University of PennsylvaniaScholarlyCommons

Neuroethics Publications Center for Neuroscience & Society

1-1-2010

Encoding Social Interactions: The NeuralCorrelates of True and False MemoriesBenjamin StraubePhilipps-University Marburg; University of Pennsylvania; RWTH Aachen University

Antonia GreenRWTH Aachen University

Anjan ChatterjeeUniversity of Pennsylvania, anjan@mail.med.upenn.edu

Tilo KircherPhilipps-University Marburg

Suggested Citation:Straube, B., Green, A. Chatterjee, A. and Kircher, T. (2010). Encoding Social Interactions: The Neural Correlates of True and False Memories. Journalof Cognitive Neuroscience. Vol. 23(2). p. 306-324.

© 2010 Massachusetts Institute of Technologyhttp://www.mitpressjournals.org/loi/jocn

This paper is posted at ScholarlyCommons. http://repository.upenn.edu/neuroethics_pubs/60For more information, please contact repository@pobox.upenn.edu.

Encoding Social Interactions: The Neural Correlates ofTrue and False Memories

Benjamin Straube1,2,3, Antonia Green3, Anjan Chatterjee2,and Tilo Kircher1

Abstract

■ In social situations, we encounter information transferred infirsthand (egocentric) and secondhand (allocentric) communica-tion contexts. However, themechanism bywhich an individual dis-tinguishes whether a past interaction occurred in an egocentricversus allocentric situation is poorly understood. This study exam-ined the neural bases for encoding memories of social interactionsthrough experimentally manipulating the communication context.During fMRI data acquisition, participantswatched video clips of anactor speaking andgesturing directly toward them (egocentric con-text) or toward an unseen third person (allocentric context). Afterscanning, a recognition task gauged participantsʼ ability to recog-nize the sentences they had just seen and to recall the context inwhich the sentences had been spoken. We found no differencesbetween the recognition of sentences spoken in egocentric and

allocentric contexts. However, when asked about the communica-tion context (“Had the actor directly spoken to you?”), participantstended to believe falsely that the actor had directly spoken to themduring allocentric conditions. Greater activity in the hippocampuswas related to correct context memory, whereas the ventral ACCwas activated for subsequent inaccurate context memory. For theinteraction between encoding context and context memory, weobserved increased activation for egocentric remembered itemsin the bilateral and medial frontal cortex, the BG, and the left pari-etal and temporal lobe. Our data indicate that memories of socialinteractions are biased to be remembered egocentrically. Self-referential encoding processes reflected in increased frontal activa-tion and decreased hippocampal activation might be the basis ofcorrect item but false context memory of social interactions. ■

INTRODUCTION

Memories of social interactions are crucial components ofhuman relationships. Nonverbal cues (such as gesture,body orientation, and eye gaze) are essential in social situa-tions: They allow an individual to recognize that he or sheis being directly addressed (e.g., Özyürek, 2002) and con-sequently define the egocentric or allocentric context inwhich communication occurs. Although the ability to recallif a past conversation occurred in an egocentric or allo-centric context is undoubtedly important for maintaininginterpersonal relationships, we know very little about theneural processes responsible for encoding the context ofa social communication. In this study, we investigate theneural encoding of the social context of spoken and ges-tural communication.

Gestures made with the hands and arms often have acommunicative function and frequently accompany speech(e.g., Kendon, 2004; Goldin-Meadow, 1999; McNeill, 1992).The neural basis of the processing such coverbal gestureshas been investigated by an increasing number of fMRIstudies (Dick, Goldin-Meadow, Hasson, Skipper, & Small,2009; Green et al., 2009; Kircher, Straube, et al., 2009;Straube, Green, Weis, Chatterjee, & Kircher, 2009; Holle,Gunter, Rüschemeyer, Hennenlotter, & Iacoboni, 2008;

Willems, Ozyurek, & Hagoort, 2007). The processing ofcoverbal gestures involves the left inferior frontal gyrus(IFG), inferior parietal cortex, posterior temporal regions,and precentral gyrus. Little is known about the neural pro-cesses responsible for creating memories of speech andgesture communication, but past research indicates thatleft inferior frontal, posterior temporal, premotor, and hip-pocampal brain regions are involved (Straube et al., 2009).Until now, there has been no evidence about how ego-centric versus allocentric contexts influence an individualʼsmemories of speech and gesture utterances.People sometimes believe they recognize things that

they have never actually encountered, for example, confus-ing a stranger with an old acquaintance. This phenomenon,known as false recognition, has been investigated in bothneuropsychological and neuroimaging research (for areview, see Schacter & Slotnick, 2004). False recognitionoften occurs in everyday social situations. Although severalstudies have compared the neural activity that accompa-nies genuine recognition of studied items (e.g., Kircheret al., 2008) and false recognition of novel items (e.g.,Slotnick & Schacter, 2004; Cabeza, Rao, Wagner, Mayer,& Schacter, 2001; Schacter, Reiman, et al., 1996), therehas been little research conducted on the true and falserecognition of social interactions. Past neuroimaging stud-ies have shown increased neural activity in regions of themedial and lateral frontal cortex during false recognition

1Philipps-University Marburg, 2University of Pennsylvania, 3RWTHAachen University

© 2010 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 23:2, pp. 306–324

(Slotnick & Schacter, 2004; Cabeza et al., 2001; Schacter,Buckner, Koutstaal, Dale, & Rosen, 1997; Schacter, Reiman,et al., 1996). In addition, patients with frontal lobe damagehave increased incidence of false recognition (e.g., Swick&Knight, 1999; Curran, Schacter, Norman, & Galluccio, 1997;Schacter, Curran, Galluccio, Milberg, & Bates, 1996). Thesestudies suggest that the frontal cortex plays an importantrole in false recognition. Other studies have shown the pari-etal lobe to be involved in both true and false recognition(for an overview, see Wagner, Shannon, Kahn, & Buckner,2005).In this study, we are especially interested in the encod-

ing processes that lead to accurate and inaccurate mem-ories of the nonverbal components of a communicationsituation. There is evidence that source memory (thememory of the context or surroundings in which onelearns a piece of information) is correlated with activity inthe hippocampal region (and possibly parahippocampalcortex) during encoding. In contrast, item memory (thememory of the information that is learned) is correlatedwith activity in the adjacent cortex (e.g., Kircher et al.,2008; Kensinger & Schacter, 2006; Ranganath et al., 2004;Davachi, Mitchell, & Wagner, 2003). Some studies havereported that item memory is correlated with widespreadactivity in the MTL and in the hippocampus (e.g., Stark &Okado, 2003; Otten, Henson, & Rugg, 2001; Kirchhoff,Wagner, Maril, & Stern, 2000). However, even during tasksthat predominantly depend on itemmemory, hippocampalactivation may help individuals associate newly learnedinformation with items already stored in memory (Kircheret al., 2007). Despite controversial findings about the spe-cific memory-related functions of the hippocampus andsurrounding brain regions, there is general agreement thatmedial-temporal structures are involved in the recognitionof source and item information. However, it is not known ifencoding related activity in the medial-temporal lobe pre-dicts subsequent memories of egocentric in comparisonwith allocentric communication situations.Beside the medial-temporal structures, frontal (e.g.,

Kirchhoff et al., 2000) and parietal brain regions (e.g.,Kensinger & Schacter, 2006) are active during encodingof items and their context. The neural correlates for theformation of false context memories are less well known.With regard to false memory encoding, it is generally as-sumed that cognitive processes engaged during encodinglead to subsequent inaccurate memories. For example,these errors may arise as a consequence of the similaritiesbetween how imagined and perceived events are encoded(Gonsalves et al., 2004; Gonsalves & Paller, 2000) as well assimilarities between the event features that are reactivatedduring retrieval (Johnson, Hashtroudi, & Lindsay, 1993;Johnson&Raye, 1981).With regard to thememory encodingof social interactions, attention processes that are elicited bysocial cues (such as body orientation, e.g., Hietanen, 2002,and eye gaze, e.g.,Hietanen,Nummenmaa,Nyman, Parkkola,& Hamalainen, 2006) may also play an important role insuccessful encoding and later remembering of contextual

information of the communication situations. Their rolemay be particularly important, as direct human-to-humaninteraction is an essential mode of communication yetcommonly results in false memories or misunderstandings.From a social psychologistʼs perspective, an egocentric biasand related self-referential encoding effects (the effect thatinformation relating to the self is preferentially encodedand organized above other types of information; Ross &Sicoly, 1979; Rogers, Kuiper, & Kirker, 1977) might be asource of biased memory encoding of social situations(for a review, see Schacter, Chiao,&Mitchell, 2003). Variousmedial cortical regions (including the medial orbital PFC,the ventromedial PFC, the sub- or pregenual ACC (PACC)and supragenual ACC [e.g., Kircher et al., 2001], the dorso-medial PFC, themedial parietal cortex, the posterior cingulatecortex, the retrosplenial cortex, and the precuneus [e.g.,Kircher et al., 2000]; for a review, see Gillihan & Farah,2005) have been found to be active during self-referentialprocessing (Northoff et al., 2006). These regions have beensubsumed under the term “cortical midline structures” andspeculatively characterized as an anatomical and functionalunit (Northoff & Bermpohl, 2004). Furthermore, one theorysuggests the existence of a self-memory system that is func-tionally dissociable from general semantic processing andevokes distinct cognitive operations (Macrae, Moran,Heatherton, Banfield, &Kelley, 2004). For example, in a studyby Macrae et al. (2004), participants evaluated the extent towhich a series of personality characteristics were self-descriptive. They found that activity in themedial PFC duringencoding of the characteristics predicted both subsequentmemory performance and judgments of self-relevance. Thisresult suggests that the advantage afforded to self-knowledgememory appears to depend on the additional recruitment ofthe medial frontal cortex (Macrae et al., 2004). However,besides this support for the theory that there is a specificself-memory system, the neural process underlying the for-mationof true and falsememories of social interactions is notwell known.

This study aimed to determine the neural correlates ofencoding processes for item memory (spoken sentencesaccompanied by gesture) and source memory (egocentricand allocentric contexts) of social interactions. During fMRIdata acquisition, participants were presented with videoclips of an actor speaking and gesturing directly towardthem (egocentric, E) or toward an unseen third person(allocentric, A; see Figure 1). After scanning, recognitionperformance for sentences (items) and communicationcontext (source) was gathered. Behaviorally, we expectedmore accurate memory of egocentric scenes (E) comparedwith allocentric scenes (A). We also anticipated that indi-viduals would more often misremember allocentric scenesthan they would egocentric scenes, as egocentric commu-nication situations may inherently be better rememberedbecause of their heightened personal relevance (Schacteret al., 2003). In accordance with the relational memory the-ory (e.g., Eichenbaum & Cohen, 2001), we predicted thathippocampal areas (specifically the cornu amonis [CA] re-

Straube et al. 307

gion) would be active during the encoding of both sen-tences and communication contexts that are later correctlyremembered. In addition, we predicted that only the en-coding of correctly remembered sentences would correlatewith activity in adjacent regions of the hippocampus (e.g.,Kensinger& Schacter, 2006; Ranganath et al., 2004;Davachiet al., 2003). We predicted that the encoding of subse-quently incorrectly remembered communication contextswould activate additional frontal (e.g., Kubota et al., 2006;Okado & Stark, 2005; Gonsalves et al., 2004) and parietalstructures (e.g., Gonsalves et al., 2004; Gonsalves & Paller,2000).

