials. Thus, the
data to be used in phase-locking analysis of the contrast of interest (SCA vs.
WCA objects) were (i) not functionally defined by this contrast and (ii)
extracted from an independent subset of the data. Last, the interareal phase
locking is, at least in principle, unrelated to the dSPM values, showing the
independence of this ROI selection (79). The extracted raw currents from
each ROI were submitted to phase-locking analysis as described below.
tfBOI Selection. Interregional communication in the upper α- (10–12 Hz) and
particularly, lower β-bands (13–25 Hz) has been found to be involved in
object recognition, showing increased phase synchrony when object recog-
nition is successful (43, 51, 55, 63–67). The α-band along with the θ-band
have been thought to be involved in memory functions (50); however, the
lower α-band (8–9 Hz) has been found to reflect activity associated with
phase-resetting, idling, or state of readiness changes, and its current role in
cognitive function is unclear (80). The β-band, particularly the lower β-band,
has been implicated in playing an important role in the synchronization of
brain activity during object recognition (review in ref. 80). Therefore, we
focused on the upper α- and lower β-bands (10–25 Hz) in our phase-locking
analyses. Because processes associated with object recognition and associa-
tive processing have been reported to occur in the poststimulus onset range
of 100–450 ms (43, 55, 64, 81, 82), we constrained our statistical phase-
locking analyses to the range of 100–450 ms, subdividing it into early (100–
250 ms) and later (250–400 ms) periods.
Phase-Locking Analysis. Phase locking is a method that directly assesses the
timing of oscillatory activity of the brain, irrespective of its amplitude (49, 83).
The waveforms from the ROIs were filtered with a continuous Morlet
wavelet transform of width 6. For the phase-locking analyses, the phase of
each waveform was extracted and averaged in each trial for every frequency
and time point of interest. The formula of the phase-locking value (PLV) was
computed using the formula below (Eq. 1):
Here, t is the time point, n is the trial number, and ɸðt;nÞ is the phase of the
MEG waveforms at a given time and trial in a pair of ROIs tested. The PLV
measures the variability across trials of the relative phases of the two signals
and can range from 0 (no phase locking) to 1 (complete phase locking). This
method is functionally equivalent to the dynamic statistical parametric
mapping, with the order of the wavelet filtering and the application of the
inverse solution reversed relative to that method (83). We describe our
statistical analyses of phase locking and power and report phase angles in
ACKNOWLEDGMENTS. This research was supported by National Institute of
Mental Health Grant K01-MH084011 (to K.K.), National Institutes of Health
Grants NS50615 and EY019477 (to M.B.), and National Science Foundation
Grant 0842947 (to M.B.).
1. Palmer SE (1975) The effects of contextual scenes on the identification of objects.
Mem Cognit 3:519–526.
2. Potter MC, Faulconer BA (1975) Time to understand pictures and words. Nature 253:437–438.
3. Biederman I, Mezzanotte RJ, Rabinowitz JC (1982) Scene perception: Detecting and
judging objects undergoing relational violations. Cognit Psychol 14:143–177.
4. Bar M, Ullman S (1996) Spatial context in recognition. Perception 25:343–352.
5. Davenport JL, Potter MC (2004) Scene consistency in object and background perception.
Psychol Sci 15:559–564.
6. Hollingworth A, Henderson JM (1998) Does consistent scene context facilitate object
perception? J Exp Psychol Gen 127:398–415.
7. Henderson JM, Hollingworth A (1999) High-level scene perception. Annu Rev Psychol
8. Bar M, Aminoff E (2003) Cortical analysis of visual context. Neuron 38:347–358.
9. Hayes SM, Ryan L, Schnyer DM, Nadel L (2004) An fMRI study of episodic memory:
Retrieval of object, spatial, and temporal information. Behav Neurosci 118:885–896.
10. Aminoff E, Gronau N, Bar M (2007) The parahippocampal cortex mediates spatial and
nonspatial associations. Cereb Cortex 17:1493–1503.
11. Park S, Intraub H, Yi DJ, Widders D, Chun MM (2007) Beyond the edges of a view:
Boundary extension in human scene-selective visual cortex. Neuron 54:335–342.