METHODS

Participants

Eighteen right-handed healthy male (Oldfield, 1971) vol-unteers, all native speakers ofGerman (mean age=24.1 years,SD= 2.8 years, range = 20 to 30 years) and without visionor auditory impairments, participated in the study. None ofthe participants had ever experienced any serious medical,neurological, or psychiatric illness. All participants gave

written informed consent and were paidA20 for participa-tion. The studywas approved by the local ethics committee.

Stimulus Construction

Spoken sentences accompanied by expressive gestureswere chosen to emphasize both the auditory and the visualaspects of a natural communication situation. The visualcomponent is of special importance because the commu-nication context was manipulated through the actorʼs bodyorientation. The actor directly faced the participant in theegocentric context, whereas the actor addressed an unseenthird person in the allocentric context. Two cameras simul-taneously filmed the actor to control for variability betweenthe two conditions (see Figure 1).We used sentences and gestures with both object-related

and person-related content. We initially created a set of 372(186 egocentric [E] and 186 allocentric [A]) short videoclips, consisting of 178 sentences with object-related con-tent accompanied by iconic gestures (89 × 2) and 194sentences with person-related content accompanied byemblematic gestures (97× 2). A comparison of emblematicand iconic gestures independent of thememory datawill be

Figure 1. Encoding and recognition procedure. (A) Scenes of the egocentric and allocentric videos that were presented during fMRI data acquisition(encoding). As seen in the example, the same items were presented in both egocentric and allocentric conditions counterbalanced acrosssubjects (S1, S2). (B) During recognition, written sentences were presented in the original (Old) and changed (New) form. (B) Examplesentences, the four questions given to participants and the possible response alternatives.

308 Journal of Cognitive Neuroscience Volume 23, Number 2

published elsewhere (Straube, Green, Jansen, Chatterjee, &Kircher, 2010) and is here not of interest. All spoken sen-tences were 5 to 10 words in length with an average dura-tion of 2.32 sec (SD= 0.36) and a similar grammatical form(subject–predicate–object). The same male actor per-formed all the speech and gestures in a natural, sponta-neous way. The procedure was supervised by two of theauthors (B.S. and A.G.) and timed digitally. All video clipswere 5 sec in length with at least 0.5 sec before and afterthe sentence onset and offset, during which the actorneither spoke nor moved.For stimulus selection, validation, and the constructionof

two similar counterbalanced stimulus sets, 17 subjects ratedeach video on a scale from 1 to 7 on “addressing” (“To whatdegree does the actor address you?”) on the sentencesʼsocial and descriptive contentand on familiarity (1 = verylow to 7 = very high). In total, 120 items were selected, halfwith iconic gestures and half with emblematic gestures.Two experimental sets were created so that the same 60items presented in the egocentric condition (E) of Set 1were presented in the allocentric condition (A) of Set 2.Conversely, the 60 items presented in the E condition ofSet 2 were presented in the A condition of Set 1 (see Fig-ure 1). This counterbalanced E andA videos across subjects.The rating parameters were analyzed with an ANOVA

for addressing, social and descriptive content, and famil-iarity between the factors set (Set 1 vs. Set 2) and com-munication context (E vs. A). We found no significantmain effects for set, F(4, 233) = 1.394, p = .237, η2 =.023, and no interactions of Set × Communication Context,F(4, 233) = 0.430, p = .787, η2 = .007, for any parameter.However, there was a significant main effect of communi-cation context, F(4, 233) = 3809.503, p < .001, η2 = .985,because of differences in ratings of “addressing,” F(1,236) = 11576.283, p < .001, η2 = .980, although not forcontent—descriptive, F(1, 236) = 0.895, p = .810, η2 =.001; social, F(1, 236) = 0.002, p = .968, η2 = .001—orfamiliarity, F(1, 236) = 2.065, p = .152, η2 = .009. There-fore, a participantʼs perception of being addressed wassignificantly enhanced in the E condition (Set 1: M = 5.13,SD = 0.32; Set 2: M = 5.15, SD = 0.29) as compared withthe A condition (Set 1: M = 1.57, SD = 0.19; Set 2: M =1.59, SD = 0.20). All other parameters were equal in allconditions. In addition, participants evaluated the materialas familiar (>5 on a scale from 1= low to 7 = high), regard-less of egocentric (Set 1: M = 5.68, SD = 0.39; Set 2: M =5.78, SD = 0.35) or allocentric condition (Set 1: M = 5.67,SD = 0.37; Set 2: M = 5.65, SD = 0.32).All other parameters (i.e., timing and movement charac-

teristics) were identical in both E and A videos, as two cam-eras simultaneously recorded the actor and videos werefurther counterbalanced across stimuli sets. For each video,time points of co-occurrence of speech and gesture (seeGreen et al., 2009; Kircher, Straube, et al., 2009; Straubeet al., 2009) were defined to model our event-related fMRIanalyses and take into account both speech and gestureinformation. For example, in the sentence “The pilot indi-

cated that everything is all right,” the point of co-occurrencewas when the end of the word (“right”) co-occurred withthe emblematic gesture (thumb-and-index-finger ring ges-ture). These timepoints of co-occurrence occurred on aver-age 2090 msec (SD = 505 msec) after the video started(1590 msec after speech onset).

Experimental Design and Procedure

During the fMRI scanning procedure, videos were pre-sented viaMR-compatible video goggles (VisuaStim©;Reso-nance Technology, Inc., Northridge, CA) and nonmagneticheadphones (audio presenting systems for stereophonicstimuli: Commander; Resonance Technology, Inc.), whichalso dampened scanner noise.

Sixty items of each of the two conditions were pre-sented in a rapid event-related design, in a pseudoran-domized order and counterbalanced across subjects.Each video was followed by a baseline condition (graybackground with a fixation cross) with a variable durationof 3750 to 6750 msec (average = 5000 msec). This rapidevent-related design was used to present the videos in arandomized sequence, but without extensive pauses,which would have reduced participantsʼ attention. Thefast presentation of stimuli may have resulted in reducedeffects of baseline contrasts for the conditions. However,the power of condition comparisons should be compa-rable with regular event-related designs because of theeffect of randomization. The latter are of special interestin this study.

Subjects were instructed to watch the videos attentively,regardless of the actorʼs orientation (egocentric or allo-centric communication context). No instructions with re-gard to memory were given to the participants. Duringthe overview scans, additional clips were presented withtheir volume adjusted to ensure that the sentences werewell understood. Each participant performed two runs with60 video clips and a total duration of 10.5 min each.

Behavioral Data Acquisition

Approximately 10 min after scanning, without prior notice,participants completed a paper and pencil test to gauge rec-ognition of the sentences seen in the experiment. All sen-tences were presented in written form. An equal number ofmodified sentences for each condition was intermixed withactual sentences presented in the experiment. The sen-tences were modified only in the word that accompaniedthe gesture. For example, the sentence “The girl asks herfriend to call” (spreading the little finger and the thumbnext to the head) was modified to “The girl asks her friendfor advice” (for another example, see Figure 1). Participantshad to report if they had seen a video in which that sentencewas spoken (indicate “old”) or not (indicate “new”) andwhether the actor was oriented toward them (E) or towardanother person (A). Participants rated their confidence

Straube et al. 309

in their responses on a scale from 1 (low confidence) to7 (high confidence; see Figure 1). Altogether, 240 sen-tences were presented in random order on a piece of paper.

fMRI Data Acquisition

MRI was performed on a 3-T Siemens scanner (SiemensMRT Trio series). Functional data were acquired withecho-planar images in 38 transversal slices (repetitiontime = 2000 msec, echo time = 30 msec, flip angle =90°, slice thickness = 3 mm, interslice gap = 0.30 mm,field of view = 220 × 199 mm, voxel resolution = 3.44 ×3.44 mm, matrix dimensions = 64 × 58 mm). Slices werepositioned to achieve whole-brain coverage. Three hun-dred fifteen volumes were acquired during each of twofunctional runs.

Data Analysis

MR images were analyzed using Statistical ParametricMapping (SPM; www.fil.ion.ucl.ac.uk) implemented inMATLAB 6.5 (Mathworks Inc., Sherborn, MA). For prac-tical reasons, minimal error-prone script-based prepro-cessing and first-level analyses were performed with SPM2.The resulting outputs are comparable with and com-patible with the newer version of SPM (SPM5), whichwas used for the second-level analyses. The first fivevolumes of every functional run were discarded to mini-mize T1 saturation effects. To correct for different ac-quisition times, we shifted the signal measured in eachslice relative to the acquisition time of the middle sliceusing a slice interpolation in time. All images of one ses-sion were realigned to the first image of a run to correctfor head movement and normalized into standard stereo-taxic anatomical Montreal Neurological Institute (MNI)space by using the transformation matrix calculated fromthe first EPI-scan of each subject and the EPI template.Afterward, the normalized data with a resliced voxel size of3.5 × 3.5 × 3.5 mm were smoothed with a 6-mm FWHMisotropic Gaussian kernel to accommodate intersubjectvariation in brain anatomy. A high-pass filter (128-sec cut-off period) was applied to remove low frequency fluctua-tions in the BOLD signal.

The anticipated hemodynamic response at the definedpoints of speech and gesture co-occurrence for eachevent-type was modeled by two response functions, a ca-nonical hemodynamic response function (Friston et al.,1998) and its temporal derivative. The temporal derivativewas included in the model to account for the residual var-iance resulting from small temporal differences in the on-set of the hemodynamic response, which is not explainedby the canonical hemodynamic response function alone.The functions were convolved with the event sequenceon the basis of the individual subsequent memory perfor-mance of the subjects (memories of correct communica-

tion context: CC; incorrect communication context: IC;and not remembered items: M) in a general linear model.In a post hoc analysis (see below), the factor communica-tion context (Egocentric: E and Allocentric: A) was also in-cluded resulting in the following six conditions: E-CC, E-IC,E-M, A-CC, A-IC, and A-M. A fixed event duration of onesecond was chosen to get a broader range of data aroundthe speech–gesture segments and has also been appliedsuccessfully in past studies of co-verbal gesture process-ing (Straube et al., 2009, 2010). On the basis of the expe-rience of prior investigations (Kircher, Blumel, et al., 2009;Kircher, Straube, et al., 2009; Straube et al., 2009), model-ing the one second window (time logged to the keywordof each sentence) gives a better signal-to-noise ratio thanmodeling the whole video duration of 5 sec within ourstimulus material.A group analysis was performed by entering contrast

images into a flexible factorial analysis as implementedin SPM5, in which subjects are treated as random vari-ables. Voxels with a significance level of p < .001 uncor-rected belonging to clusters with at least 10 voxels arereported. A Monte Carlo simulation of the brain volumeof the current study was conducted to establish an appro-priate voxel contiguity threshold (Slotnick, Moo, Segal, &Hart, 2003). Assuming an individual voxel type I error ofp < .001, a cluster extent of 10 contiguous resampledvoxels was indicated as sufficient to correct for multiplevoxel comparisons at p < .05. Activation peaks of most ofthe activation clusters also hold a false discovery rate(FDR). Corresponding corrected p values for each activa-tion peak were included in Tables 1 and 2 and the Resultssection.The reported voxel coordinates of activation peaks are

located in MNI space. For anatomical localization, thefunctional data were referenced to probabilistic cytoar-chitectonic maps (Eickhoff et al., 2005, 2007). For thecomparison and presentation of distinct hippocampalactivation within the CA subregion of the hippocampus(which is thought to be involved in memory binding pro-cesses and episodic memory; e.g., Eichenbaum, 2004), asmall volume correction was applied to the contrast CCversus IC (see Results section) using an ROI, defined bythe CA probability maps from Amunts et al. (2005). Forthe small volume correction, a more liberal whole-brainthreshold was applied ( p < .005 uncorrected), and theVOI was then restricted to the CA region of the hippo-campus. The corresponding statistics of this procedureare reported in the Results section. All contrast estimatesprovided in Figures 3–5 present activation change withinthe whole activation clusters of the corresponding con-trast. For this purpose, we calculated the average beta valuefor each subject and condition for the correspondingactivation cluster of the group analysis. Thus, contrast barspresent the average beta value of the subjectsʼ cluster aver-ages. Error bars represent as the SEM (between-subject var-iability). Parameter estimates of Figure 6 present activationchange within the entire subregions (CA and entorhinal

310 Journal of Cognitive Neuroscience Volume 23, Number 2

cortex [EC]) of the hippocampus created with the sameaveraging procedure.Statistical analyses of data other than fMRI were per-

formed using SPSS version 14.0 for Windows (SPSSInc., Chicago, IL). t Tests and ANOVAs were used forthe analyses. Greenhouse–Geisser correction was ap-plied whenever necessary. Discrimination performance(d0) and response criterion (c) were calculated followingthe signal detection theory (d0 = z(hits) − z(FA); c =−1/2 [z(hits) + z(FA)]; e.g., Macmillan & Creelman,1991). Statistical analyses are two-tailed with α levels ofsignificance of p < .05.

Contrasts of Interest

The fMRI analyses are structured in the following man-ner: First, comparisons between subsequently remem-bered sentences in the correct context (CC) and in theincorrect context (IC) in contrast to nonrememberedsentences (misses: M) were performed to give an over-view of the structures involved in memory processes withand without correct contextual information.

Second, subsequently remembered sentences in the cor-rect context (CC) were contrasted with subsequently re-membered sentences in the incorrect context (IC) to testthe hypothesis that predominantly medial frontal brain re-gions are related to inaccurate context memory and hip-pocampal regions to correct context memory. Third, apost hoc analysis was performed to explore if the analysesreported above were influenced by the communicationcontext (ego- or allocentric). Finally, an interaction analysisbetween communication context (ego- or allocentric) andmemory quality (correct/incorrect context) was performedto explore brain regions related to egocentric encoding(sentences subsequently remembered as egocentric).

Thus, there was no separation of egocentric and allo-centric items in the first two analyses to obtain more eventsand more reliable results for CC, IC, and M (given that CCare predominantly egocentric, IC are mainly allocentric,and M contains equal proportions of egocentric and allo-centric stimuli; see Behavioral results). Because 10 of18 subjects in the egocentric incorrect-context condition(E-IC) and 4 of 18 subjects in the allocentric correct-contextcondition (A-CC) had fewer than three trials per condition,

Table 1. Activation for Correct (CC) and Incorrect Context Information (IC) versus Misses (M)

Contrast Anatomical Region Lat. BA

Coordinates

t *FDR No. Voxelsx y z

CC > M Supramarginal gyrus R 40 56 −28 46 5.38 0.018 23

Postcentral/supramarginalgyrus

R 2 35 −39 49 5.35 0.018 101

Inferior/middle temporalgyrus

L 37 −53 −56 −4 5.06 0.030 81

Postcentral gyrus L 2 −39 −42 60 4.89 0.033 29

Inferior frontal gyrus L 45 −49 28 7 4.71 0.033 31

Parahippocampal gyrus (CA) L −28 −28 −14 4.63 0.033 21

Supramarginal gyrus L 40 −46 −39 28 4.45 0.041 14

Inferior frontal gyrus L 44/45 −53 11 7 4.35 0.041 26

Calcarine gyrus R 17 14 −98 7 4.30 0.041 10

IC > M Premotor cortex L 6 −4 −25 56 5.19 0.043 18

Postcentral gyrus L 2/3 −35 −42 60 5.08 0.043 34

Premotor cortex R 6 11 −7 56 4.90 0.043 33

Parahippocampal gyrus (CA) L −28 −28 −14 4.68 0.046 10

Parahippocampal gyrus (EC) R 25 −4 −25 4.65 0.046 17

Postcentral gyrus R 2/3 32 −35 42 4.65 0.046 43

Fusiform/hippocampus (SUB/CA) L −32 −18 −25 4.49 0.051 10

Middle temporal gyrus R 37 56 −46 11 3.63 0.087 10

Significance level (t value), size of the respective activation cluster (No. voxels ⩾ 10) at p < .001 uncorrected ( p < .05 corrected, see Methods), andpossible false discovery rate correction (*FDR). Coordinates are listed in MNI space. BA is the Brodmannʼs area nearest to the coordinate and shouldbe considered approximate. Contrasts were named as follows: CC = correct context; IC = incorrect context; M = misses.

Straube et al. 311

the post hoc analysis was performed in an unbalanceddesign (E-CC: n = 18, E-IC: n = 8; E-M: n = 18; A-CC:n = 14; A-IC = 18; A-M: n = 18). For the remaining sam-ple the average number of trials were for E-CC = 29,E-IC = 12, E-M = 26, A-CC = 19, A-IC = 22 and A-M =27. Contrast estimates were provided for the analyses thatcontrasted CC, IC, and M to explore the effect of the con-founded variable (stimulus material; see Figures 3D and 4Band D).

RESULTS

Behavioral Results

Memory of Sentences

The percentage of study items correctly reported as old(hits) was 56.49% (SD = 19.54%) for the egocentric (E)and 54.82% (SD = 17.51%) for the allocentric condition(A). The corresponding false alarm rate (FA; the percent-age of new modified sentences that were judged as old)was 12.04% (SD= 13.51%) for the E condition and 11.58%(SD= 12.56%) for the A condition. A 2 × 2 within-subjectsANOVA, with communication context (E vs. A) and mem-ory performance (hits vs. FA), revealed a significant maineffect of memory performance (hits > FA; F(1, 17) =91.130, p < .001, η2 = .843), no significant effect of thecommunication context, F(1, 17) = 1.405, p = .252, η2 =.076, and no significant interaction between factors, F(1,17) = 0.508, p = .486, η2 = .094.

To assess differences in signal detection parameters acrossconditions, we calculated the discrimination index d0. Weobtained an average discrimination performance of d0 =.586 (SD = 0.274) for the E condition and d0 = .570 (SD =0.240) for the A condition. The corresponding responsecriteria (c) were 0.209 (SD = 0.170) for the E conditionand 0.223 (SD = 0.159) for the A condition. Both memoryparameters did not differ between the E and the A condi-tions:d0, t(17)=0.718,p=.483; c, t(17)=−1.111,p=.282.

Memory of Communication Context

The percentage of study items correctly reported as pre-viously seen in an egocentric context (E hits) was 49.81%(SD= 17.28%) of all egocentric encoded items (thus 88.5%of correct remembered items of the egocentric condition;SD = 30.7%) with a corresponding failure rate (E failure)of 6.47% (SD = 8.82%) of all egocentric encoded items(thus 11.5% of correct remembered items of the egocen-tric condition; SD = 15.7%). In contrast, the proportionof study items correctly reported as previously seen inan allocentric context (A hits) was 15.11% (SD = 9.93%)of all allocentric encoded items (thus 27.91% of correct re-membered items of the allocentric condition; SD = 18.3%)with a corresponding failure rate (A failure) of 39.03% (SD=15.42%) of all allocentric encoded items (thus 72.09% ofcorrect remembered items of the allocentric condition;SD = 28.48%). A 2 × 2 within-subject ANOVA betweencommunication context (E vs. A) and memory performance

Table 2. Interaction between Communication Context and Memory (E-CC > E-IC) > (A-CC > A-IC)

Anatomical Region Lat. BA

Coordinates

t *FDR No. Voxelsx y z

Superior medialfrontal gyrus

R 6/8 7 25 60 5.63 0.004 59

Inferior parietal lobule L 40/7 −49 −60 42 5.39 0.004 113

Basal ganglia (putamen) R 21 4 −11 4.84 0.007 69

Middle temporal gyrus L 37 −56 −42 0 4.64 0.012 16

Middle frontal gyrus L 44/9 −35 11 46 4.63 0.012 48

Middle frontal/orbitalgyrus

L 46 −39 46 14 4.42 0.017 30

Middle cingulate cortex R 23/31 7 −25 35 4.14 0.021 10

Basal ganglia (putamen) L −28 7 −4 4.14 0.021 28

Superior frontal gyrus R 46/10 18 46 35 4.13 0.021 17

Cerebellum R 25 −81 −28 4.10 0.021 12

Inferior frontal gyrus L 47/10 −35 42 −11 3.87 0.029 12

Basal ganglia (putamen) L −21 0 11 3.86 0.030 10

Significance level (t value), size of the respective activation cluster (No. voxels ⩾ 10) at p < .001 uncorrected ( p < .05 corrected, see Methods), andpossible false discovery rate correction (*FDR). Coordinates are listed in MNI space. BA is the Brodmannʼs area nearest to the coordinate and shouldbe considered approximate.