12. Hayes SM, Nadel L, Ryan L (2007) The effect of scene context on episodic object
recognition: Parahippocampal cortex mediates memory encoding and retrieval success.
13. Diana RA, Yonelinas AP, Ranganath C (2008) High-resolution multi-voxel pattern
analysis of category selectivity in the medial temporal lobes. Hippocampus 18:536–541.
14. Szpunar KK, Chan JC, McDermott KB (2009) Contextual processing in episodic future
thought. Cereb Cortex 19:1539–1548.
15. Aminoff E, Schacter DL, Bar M (2008) The cortical underpinnings of context-based
memory distortion. J Cogn Neurosci 20:2226–2237.
16. Gronau N, Neta M, Bar M (2008) Integrated contextual representation for objects’
identities and their locations. J Cogn Neurosci 20:371–388.
Kveraga et al. PNAS
| February 22, 2011
| vol. 108
| no. 8
17. Peters J, Daum I, Gizewski E, Forsting M, Suchan B (2009) Associations evoked during
memory encoding recruit the context-network. Hippocampus 19:141–151.
18. Litman L, Awipi T, Davachi L (2009) Category-specificity in the human medial
temporal lobe cortex. Hippocampus 19:308–319.
19. Aguirre GK, Detre JA, Alsop DC, D’Esposito M (1996) The parahippocampus subserves
topographical learning in man. Cereb Cortex 6:823–829.
20. Epstein R, Kanwisher N (1998) A cortical representation of the local visual
environment. Nature 392:598–601.
21. Epstein R, Harris A, Stanley D, Kanwisher N (1999) The parahippocampal place area:
Recognition, navigation, or encoding? Neuron 23:115–125.
22. Gabrieli JDE, Keane MM, Zarella MM, Poldrack RA (1997) Preservation of implicit
memory for new associations in global amnesia. Psychol Sci 8:326–329.
23. Wagner AD, et al. (1998) Building memories: Remembering and forgetting of verbal
experiences as predicted by brain activity. Science 281:1188–1191.
24. Ranganath C, et al. (2004) Dissociable correlates of recollection and familiarity within
the medial temporal lobes. Neuropsychologia 42:2–13.
25. Daselaar SM, Fleck MS, Cabeza R (2006) Triple dissociation in the medial temporal
lobes: Recollection, familiarity, and novelty. J Neurophysiol 96:1902–1911.
26. Bar M (2004) Visual objects in context. Nat Rev Neurosci 5:617–629.
27. Bar M, Aminoff E, Schacter DL (2008) Scenes unseen: The parahippocampal cortex
intrinsically subserves contextual associations, not scenes or places per se. J Neurosci
28. Maguire EA (2001) The retrosplenial contribution to human navigation: A review of
lesion and neuroimaging findings. Scand J Psychol 42:225–238.
29. Svoboda E, McKinnon MC, Levine B (2006) The functional neuroanatomy of
autobiographical memory: A meta-analysis. Neuropsychologia 44:2189–2208.
30. Henderson JM, Larson CL, Zhu DC (2008) Full scenes produce more activation than
close-up scenes and scene-diagnostic objects in parahippocampal and retrosplenial
cortex: An fMRI study. Brain Cogn 66:40–49.
31. Vann SD, Aggleton JP, Maguire EA (2009) What does the retrosplenial cortex do? Nat
Rev Neurosci 10:792–802.
32. Park S, Chun MM (2009) Different roles of the parahippocampal place area (PPA) and
retrosplenial cortex (RSC) in panoramic scene perception. Neuroimage 47:1747–1756.
33. Raichle ME, et al. (2001) A default mode of brain function. Proc Natl Acad Sci USA 98:
34. Buckner RL, Carroll DC (2007) Self-projection and the brain. Trends Cogn Sci 11:49–57.
35. Mason MF, et al. (2007) Wandering minds: The default network and stimulus-
independent thought. Science 315:393–395.