312 Journal of Cognitive Neuroscience Volume 23, Number 2

(hits vs. failures) indicated a significant main effect ofmemory performance, hits > failures, F(1, 17) = 37.991,p < .001, η2 = .691, no significant effect of the commu-nication context, F(1, 17) = 2.166, p = .159, η2 = .113,and a significant interaction between factors, F(1, 17) =66.198, p < .001, η2 = .796 (see Figure 2A).To assess differences in signal detection parameters

across conditions, we further defined the discriminationindex d0 for each orientation. We obtained an average dis-crimination performance of d0 = .577 (SD = 0.252) forthe E condition and d0 = −.315 (SD = 0.238) for the Acondition. The corresponding response criteria (c) were.291 (SD = 0.126) for the E condition and .300 (SD =

0.123) for the A condition. The ability to recall the com-munication context during the recognition task was sig-nificantly enhanced for the E condition in contrast to theA condition, d0, t(17) = 8.295, p < .001.

Our data reveal an interaction between egocentric con-text and subsequent recollection of communication con-text. Egocentric videos were predominantly correctlyremembered as egocentric (8 of 18 participants did notmake a single error). However, allocentric videos tendedto be falsely remembered as egocentric (3 of 18 partici-pants did not make a single correct response). This effectis independent of the participantsʼ memory of the sen-tences, which did not differ between conditions.

Figure 2. Memoryperformance and confidenceratings for item and contextmemory. (A) The hit (left)and false alarms rate (right)for items (white) andcommunication contexts (gray)of ego- and allocentric videoclips. (B) Correspondingdiscrimination performance(d0; left) and response criteria(c; right) for items (white)and context (gray) of ego-and allocentric video clips.(C) Results of the confidenceratings for correct (left) andincorrect responses (right).For conditions in which notall subjects providedresponses, the number ofremaining subjects is notedin the bar graph. The errorbars indicate the SEM.

Straube et al. 313

Analyses of false alarms (responses to new items dur-ing recognition) with regard to communication contextwere performed to investigate possible response biasesunaffected by the actual encoding episode (these sen-tence were not actually encountered). We found moreegocentric than allocentric responses for new sentencespresented during the recognition task (E-CCnew: M =9.44%, SD = 9.24%; E-ICnew: M = 2.26%, SD = 4.49%;E-CCnew > E-ICnew, t(17) = 4.867, p < .001; and A-CCnew: M = 3.24%, SD = 5.84%; A-ICnew: M = 7.72%,SD = 7.63%; A-ICnew > A-CCnew, t(17) = 3.767, p <.005). However, this effect is still smaller than the amountof egocentric responses in the egocentric condition as wellas allocentric condition (E-CCold: M = 49.81%, SD =17.34%; E-ICold: M = 6.48%, SD = 8.72%; [E-CCnew − E-ICnew] < [E-CCold > E-ICold], t(17) = 7.591, p< .001; A-CCnew: M = 15.09%, SD = 9.99%; A-ICnew: M = 38.98%,SD=15.37%; [A-ICnew−A-CCnew]<[A-ICold−A-CCold],t(17) = 4.869, p< .001). This is also reflected in the signif-icant interaction effect in a three-factorial ANOVA, with thefactors stimulus material (Ego-/Allocentric), responses toold and new items (Old/New), and memory about thecontext (CC/IC); main effect: Ego-/Allocentric, F(1, 17) =2.582, p = .127, η2 = .132; main effect: Old/New, F(1,17) = 88.059, p < .001, η2 = .838; main effect: CC/IC,F(1, 17) = 46.313, p < .001, η2 = .731; interaction: Ego-/Allocentric × Old/New, F(1, 17) = .858, p = .367, η2 =.048; interaction: Ego-/Allocentric × CC/IC, F(1, 17) =74.518, p < .001, η2 = .814; interaction: Old/New × CC/IC, F(1, 17) = 21.236, p < .001, η2 = .555; interaction:Ego-/Allocentric × Old/New × CC/IC, F(1, 17) = 48.287,p < .001, η2 = .740. Correspondingly, the “correct minusfalse” (old − new) proportion for egocentric responses(M= 0.4028, SD= 0.1968) is higher than for allocentric re-sponses (M = 0.1178, SD = 0.0897), t(17) = 6.479, p <.001, suggesting that egocentric encoding leads to bettermemory.

Confidence Rating

Participantsʼ confidence ratings are presented in Figure 2C.For memories of sentences (white bars), we found higherconfidence ratings for hits compared with false alarms, F(1,17) = 37.539, p < .001, η2 = .688, independent of com-munication context; egocentric vs. allocentric: main effect,F(1, 17) = 0.010, p = .922, η2 = .001; interaction, F(1,17) = 0.007, p = .934, η2 < .001. We found that eightparticipants made no mistakes in the egocentric conditionand three participants had no correct responses in theallocentric condition. Furthermore, the descriptive data in-dicate an interaction effect between confidence and con-dition, as the participants gave the highest confidenceratings for both correctly recognized E conditions and in-correctly recognized A conditions (see Figure 2). A compar-ison after participants with missing values were excludedconfirms the interaction effect, F(1, 7) = 0.007, p < .001,η2 < .863, demonstrating a trend for a main effect of

communication context, E > A, F(1, 7) = 4.804, p =.064, η2 < .407, and revealing no significant main effectof correct versus false responses, F(1, 7) = 0.370, p =.562, η2 < .050.There did not appear to be any differences in confi-

dence ratings about item memory between egocentric andallocentric conditions. However, participants appeared tobe more confident in their responses when they falsely be-lieved that they were presented with an egocentric video ascompared with when they falsely believed that they saw anallocentric video.

fMRI Results

Neural Encoding Processes Predicting Correct (CC) andIncorrect (IC) Recognition of the CommunicationContext versus Misses (M)

To identify the brain regions that aremore active during theencoding of sentences that are subsequently remembered(in either the correct context [CC] or incorrect context [IC])in contrast to nonremembered sentences (misses [M]), weincluded the corresponding contrast images of the indi-vidual first-level analyses in the flexible factorial analysesof SPM5. Further separation of egocentric and allocentricitems was not performed to obtain more events and morereliable results for CC, IC, and M. However, contrast esti-mates resulting from the post hoc analyses presented inFigure 3D illustrate the activations of the hippocampal re-gions for CC, IC, and M separated into egocentric (E-CC,E-IC, E-M) and allocentric conditions (A-CC, A-IC, A-M).For CC > M, the data revealed activity in the bilateral

parietal and left lateral inferior frontal, posterior temporal,and hippocampal brain regions (see Table 1 and Figure 3;red). For IC > M, we observed bilateral parietal and pre-central activation as well as activity in the bilateral medial-temporal lobe, predominantly in the EC and subiculum(see Table 1 and Figure 3; green). To identify the com-mon activity between CC and IC conditions, we used aconjunction analysis and demonstrated bilateral superiorparietal/postcentral activation for both conditions ([CC >M] ∩ [IC > M]; MNIx, y, z: 32,−35, 42, t= 4.65, 28 voxels;MNIx, y, z: −39, −42, 60, t = 4.89, 14 voxels; see Figure 3;yellow).Probabilistic cytoarchitectonic maps (Eickhoff et al.,

2007) were applied to determine the specific anatomicallocation of the medial-temporal activation. For CC > M,the activation peak was most likely located in the CA re-gion of the left hippocampus (MNIx, y, z: −28, −28, −14;assigned to CA: probability = 70% [40–90%]; FD: prob-ability = 70% [0–80]). However, for IC > M, additionalregions of the bilateral EC and subiculum (SUB) were ac-tivated (right: MNIx, y, z:−25,−4,−25, assigned to amyg-dala: probability = 60% [10–100%]; EC: probability =60% [10–100]; left: MNIx, y, z: −32, −18, −25, assignedto SUB: probability = 50% [30–70%]; CA: probability =40% [20–80]; see Figure 3).

314 Journal of Cognitive Neuroscience Volume 23, Number 2

These data provide insight into the common and dis-tinct neural encoding mechanisms that yield both accurateand inaccurate memories of the context in which a sen-tence was observed. Parietal and medial-temporal brainareas appear to be more active during the encoding oflater correctly remembered sentences (hits) than duringthe encoding of nonremembered sentences (misses) in-dependent of the communication context. The parameterestimates of the post hoc analyses (Figure 3D) indicate

that activation of the CA region in the CC condition is pre-dominantly due to the activation increase of the E-CC andnot the A-CC condition. By contrast, activationof theEC andsubiculum in the IC condition appears to result from theactivation of both the E-IC and the A-IC conditions. Thissuggests that memory processes related to the binding ofsentence and context within the CA region of the hip-pocampus are less involved in the encoding of allocentricconditions.

Figure 3. Activation for sentence memory with correct (CC) and incorrect memory for the communication context (IC) versus misses (M).(A) The cortical activations for sentence memory with correct (CC, red) and incorrect recollection about the communication context (IC, green)in contrast to sentences that were not remembered (misses, M). (B) The medial-temporal activation pattern (from the same analyses aspanel A) in sagittal (right) slices of the probability maps (Eickhoff et al., 2005, 2007; see Results section). Contrast estimates in panel C (middle)illustrate the differences for the clusters of medial-temporal activations for CC, IC, and M. Contrast estimates in panel D (right) illustrate thedifferences for the same regions in the post hoc analysis including the factor actor orientation (Ego- vs. Allocentric). CC: correct context;IC = incorrect context; M = misses; E-CC = egocentric correct context; E-IC = egocentric incorrect context; E-M = egocentric misses;A-CC = allocentric correct context; A-IC = allocentric incorrect context; A-M = allocentric misses; CA = cornu amonis; EC = entorhinal cortex;SUB = subiculum cortex.

Straube et al. 315

Comparison of Neural Encoding Processes PredictingSubsequent Correct and Incorrect Recognition of theCommunication Context (CC vs. IC)

We identified brain regions related to the encoding of accu-rate and inaccurate memories of the communication con-text through direct comparisons of CC and IC. As the CAsubregion of the hippocampus is thought to have a specificrole in episodic and associative memory processes (e.g.,Eichenbaum, 2004), a small volume correction (SVC) forthis region was also applied (see Methods section) to bothcontrasts (CC > IC and IC > CC).

Whole-brain analysis revealed an activation of the rightcalcarine gyrus for CC > IC (Brodmannʼs area [BA] 17;MNIx, y, z: 14, −98, 7, t = 7.14, 211 voxels, FDR < 0.001).The analysis using SVC revealed a significant activation ofthe right CA region of the hippocampus (MNIx, y, z: 32,−32, −7, t = 3.98, 8 voxels, FDR = 0.034; see Figure 4A).

For IC > CC, whole-brain analysis revealed activationof the PACC (BA24, 25, 32; MNIx, y, z: 0, 32, −4, t =4.70, 15 voxels, FDR = 0.173; see Figure 4A), the superiorfrontal cortex (BA9/6; MNIx, y, z: −21, −7, 67, t = 4.86,16 voxels, FDR = 0.173), and the occipital lobe (BA18;MNIx, y, z: −11, −91, −7, t = 4.40, 19 voxels, FDR =0.207). The SVC analysis revealed no additional significantactivation of the CA region of the hippocampus.

Our results suggest that there are particular encodingprocesses that lead to accurate and inaccurate memoriesof the social context of a communication situation. Activ-ity in the right CA region appears to be related to contextmemory, whereas the PACC and superior frontal cortexmay be involved in the encoding of inaccurate memories.

The bar graphs of the post hoc analyses (Figure 4B) indi-cate that activation of the CA region in the CC condition ispredominantly due to increased activation of the E-CC andnot the A-CC condition. This suggests that memory pro-cesses related to the binding of sentence and context withinthe CA region of the hippocampus are less involved in theencoding of allocentric conditions. By contrast, increasedactivation within the PACC for IC > CC is predominantlydue to differences in the allocentric condition (A-IC > A-CC) and not the egocentric condition (E-IC = E-CC). Thissuggests that increased activation in the PACC is not gener-ally related to incorrect context memory but is unique tothe encoding of incorrect context for allocentric situations.Contrast estimates for the superior frontal and occipitalactivations indicate that activity in these regions is predomi-nantly related to the stimulus condition (Ego-/Allocentric)and less to the subsequent memory performance.

Interaction between Communication Context andMemory (E-CC > E-IC) > (A-CC > A-IC)

Previous analyses focused on the neural substrates of cor-rect and incorrect context memory of social interactions.However, the post hoc analyses that differentiated betweenegocentric and allocentric encoding episodes have indi-

cated that the encoding context has an effect on thesememory processes. We calculated an interaction betweenencoding context and context memory (E-CC > E-IC) >(A-CC > A-IC) to provide insight into the influence of en-coding context on these memory processes. We found in-creased activation in bilateral and medial frontal and lefttemporal and parietal brain regions as well as the bilateralBG in this analysis (see Figure 4 and Table 2 for coordinatesand statistics). Bar graphs in Figure 5 illustrate that theseregions are more activated for sentences subsequently re-membered as egocentric (E-CC and A-IC) than sentencesremembered as allocentric (E-IC andA-CC). Thus, encodingprocesses that predict subsequent egocentric memories(regardless of if the episode was actually encountered inan egocentric condition) are related to bilateral and medialfrontal and left temporal andparietal brain regions aswell asthe bilateral BG.We extracted the average brain activation for each partic-

ipant within ROIs of the left and right CA and the EC to ex-plore the influence of encoding context on activation withinmedial-temporal structures. As illustrated by Figure 6,we found significant modulation of memory-related activa-tion (CC > M) within all ROIs for the egocentric condition(E-CC > E-M), all ROIs together, t(17) = 2.686, p< .05, butnot for the allocentric condition (A-CC > A-M), all ROIstogether, t(13) = −.080, p = .937. Directed comparisonssupport the finding that the effect observed in the ego-centric condition is present in all ROIs: left CA, t(17) =3.249, p < .005; right CA, t(17) = 2.145, p < .05; left EC,t(17) = 2.552, p< .05; right EC, t(17) = 1.766, p< .05 (seeFigure 6). The differences between subregions of the hip-pocampus did not reach the significant threshold.These results indicate that hippocampal activations are

modulated in relation to memory performance only inthe egocentric condition. This suggests that processesthat bind communication context and spoken sentenceoccur only for sentences remembered as egocentric thatwere actually encountered in an egocentric context.

DISCUSSION

Memories of interpersonal interactions comprise morethan just the words spoken; they include additional im-portant information about the context in which the com-munication occurs. Our study aimed to determine theneural encoding processes underlying successful recogni-tion of earlier communications, with and without recol-lection of the context in which they were observed.Our results indicate that memories of social interactions

are egocentrically biased. Participants tended to reportinaccurately that they had been directly addressed whenthe original event was presented in an allocentric context.At the neural level, we observed that activation of specificcortical and hippocampal brain regions during encodingpredicted subsequent accurate and inaccurate recollectionof contextual information. Activity in the right hippocam-pus was associated with successful encoding of speech,

316 Journal of Cognitive Neuroscience Volume 23, Number 2

gesture, and their social context. In contrast, the anteriorcingulate, superior frontal, and bilateral EC/subiculumwas associated with in the inaccurate encoding of the socialcontext of the original communications. These resultsagree with previous research on the memory processes in-volved in speech and gesture binding (Straube et al., 2009)and memory theories that suggest the hippocampal CAregion plays a role in episodic or relational memory pro-cesses (e.g., Kensinger & Schacter, 2006; Eichenbaum,

2004; Ranganath et al., 2004; Davachi et al., 2003). Further-more, our data suggest the frontal lobe is involved in pro-cesses that result in inaccuratememories, which agrees withpast research on memory encoding (e.g., Kubota et al.,2006; Okado & Stark, 2005; Gonsalves et al., 2004) and rec-ognition (Slotnick & Schacter, 2004; Cabeza et al., 2001;Swick&Knight, 1999; Schacter et al., 1997; Schacter, Curran,et al., 1996; Schacter, Reiman, et al., 1996; for a review, seeSchacter & Slotnick, 2004).

Figure 4. Neural correlates of correct versus incorrect memory for the encoding context. (A) Results from the ROI analyses of the CA region ofthe hippocampus. The right hippocampus is activated for the direct comparison of correct versus incorrect memory for the communicationcontext. Contrast estimates of the post hoc analyses (B; right) illustrate that this difference is mainly due to the difference in the egocentric condition.(C) Result of the reverse contrast (IC > CC). Contrast estimates on the right of panels A and C illustrate the differences between conditionsfrom the original analyses. Contrast estimates in panels B and D illustrate the differences between conditions within the same regions from thepost hoc analysis, including the factor actor orientation (Ego- vs. Allocentric). CC = correct context; IC = incorrect context; M = misses;E-CC = egocentric correct context; E-IC = egocentric incorrect context; E-M = egocentric misses; A-CC = allocentric correct context; A-IC =allocentric incorrect context; A-M = allocentric misses; CA = cornu amonis; OL = occipital lobe; SFG = superior frontal gyrus; L = left; R = right.

Straube et al. 317

We found that the left inferior frontal gyrus and the pos-terior temporal lobe were more active during the encodingof items whose contextsʼ were later correctly remembered(CC) than they were for misses (M). Past studies haveshown that inferior frontal and posterior temporal brain re-gions are involved in speech–gesture integration (Kircher,Straube, et al., 2009) and are also related to memory bind-ing processes (Straube et al., 2009). These memory andintegration processes are most likely engaged during the

encoding of memories with accurate contextual infor-mation as opposed to the encoding of memories with in-accurate contextual information. Straube et al. (2009)demonstrated that the individual subsequent memory per-formance for related abstract speech and gesture informa-tion is correlated with activity in the left inferior frontal andthe posterior middle temporal gyrus during encoding.Although we presented written sentences without gesturesduring the recognition phase of this experiment, it is most

Figure 5. Interaction of communication context and context memory. Results from the post hoc analyses, showing activations for the interactionbetween encoding context (Ego- vs. Allocentric) and context memory [(E-CC > E-IC) > (A-CC > AIC)]. Results are presented on a single subjectʼsrendered brain (top and middle) and an axial slice of the anatomical toolbox (bottom). Contrast estimates represent the averaged beta value(arbitrary units [a.u.]) across all voxels of each activation cluster. Error bars represent the SEM. Contrast estimates show that all presented regionsare more activated for trials subsequently remembered as egocentric (E-CC and A-IC) in contrast to those remembered as allocentric (E-IC andA-CC). CC = correct context; IC = incorrect context; M = misses; E-CC = egocentric correct context; E-IC = egocentric incorrect context; E-M =egocentric misses; A-CC = allocentric correct context; A-IC = allocentric incorrect context; A-M = allocentric misses; IPL = inferior parietal lobe;MTG = middle temporal gyrus; MFG = middle frontal gyrus; SFG = superior frontal gyrus; L = left; R = right.

318 Journal of Cognitive Neuroscience Volume 23, Number 2

likely that gesture information is better incorporated in theneural representation of the communication episode whencontextual information (such as the speakersʼ body orienta-tion) is also remembered. Therefore, the left lateralizedfronto-temporal activation pattern for CC versus M possiblyindicates that memory-binding processes of speech andgesture information are occurring.In both CC and IC conditions (in contrast to M condi-

tions), we observed bilateral postcentral/parietal brain acti-vation during encoding. This observed activitymight bedueto heightened gesture processing in the dorsal visuospatialstream, eventually yielding more accurate memories (e.g.,Cunnington, Windischberger, Robinson, & Moser, 2006;Weiss et al., 2006;Mühlau et al., 2005;Wheaton, Thompson,Syngeniotis, Abbott, & Puce, 2004). However, there is alsoevidence that the parietal lobe plays a role in episodicmem-ory retrieval (e.g., Cabeza, Ciaramelli, Olson, & Moscovitch,2008; Vilberg & Rugg, 2008; Naghavi & Nyberg, 2005;Wagner et al., 2005; Cabeza & Nyberg, 2000), working

memory, and verbal storage processes (e.g., Herwig et al.,2003). Kahn, Davachi, and Wagner (2004) and Wheelerand Buckner (2003) observed activation in the left inferiorparietal cortex during false alarms. In the study by Kahnet al., participants were presented with words while in ascanner and then were asked to identify the words theyhad seen in a recognition test. During the recognitionphase, activity was observed in the left inferior parietal cor-tex during false alarms, which suggests that activation ofthis region is involved in perceived recognition. In addi-tion, parietal regions appeared to interact with the hippo-campal formation during successful recognition (Vincentet al., 2006). Past observations of parietal/postcentral activ-ity during correct and incorrect recognition are consistentwith our current finding that the parietal lobe is involved inencoding processes that lead to accurate and inaccuratememories of context information. Processes of imaginingthat occur during encoding, which might be related toparietal activations, are perhaps the basis of the formation

Figure 6. Medial-temporallobe activation. Results fromthe ROI analyses of the CAand the EC of the left and righthippocampus. The sagittal (left)and coronal slices of a templatebrain illustrate the locationof the subregions of thehippocampus (EC: black; CA:white). Contrast estimatesrepresent the averaged betavalue (arbitrary units [a.u.])across voxels of each ROI.Error bars represent the SEM.Contrast estimates illustrate amodulation of activation withregard to memory qualitywithin the CA region for theegocentric condition (CC >IC > M). The EC seemsto be involved if item, butnot context, is correctlyremembered (E-IC and A-IC).For the allocentric context,the A-IC condition led tohighest activation within CAand EC. L = left; R = right.

Straube et al. 319

of false memories (Gonsalves et al., 2004; Gonsalves &Paller, 2000).