36. Thorpe SJ, Rolls ET, Maddison S (1983) The orbitofrontal cortex: Neuronal activity in
the behaving monkey. Exp Brain Res 49:93–115.
37. Tanaka K (1996) Inferotemporal cortex and object vision. Annu Rev Neurosci 19:
38. Logothetis NK, Sheinberg DL (1996) Visual object recognition. Annu Rev Neurosci 19:
39. Rolls ET (2000) Functions of the primate temporal lobe cortical visual areas in
invariant visual object and face recognition. Neuron 27:205–218.
40. Haxby JV, et al. (2001) Distributed and overlapping representations of faces and
objects in ventral temporal cortex. Science 293:2425–2430.
41. Bar M, et al. (2001) Cortical mechanisms specific to explicit visual object recognition.
42. Rolls ET (2004) The functions of the orbitofrontal cortex. Brain Cogn 55:11–29.
43. Bar M, et al. (2006) Top-down facilitation of visual recognition. Proc Natl Acad Sci USA
44. Kveraga K, Boshyan J, Bar M (2007) Magnocellular projections as the trigger of top-
down facilitation in recognition. J Neurosci 27:13232–13240.
45. Singer W (1999) Neuronal synchrony: A versatile code for the definition of relations?
46. Varela F, Lachaux JP, Rodriguez E, Martinerie J (2001) The brainweb: Phase
synchronization and large-scale integration. Nat Rev Neurosci 2:229–239.
47. Buzsáki G, Draguhn A (2004) Neuronal oscillations in cortical networks. Science 304:
48. Fries P (2005) A mechanism for cognitive dynamics: Neuronal communication through
neuronal coherence. Trends Cogn Sci 9:474–480.
49. Lachaux JP, Rodriguez E, Martinerie J, Varela FJ (1999) Measuring phase synchrony in
brain signals. Hum Brain Mapp 8:194–208.
50. Klimesch W, Freunberger R, Sauseng P, Gruber W (2008) A short review of slow phase
synchronization and memory: Evidence for control processes in different memory
systems? Brain Res 1235:31–44.
51. Düzel E, Neufang M, Heinze HJ (2005) The oscillatory dynamics of recognition
memory and its relationship to event-related responses. Cereb Cortex 15:1992–2002.
52. Lieberman MD, Cunningham WA (2009) Type I and Type II error concerns in fMRI
research: Re-balancing the scale. Soc Cogn Affect Neurosci 4:423–428.
53. VanRullen R, Thorpe SJ (2001) The time course of visual processing: From early
perception to decision-making. J Cogn Neurosci 13:454–461.
54. Rousselet GA, Macé MJ, Thorpe SJ, Fabre-Thorpe M (2007) Limits of event-related
potential differences in tracking object processing speed. J Cogn Neurosci 19:
55. Ghuman AS, Bar M, Dobbins IG, Schnyer DM (2008) The effects of priming on frontal-
temporal communication. Proc Natl Acad Sci USA 105:8405–8409.
56. Epstein RA, Ward EJ (2010) How reliable are visual context effects in the
parahippocampal place area? Cereb Cortex 20:294–303.
57. Tomita H, Ohbayashi M, Nakahara K, Hasegawa I, Miyashita Y (1999) Top-down signal
from prefrontal cortex in executive control of memory retrieval. Nature 401:699–703.
58. Garrido MI, Kilner JM, Kiebel SJ, Friston KJ (2007) Evoked brain responses are
generated by feedback loops. Proc Natl Acad Sci USA 104:20961–20966.
59. Düzel E, et al. (1999) Task-related and item-related processes in episodic and semantic
retrieval: A combined PET and ERP study. Proc Natl Acad Sci USA 96:1794–1799.
60. Ganis G, Kutas M (2003) An electrophysiological study of scene effects on object
identification. Brain Res Cogn Brain Res 16:123–144.
61. Hagoort P (2008) The fractionation of spoken language understanding by measuring
electrical and magnetic brain signals. Philos Trans R Soc Lond B Biol Sci 363:
62. Sitnikova T, Holcomb PJ, Kiyonaga KA, Kuperberg GR (2008) Two neurocognitive
mechanisms of semantic integration during the comprehension of visual real-world
events. J Cogn Neurosci 20:2037–2057.