We found that PACC and superior frontal cortex weremore active during encodingwhen communication contextwas later incorrectly remembered. This result is consistentwith the view that the frontal cortex plays a role in false rec-ognition (Slotnick & Schacter, 2004; Cabeza et al., 2001;Schacter et al., 1997; Schacter, Reiman, et al., 1996) andagrees with evidence that the ACC is involved in false recog-nition (Slotnick & Schacter, 2004). Interestingly, Kim andCabeza (2007) found that the ACC (among other regions)is active during trials that participants are confident theyhave previously seen, regardless if they have actually en-countered them before. The confidence ratings obtainedin our study indicate that participantsʼ confidence in theirmemories of the communication context did not differ be-tween true and false recognition. However, we observedactivation of the ACC only during the encoding of trials inwhich the context was inaccurately remembered. Our posthoc analysis indicate that this effect is mainly due to differ-ences in the allocentric condition (A-IC>A-CC), suggestingthat the PACC has a specific role in trials subsequentlyfalsely remembered as egocentric (self-referring). By con-trast, activation of the occipital and superior frontal cortexseems to be especially affected by the stimulus condition(ego-/allocentric), and results of the collapsed conditionsmay predominantly represent differences in the amountof egocentric and allocentric trials between CC, IC, and M(see Figure 4C and D).

Although most past research on the neural basis of trueand false memories investigated retrieval processes, welooked at neural activity during encoding and comparedthese data to subsequent recognition performance. Partici-pants remembered all sentences equally well, regardless ofthe communication context. However, memory of just thecommunication context was not only worse in allocentricconditions, participants tended to recall allocentric stimulias having been encountered egocentrically. One interpre-tation of this finding is that the contextual information inthe allocentric condition was less personally relevant forparticipants and therefore was less accurately remembered(e.g., Symons & Johnson, 1997). However, this interpreta-tion accounts for poorer performance in the allocentriccondition but does not explain whymemories of social con-text are biased toward the egocentric condition. Further-more, the fact that participants remembered sentences inboth conditions equally well contradicts the theory that in-dividuals were less attentive in allocentric situations. An al-ternative explanation is that participantsmay have imaginedthemselves as the unseen addressed person to get a betterunderstanding of the communication and performed amental rotation of the observed event during encoding.Processes responsible for such shifts in mental perspectiveare often referred to as “theory of mind” or “mentalizing”mechanisms (e.g., Frith & Frith, 1999). The occurrence ofsuch processes during encoding may change the partici-pantʼs perception of the situation, leading to correct item

memory (comparable with the egocentric condition), buta false memory about the encoding context of allocentriccommunications. According to this theory, the observedactivation of the ACC could also reflect shifts in imaginedperspective, which have been reported in other studies thatinvestigated mentalizing or theory of mind processes (e.g.,Kircher, Blumel, et al., 2009; Krach et al., 2008, 2009; Kedia,Berthoz,Wessa, Hilton, &Martinot, 2008; Cheng et al., 2007;Schulte-Ruther, Markowitsch, Fink, & Piefke, 2007; Amodio& Frith, 2006; Baird et al., 2006; David et al., 2006; Lawrenceet al., 2006; Vogeley et al., 2004) and other social processingmechanisms (for a review, see Lieberman, 2007). With re-gard to memory processes, an egocentric bias reflects thestrong role played by the self in the encoding and retrievalof episodic memories (for a review, see Schacter et al.,2003). It has been shown that when information is en-coded in relation to ourselves, it is usually better remem-bered than other types of semantic information (Rogers et al.,1977; for a review, see Symons & Johnson, 1997). Kelleyet al. (2002) have used fMRI to reveal brain regions thatplay a role in the “self-reference effect.” Participants werescanned while they either decided whether a series of traitadjectives (e.g., honest, friendly) described themselves ordecided whether they described a familiar other person (inthis case, George Bush). The first condition has been usedextensively to engage processing in relation to the self,whereas the latter condition is assumed to involve semantic,but not self-referential, encoding. Semantic but non-self-referential encodingwas associatedwith activation in the leftinferior PFC, whereas self-referential encoding was asso-ciated with activation in a inferior portion of the medialPFC (including inferior parts of the ACC). Consistent withour data, these findings suggest a specific role of the inferiormedial frontal/ACC in self-referential encoding of informa-tion. Therefore, an egocentric bias and related self-referentialencoding processes may be the basis of false context mem-ory in the allocentric condition. To test this assumption inour fMRI data set, we performed an interaction analysisbetween communication context and context memory (E-CC > E-IC) > (A-CC > A-IC). In this analysis, we found in-creased activation in bilateral and medial frontal and lefttemporal and parietal brain regions as well as the bilateralBG. These regions are more activated for sentences subse-quently remembered as egocentric than sentences remem-bered as allocentric. Thus, encoding processes that predictsubsequent egocentricmemories (regardless of if the episodewas actually encountered in an egocentric condition) are re-lated to increased activation in a distributed network, includ-ing regions previously found to be involved self-referentialencoding (e.g., Macrae et al., 2004; Kelley et al., 2002), menta-lizing (e.g., Schulte-Ruther et al., 2007), and perspective taking(e.g., David et al., 2006; Vogeley et al., 2004; for ameta-analysisabout self-referential processes, see Northoff et al., 2006).However, some of these regions are also thought to beinvolved in false memory encoding (e.g., Gonsalves et al.,2004; PACC) and false recognition (e.g., Kim & Cabeza, 2007;left parietal and superior frontal cortex). Specifically, the

320 Journal of Cognitive Neuroscience Volume 23, Number 2

medial cortical regions appear to be involved in self-referential processing in the brain (Northoff et al., 2006;Northoff & Bermpohl, 2004; Kircher et al., 2000). Althoughthe observation that the frontal lobe is related to correct andfalse memory processes is a consistent finding, there is a highheterogeneity in the specific location of frontal activationbetween the studies, which suggests functional differen-tiation within this brain region. In our study, we found thatthe PACC was specifically activated for incorrectly egocen-tric remembered sentences (A-IC > A-CC), whereas dorso-medial frontal and superior frontal regions appear to berelated to general egocentric encoding (subsequent ego-centric remembering: E-CC > E-IC and A-IC > A-CC). Thissuggests at least a functional differentiation between thePACC and the dorsomedial frontal cortex with regard to thecreation of false memories about the communication situa-tion. Findings from studies by Kim and Cabeza (2007),Gonsalves et al. (2004), and Slotnick and Schacter (2004)actually refer to regions located quite close to the activatedlocation of the PACC found in our study. The correspond-ing cluster extent allows us to assume that there is at leastan overlap of the reported regions. This indicates that thePACC is involvednot only in false recognition (Kim&Cabeza,2007; Slotnick & Schacter, 2004) but also in the creation offalse memories (e.g., Gonsalves et al., 2004), whichmight berelated to self-referential encoding (e.g., Kelley et al., 2002)and mentalizing processes (e.g., Schulte-Ruther et al., 2007)triggered by the communication situation (seeNorthoff et al.,2006). This is in line with the theory that neural activity inthe ventral medial cortical structures (including the PACC)is involved in encoding the self-relatedness of stimuli andtherefore representing them as self-referential (see Northoffet al., 2006). Our data suggest that general egocentricencoding processes exist within the medial and left lateralfrontal cortex, whereas the PACC seems to be uniquelyinvolved in additional processes related to the creationof a self-referential episode in an allocentric context. Self-referential processing (Northoff et al., 2006) and vividimagining during stimulus processing (e.g., Gonsalves et al.,2004) might be the basis of processes that lead to correctitem memory but incorrect context memory. As indicatedby the association between bilateral activation of the BGand egocentric encoding (subsequent egocentric vs. allo-centric remembering), this effect might be mediated bythe reward system activated through self-referential pro-cesses (De Greck et al., 2008).We also found that activity in the hippocampus was re-

lated to sentence and context memory. The left hippo-campus (CA) appears to be related to memory of thesentence itself, independent of if the context in which itwas observed is subsequently remembered. Activation ofthe bilateral EC and subiculum was observed for correctlyremembered sentences with inaccurate memories of thecommunication context. In contrast, the right CA regionof the hippocampus is more active in trials in which thecontext is later correctly recalled. These findings are consis-tent with evidence that the hippocampus (predominantly

the CA region) plays a specific role in episodic or relationalmemory (e.g., Straube et al., 2009; Kircher et al., 2008;Kensinger & Schacter, 2006; Achim & Lepage, 2005;Eichenbaum, 2004; Ranganath et al., 2004; Davachi et al.,2003), but conflict with previous evidence that suggeststhat item and source memory both depend on the samemedial-temporal lobe structures (e.g., Gold, Hopkins, &Squire, 2006). We found unique patterns of activation inthe medial-temporal lobe structures for the encoding ofaccurate versus inaccurate episodic contextual information.The lateralization of hippocampal activity suggests that spa-tial encoding processes are involved when the communica-tion context is correctly remembered (e.g., Bird, Shallice, &Cipolotti, 2007; Crane & Milner, 2005; Parslow et al., 2005;Treyer, Buck, & Schnider, 2005; Feigenbaum & Morris,2004; Kessels, Hendriks, Schouten, van Asselem, & Postma,2004). This finding reflects the direct manipulation of a spa-tial variable (communication context/actorʼs orientation) ofthe communication episode. This result may suggest thataccurate memories of a communication context rely moreon visuospatial processes than on abstract-social processes.However, alternative explanations for the right hippocam-pal activation (like the encoding of interesting emotionalstimuli) are also plausible (see Hamann, Ely, Grafton, &Kilts, 1999).

The post hoc analyses of activation of the subregionsof the hippocampus (CA and EC) indicate that the hippo-campal activations are modulated in relation to memoryperformance only in the egocentric condition. This sug-gests that processes that bind communication contextand spoken sentence occur only for sentences remem-bered as egocentric that were actually encountered in anegocentric context. For items in the allocentric condition,the reducedmodulation of activation within the hippocam-pal subregions seems to be related to false memories of theencoding situation.

Further investigations into the unique encoding processesof multimodal communication (as well as the general pro-cesses responsible for item and source memory) are neces-sary to disentangle the underlying neural mechanisms. It isalso important to note that the influence of gestures on self-referential encoding processes cannot be disentangled inthis study. Gestures were included primarily to emphasizethe visual component of the communication. However, theymay have a fundamental influence on related memory pro-cesses. This should be clearly addressed in future studiesbecause gestures themselves play an important role inmem-ory encoding (Straube et al., 2009) and may be sufficient toelicit specific self-referential encoding processes. Further-more, it might be interesting in future studies to investigatethe effect of communication context with an implicit rec-ognition task and then compare the neural encoding andretrieval processes in relation to self-relevant stimuli.

Here we investigated implicit encoding without anyself or memory-related instruction. The previous rating(performed by an independent subject group) indicatesno differences in familiarity and understanding of de-

Straube et al. 321

scriptive or social contents of utterances in an egocentricor allocentric context. However, a participantʼs percep-tion of being addressed was significantly enhanced inthe egocentric in contrast to the allocentric condition.Egocentric encoding processes might support item mem-ory because egocentric and allocentric presented sentenceswere equally well remembered, although both were morelikely to be remembered as having been encountered in anegocentric context than items not seen during fMRI mea-surement. With regard to the idea of self-referential encod-ing, increased activation in the PACC might reflect theincreased processing effort for self-referential encoding inallocentric situations. The corresponding decreased activa-tion in the CA region of the hippocampus is possibly relatedto reduced attention or even inhibition of spatial informa-tion in allocentric contexts. This suggests that an egocentricbias and related self-referent encoding processes are thebasis of correct item memory but false memory about theencoding context of allocentric conditions.

In conclusion, our study demonstrated that distinct pat-terns of activation in cortical and hippocampal areas duringencoding are associated with accurate and inaccuratememories of past communications. In addition to support-ing theories that the hippocampus is involved in relational-associative memory, our results provide new insight intoprocesses that lead to false memories of social interactions.One such process may involve an egocentric bias that leadsparticipants to believe incorrectly they were directly ad-dressed in past allocentric social situations. Self-referentialencoding processes reflected in predominantly frontal acti-vation increase and hippocampal activation decrease mightbe the basis of successful item memory but false memoryof the social encoding context.

Acknowledgments

This research project was supported by a grant from the Inter-disciplinary Center for Clinical Research “BIOMAT” (IZKF VVN68) within the Faculty of Medicine at the RWTH Aachen Uni-versity, Germany. The authors thank Katharina Augustin, BettinaFreese, and Simone Schröder for help with the preparation andevaluation of the stimulus material, Thilo Kellermann for assis-tance with the fMRI analyses, and Bianca Bromberger for helpwith the editing of the manuscript.

Reprint requests should be sent to Dr. Benjamin Straube, Depart-ment of Psychiatry undPsychotherapy, Philipps-UniversityMarburg,Rudolf-Bultmann Str. 8, 35039 Marburg, Germany, or via e-mail:straubeb@med.uni-marburg.de.

REFERENCES

Achim, A. M., & Lepage, M. (2005). Neural correlates of memoryfor items and for associations: An event-related functionalmagnetic resonance imaging study. Journal of CognitiveNeuroscience, 17, 652–667.

Amodio, D. M., & Frith, C. D. (2006). Meeting of minds: Themedial frontal cortex and social cognition. Nature ReviewsNeuroscience, 7, 268–277.

Amunts, K., Kedo, O., Kindler, M., Pieperhoff, P., Mohlberg, H.,Shah, N. J., et al. (2005). Cytoarchitectonic mapping of thehuman amygdala, hippocampal region and entorhinal cortex:Intersubject variability and probability maps. Anatomy andEmbryology, 210, 343–352.

Baird, A., Dewar, B. K., Critchley, H., Dolan, R., Shallice, T.,& Cipolotti, L. (2006). Social and emotional functions inthree patients with medial frontal lobe damage includingthe anterior cingulate cortex. Cognitive Neuropsychiatry,11, 369–388.

Bird, C. M., Shallice, T., & Cipolotti, L. (2007). Fractionation ofmemory in medial temporal lobe amnesia. Neuropsychologia,45, 1160–1171.

Cabeza, R., Ciaramelli, E., Olson, I. R., & Moscovitch, M. (2008).The parietal cortex and episodic memory: An attentionalaccount. Nature Reviews Neuroscience, 9, 613–625.

Cabeza, R., & Nyberg, L. (2000). Neural bases of learning andmemory: Functional neuroimaging evidence. CurrentOpinion in Neurology, 13, 415–421.

Cabeza, R., Rao, S. M., Wagner, A. D., Mayer, A. R., &Schacter, D. L. (2001). Can medial temporal lobe regionsdistinguish true from false? An event-related functionalMRI study of veridical and illusory recognition memory.Proceedings of the National Academy of Sciences, U.S.A.,98, 4805–4810.

Cheng, Y., Lin, C. P., Liu, H. L., Hsu, Y. Y., Lim, K. E., Hung, D.,et al. (2007). Expertise modulates the perception of pain inothers. Current Biology, 17, 1708–1713.

Crane, J., & Milner, B. (2005). What went where? Impairedobject-location learning in patients with right hippocampallesions. Hippocampus, 15, 216–231.

Cunnington, R., Windischberger, C., Robinson, S., & Moser, E.(2006). The selection of intended actions and the observationof othersʼ actions: A time-resolved fMRI study. Neuroimage,29, 1294–1302.

Curran, T., Schacter, D. L., Norman, K. A., & Galluccio, L.(1997). False recognition after a right frontal lobeinfarction: Memory for general and specific information.Neuropsychologia, 35, 1035–1049.

Davachi, L., Mitchell, J. P., & Wagner, A. D. (2003). Multipleroutes to memory: Distinct medial temporal lobe processesbuild item and source memories. Proceedings of theNational Academy of Sciences, U.S.A., 100, 2157–2162.

David, N., Bewernick, B. H., Cohen, M. X., Newen, A., Lux, S.,Fink, G. R., et al. (2006). Neural representations of self versusother: Visual-spatial perspective taking and agency in a virtualball-tossing game. Journal of Cognitive Neuroscience, 18,898–910.

De Greck, M., Rotte, M., Paus, R., Moritz, D., Thiemann, R.,Proesch, U., et al. (2008). Is our self based on reward? Self-relatedness recruits neural activity in the reward system.Neuroimage, 39, 2066–2075.

Dick, A. S., Goldin-Meadow, S., Hasson, U., Skipper, J. I., &Small, S. L. (2009). Co-speech gestures influence neuralactivity in brain regions associated with processing semanticinformation. Human Brain Mapping, 30, 3509–3526.

Eichenbaum, H. (2004). Hippocampus: Cognitive processesand neural representations that underlie declarative memory.Neuron, 44, 109–120.

Eichenbaum, H., & Cohen, N. J. (2001). From conditioning toconscious recollection: Memory systems of the brain.Oxford: Oxford University Press.

Eickhoff, S. B., Paus, T., Caspers, S., Grosbras, M. H., Evans,A. C., Zilles, K., et al. (2007). Assignment of functionalactivations to probabilistic cytoarchitectonic areas revisited.Neuroimage, 36, 511–521.

Eickhoff, S. B., Stephan, K. E., Mohlberg, H., Grefkes, C., Fink,G. R., Amunts, K., et al. (2005). A new SPM toolbox for

322 Journal of Cognitive Neuroscience Volume 23, Number 2

combining probabilistic cytoarchitectonic maps andfunctional imaging data. Neuroimage, 25, 1325–1335.

Feigenbaum, J. D., & Morris, R. G. (2004). Allocentric versusegocentric spatial memory after unilateral temporallobectomy in humans. Neuropsychology, 18, 462–472.

Friston, K. J., Fletcher, P., Josephs, O., Holmes, A., Rugg, M. D.,& Turner, R. (1998). Event-related fMRI: Characterizingdifferential responses. Neuroimage, 7, 30–40.

Frith, C. D., & Frith, U. (1999). Interacting minds-A biologicalbasis. Science, 286, 1692–1695.

Gillihan, S. J., & Farah, M. J. (2005). Is self special? A criticalreview of evidence from experimental psychology andcognitive neuroscience. Psychological Bulletin, 131, 76–97.

Gold, J. J., Hopkins, R. O., & Squire, L. R. (2006). Single-itemmemory, associative memory, and the human hippocampus.Learning and Memory, 13, 644–649.

Goldin-Meadow, S. (1999). The role of gesture in communicationand thinking. Trends in Cognitive Sciences, 3, 419–429.

Gonsalves, B., & Paller, K. A. (2000). Neural events that underlieremembering something that never happened. NatureNeuroscience, 3, 1316–1321.

Gonsalves, B., Reber, P. J., Gitelman, D. R., Parrish, T. B.,Mesulam, M. M., & Paller, K. A. (2004). Neural evidence thatvivid imagining can lead to false remembering. PsychologicalScience, 15, 655–660.

Green, A., Straube, B., Weis, S., Jansen, A., Willmes, K., Konrad,K., et al. (2009). Neural integration of iconic and unrelatedcoverbal gestures: A functional MRI study. Human BrainMapping, 30, 3309–3024.

Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999).Amygdala activity related to enhanced memory for pleasantand aversive stimuli. Nature Neuroscience, 2, 289–293.

Herwig, U., Abler, B., Schonfeldt-Lecuona, C., Wunderlich, A.,Grothe, J., Spitzer, M., et al. (2003). Verbal storage ina premotor-parietal network: Evidence from fMRI-guidedmagnetic stimulation. Neuroimage, 20, 1032–1041.

Hietanen, J. K. (2002). Social attention orienting integratesvisual information from head and body orientation.Psychological Research, 66, 174–179.

Hietanen, J. K., Nummenmaa, L., Nyman, M. J., Parkkola, R., &Hamalainen, H. (2006). Automatic attention orienting bysocial and symbolic cues activates different neural networks:An fMRI study. Neuroimage, 33, 406–413.

Holle, H., Gunter, T. C., Rüschemeyer, S.-A., Hennenlotter, A.,& Iacoboni, M. (2008). Neural correlates of the processing ofco-speech gestures. Neuroimage, 39, 2010–2024.

Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Sourcemonitoring. Psychological Bulletin, 114, 3–28.

Johnson, M. K., & Raye, C. L. (1981). Reality monitoring.Psychological Review, 88, 67–85.

Kahn, I., Davachi, L., & Wagner, A. D. (2004). Functional-neuroanatomic correlates of recollection: Implications formodels of recognition memory. Journal of Neuroscience,24, 4172–4180.

Kedia, G., Berthoz, S., Wessa, M., Hilton, D., & Martinot, J. L.(2008). An agent harms a victim: A functional magneticresonance imaging study on specific moral emotions.Journal of Cognitive Neuroscience, 20, 1788–1798.

Kelley, W. M., Macrae, C. N., Wyland, C. L., Caglar, S., Inati, S., &Heatherton, T. F. (2002). Finding the self? An event-relatedfMRI study. Journal of Cognitive Neuroscience, 14, 785–794.

Kendon, A. (2004). Some contrasts in gesticulation inNeapolitan speakers and speakers in Northamptonshire. In R.Posner & C. Mueller (Eds.), The semantics and pragmaticsof everyday gesture (pp. 173–193). Berlin: WeidlerBuchverlag.

Kensinger, E. A., & Schacter, D. L. (2006). Amygdala activity isassociated with the successful encoding of item, but not

source, information for positive and negative stimuli.Journal of Neuroscience, 26, 2564–2570.

Kessels, R. P., Hendriks, M., Schouten, J., van Asselem, M., &Postma, A. (2004). Spatial memory deficits in patientsafter unilateral selective amygdalohippocampectomy.Journal of the International Neuropsychological Society,10, 907–912.