63. Tarokh L (2009) EEG oscillations induced by contour closure in a noisy visual field.
Behav Brain Res 196:55–62.
64. Sehatpour P, et al. (2008) A human intracranial study of long-range oscillatory
coherence across a frontal-occipital-hippocampal brain network during visual object
processing. Proc Natl Acad Sci USA 105:4399–4404.
65. Okazaki M, Kaneko Y, Yumoto M, Arima K (2008) Perceptual change in response to
a bistable picture increases neuromagnetic beta-band activities. Neurosci Res 61:
66. Supp GG, et al. (2005) Semantic memory retrieval: Cortical couplings in object
recognition in the N400 window. Eur J Neurosci 21:1139–1143.
67. von Stein A, Rappelsberger P, Sarnthein J, Petsche H (1999) Synchronization between
temporal and parietal cortex during multimodal object processing in man. Cereb
68. Bar M, Aminoff E, Mason M, Fenske M (2007) The units of thought. Hippocampus 17:
69. Boyce SJ, Pollatsek A (1992) Identification of objects in scenes: The role of scene
background in object naming. J Exp Psychol Learn Mem Cogn 18:531–543.
70. Ullman S (1996) High-Level Vision: Object Recognition and Visual Cognition (MIT
Press, Cambridge, MA).
71. Hollingworth A, Henderson JM (1999) Object identification is isolated from scene
semantic constraint: Evidence from object type and token discrimination. Acta
Psychol (Amst) 102:319–343.
72. Bar M (2007) The proactive brain: Using analogies and associations to generate
predictions. Trends Cogn Sci 11:280–289.
73. Pelli DG (1997) The VideoToolbox software for visual psychophysics: Transforming
numbers into movies. Spat Vis 10:437–442.
74. Hämäläinen MS, Sarvas J (1989) Realistic conductivity geometry model of the human
head for interpretation of neuromagnetic data. IEEE Trans Biomed Eng 36:165–171.
75. Hämäläinen MS, Hari R (2002) Magnetoencephalographic characterization of
dynamic brain activation: Basic principles and methods of data collection and source
analysis. Brain Mapping: The Methods, eds Toga AW, Mazziotta JC (Academic, San
Diego), 2nd Ed, pp 227–254.
76. Fischl B, Sereno MI, Dale AM (1999) Cortical surface-based analysis. II: Inflation,
flattening, and a surface-based coordinate system. Neuroimage 9:195–207.
77. Dale AM, et al. (2000) Dynamic statistical parametric mapping: Combining fMRI and
MEG for high-resolution imaging of cortical activity. Neuron 26:55–67.
78. Bar M, Aminoff E, Ishai A (2008) Famous faces activate contextual associations in the
parahippocampal cortex. Cereb Cortex 18:1233–1238.
79. Kriegeskorte N, Simmons WK, Bellgowan PS, Baker CI (2009) Circular analysis in
systems neuroscience: The dangers of double dipping. Nat Neurosci 12:535–540.
80. Uhlhaas PJ, Haenschel C, Nikolić D, Singer W (2008) The role of oscillations and
synchrony in cortical networks and their putative relevance for the pathophysiology
of schizophrenia. Schizophr Bull 34:927–943.
81. Vanni S, Revonsuo A, Saarinen J, Hari R (1996) Visual awareness of objects correlates
with activity of right occipital cortex. Neuroreport 8:183–186.
82. Tallon-Baudry C, Bertrand O (1999) Oscillatory gamma activity in humans and its role
in object representation. Trends Cogn Sci 3:151–162.
83. Lin FH, et al. (2004) Spectral spatiotemporal imaging of cortical oscillations and
interactions in the human brain. Neuroimage 23:582–595.
84. Maris E, Oostenveld R (2007) Nonparametric statistical testing of EEG- and MEG-data.
J Neurosci Methods 164:177–190.
| www.pnas.org/cgi/doi/10.1073/pnas.1013760108Kveraga et al.