Kim, H., & Cabeza, R. (2007). Trusting our memories:Dissociating the neural correlates of confidence in veridicalversus illusory memories. Journal of Neuroscience, 2745,12190–12197.

Kircher, T., Blumel, I., Marjoram, D., Lataster, T., Krabbendam,L., Weber, J., et al. (2009). Online mentalising investigatedwith functional MRI. Neuroscience Letters, 454, 176–181.

Kircher, T., Straube, B., Leube, D., Weis, S., Sachs, O., Willmes,K., et al. (2009). Neural interaction of speech and gesture:Differential activations of metaphoric co-verbal gestures.Neuropsychologia, 471, 169–179.

Kircher, T., Weis, S., Freymann, K., Erb, M., Jessen, F., Grodd, W.,et al. (2007). Hippocampal activation in patients with mildcognitive impairment is necessary for successful memoryencoding. Journal of Neurology, Neurosurgery, andPsychiatry, 78, 812–818.

Kircher, T., Weis, S., Leube, D., Freymann, K., Erb, M., Jessen, F.,et al. (2008). Anterior hippocampus orchestrates successfulencoding and retrieval of non-relational memory: An event-related fMRI study. European Archives of Psychiatry andClinical Neuroscience, 258, 363–372.

Kircher, T. T. J., Senior, C., Phillips, M. L., Benson, P. J.,Bullmore, E. T., Brammer, M., et al. (2000). Towards afunctional neuroanatomy of self processing: Effects of facesand words. Cognitive Brain Research, 10, 133–144.

Kircher, T. T. J., Senior, C., Phillips, M. L., Rabe-Hesketh, S.,Benson, P. J., Bullmore, E. T., et al. (2001). Recognizing oneʼsown face. Cognition, 78, B1–B15.

Kirchhoff, B. A., Wagner, A. D., Maril, A., & Stern, C. E. (2000).Prefrontaltemporal circuitry for episodic encoding andsubsequent memory. Journal of Neuroscience, 20, 6173–6180.

Krach, S., Blümel, I., Marjoram, D., Lataster, T., Krabbendam, L.,Weber, J., et al. (2009). Are women better mindreaders? Sexdifferences in neural correlates of mentalizing detected withfunctional MRI. BMC Neuroscience, 10, 9.

Krach, S., Hegel, F., Wrede, B., Sagerer, G., Binkofski, F., &Kircher, T. (2008). Can machines think? Interaction andperspective taking with robots investigated via fMRI. PLoSONE, 3, e2597.

Kubota, Y., Toichi, M., Shimizu, M., Mason, R. A., Findling, R. L.,Yamamoto, K., et al. (2006). Prefrontal hemodynamic activitypredicts false memory-A near-infrared spectroscopy study.Neuroimage, 31, 1783–1789.

Lawrence, E. J., Shaw, P., Giampietro, V. P., Surguladze, S.,Brammer, M. J., & David, A. S. (2006). The role of “sharedrepresentations” in social perception and empathy: An fMRIstudy. Neuroimage, 29, 1173–1184.

Lieberman, M. D. (2007). Social cognitive neuroscience: Areview of core processes. Annual Review of Psychology, 58,259–289.

Macmillan, N. A., & Creelman, C. D. (1991). Detection theory: Auserʼs guide. New York: Cambridge University Press.

Macrae, C. N., Moran, J. M., Heatherton, T. F., Banfield, J. F., &Kelley, W. M. (2004). Medial prefrontal activity predictsmemory for self. Cerebral Cortex, 14, 647–654.

McNeill, D. (1992). Hand and mind: What gestures revealabout thought. Chicago, IL: University of Chicago Press.

Mühlau, M., Hermsdorfer, J., Goldenberg, G., Wohlschlager,A. M., Castrop, F., Stahl, R., et al. (2005). Left inferiorparietal dominance in gesture imitation: An fMRI study.Neuropsychologia, 43, 1086–1098.

Straube et al. 323

Naghavi, H. R., & Nyberg, L. (2005). Common fronto-parietalactivity in attention, memory, and consciousness: Shareddemands on integration? Consciousness and Cognition,14, 390–425.

Northoff, G., & Bermpohl, F. (2004). Cortical midline structuresand the self. Trends in Cognitive Sciences, 8, 102–107.

Northoff, G., Heinzel, A., de Greck, M., Bermpohl, F.,Dobrowolny, H., & Panksepp, J. (2006). Self-referentialprocessing in our brain-A meta-analysis of imaging studies onthe self. Neuroimage, 31, 440–457.

Okado, Y., & Stark, C. (2005). Neural activity during encodingpredicts false memories created by misinformation. Learningand Memory, 12, 3–11.

Oldfield, R. C. (1971). The assessment and analysis of handedness:The Edinburgh inventory. Neuropsychologia, 9, 97–113.

Otten, L. J., Henson, R. N., & Rugg, M. D. (2001). Depth ofprocessing effects on neural correlates of memory encoding:Relationship between findings from across- and within-taskcomparisons. Brain, 124, 399–412.

Özyürek, A. (2002). Do speakers design their cospeech gesturesfor their addressees? The effects of addressee locationon representational gestures. Journal of Memory andLanguage, 46, 688–704.

Parslow, D. M., Morris, R. G., Fleminger, S., Rahman, Q.,Abrahams, S., & Recce, M. (2005). Allocentric spatial memoryin humans with hippocampal lesions. Acta Psychologica, 118,123–147.

Ranganath, C., Yonelinas, A. P., Cohen, M. X., Dy, C. J., Tom,S. M., & DʼEsposito, M. (2004). Dissociable correlates ofrecollection and familiarity within the medial temporal lobes.Neuropsychologia, 42, 2–13.

Rogers, T. B., Kuiper, N. A., & Kirker, W. S. (1977). Self-reference and the encoding of personal information. Journalof Personality and Social Psychology, 35, 677–688.

Ross, M., & Sicoly, F. (1979). Egocentric biases in availability andattribution. Journal of Personality and Social Psychology,37, 322–336.

Schacter, D. L., Buckner, R. L., Koutstaal, W., Dale, A. M., &Rosen, B. R. (1997). Late onset of anterior prefrontal activityduring true and false recognition: An event-related fMRIstudy. Neuroimage, 6, 259–269.

Schacter, D. L., Chiao, J. Y., & Mitchell, J. P. (2003). The sevensins of memory: Implication for self. Annals of the New YorkAcademy of Sciences, 1001, 226–239.

Schacter, D. L., Curran, T., Galluccio, L., Milberg, W. P., & Bates,J. F. (1996). False recognition and the right frontal lobe: Acase study. Neuropsychologia, 34, 793–808.

Schacter, D. L., Reiman, E., Curran, T., Yun, L. S., Bandy, D.,McDermott, K. B., et al. (1996). Neuroanatomical correlatesof veridical and illusory recognition memory: Evidence fromposition emission tomography. Neuron, 17, 267–274.

Schacter, D. L., & Slotnick, S. D. (2004). The cognitiveneuroscience of memory distortion. Neuron, 44, 149–160.

Schulte-Ruther, M., Markowitsch, H. J., Fink, G. R., & Piefke, M.(2007). Mirror neuron and theory of mind mechanismsinvolved in face-to-face interactions: A functional magneticresonance imaging approach to empathy. Journal ofCognitive Neuroscience, 19, 1354–1372.

Slotnick, S. D., Moo, L. R., Segal, J. B., & Hart, J., Jr. (2003).Distinct prefrontal cortex activity associated with itemmemory and source memory for visual shapes. BrainResearch, Cognitive Brain Research, 17, 75–82.

Slotnick, S. D., & Schacter, D. L. (2004). A sensory signaturethat distinguishes true from false memories. NatureNeuroscience, 7, 664–672.

Stark, C. E., & Okado, Y. (2003). Making memories withouttrying: Medial temporal lobe activity associated withincidental memory formation during recognition. Journal ofNeuroscience, 17, 6748–6753.

Straube,B.,Green, A., Jansen, A., Chatterjee, A.,&Kircher, T. (2010).Social cues, mentalizing and the neural processing of speechaccompanied by gestures. Neuropsychologia, 48, 382–393.

Straube, B., Green, A., Weis, S., Chatterjee, A., & Kircher, T.(2009). Memory effects of speech and gesture binding:Cortical and hippocampal activation in relation to subsequentmemory performance. Journal of Cognitive Neuroscience,21, 821–836.

Swick, D., & Knight, R. T. (1999). Contributions of prefrontalcortex to recognition memory: Electrophysiological andbehavioral evidence. Neuropsychology, 13, 155–170.

Symons, C. S., & Johnson, B. T. (1997). The self-referenceeffect in memory: A meta-analysis. Psychological Bulletin,121, 371–394.

Treyer, V., Buck, A., & Schnider, A. (2005). Processing contentor location: Distinct brain activation in a memory task.Hippocampus, 15, 684–689.

Vilberg, K. L., & Rugg, M. D. (2008). Memory retrieval and theparietal cortex: A review of evidence from a dual-processperspective. Neuropsychologia, 46, 1787–1799.

Vincent, J. L., Snyder, A. Z., Fox, M. D., Shannon, B. J., Andrews,J. R., Raichle, M. E., et al. (2006). Coherent spontaneousactivity identifies a hippocampal-parietal memory network.Journal of Neurophysiology, 96, 3517–3531.

Vogeley, K., May, M., Ritzl, A., Falkai, P., Zilles, K., & Fink, G. R.(2004). Neural correlates of first-person perspective as oneconstituent of human self-consciousness. Journal ofCognitive Neuroscience, 16, 817–827.

Wagner, A. D., Shannon, B. J., Kahn, I., & Buckner, R. L. (2005).Parietal lobe contributions to episodic memory retrieval.Trends in Cognitive Sciences, 99, 445–453.

Weiss, P. H., Rahbari, N. N., Lux, S., Pietrzyk, U., Noth, J., &Fink, G. R. (2006). Processing the spatial configuration ofcomplex actions involves right posterior parietal cortex:An fMRI study with clinical implications. Human BrainMapping, 27, 1004–1014.

Wheaton, K. J., Thompson, J. C., Syngeniotis, A., Abbott, D. F.,& Puce, A. (2004). Viewing the motion of human body partsactivates different regions of premotor, temporal, andparietal cortex. Neuroimage, 22, 277–288.

Wheeler, M. E., & Buckner, R. L. (2003). Functionaldissociation among components of remembering: Control,perceived oldness, and content. Journal of Neuroscience,23, 3869–3880.

Willems, R. M., Ozyurek, A., & Hagoort, P. (2007). Whenlanguage meets action: The neural integration of gesture andspeech. Cerebral Cortex, 17, 2322–2333.

324 Journal of Cognitive Neuroscience Volume 23, Number 2