Implicit emotion during recollection of past events: A nonverbal fMRI study
Introduction
Remembering by consciously re-experiencing detailed spatially and temporally specific events of one's own life characterises autobiographical memory (Tulving, 2002). Functional neuroimaging techniques used to explore the neural bases of this memory system in healthy people show an extended medial and predominantly left-sided cerebral network (for reviews, see Gilboa 2004; Maguire 2001) more frequently than right-lateralised (e.g., Fink et al., 1996) or bilateral (Graham et al., 2003) activations. Refreshment of the memory trace before the imaging experiment and the use of verbal stimuli to cue autobiographical recollections were the two methodological issues deemed likely to influence the left-lateralisation findings (Maguire 2001; Piefke et al. 2003). Recently, Gilboa et al. (2004) investigated retrieval of personally experienced events by means of family photographs, a technique that presents the advantage of avoiding reactivation of the memories prior to the scanning session using fMRI devices and an ecological method of studying recollection of past events. The authors reported MTL involvement during the remembering of both recent and remote vivid episodes and predominantly left-lateralised prefrontal activation which seemed to characterise autobiographical memory retrieval (Gilboa, 2004).#
Human faces are the most frequently used nonverbal stimuli in neuropsychology (McCarthy and Warrington, 1990). According to the degree of familiarity, facial stimuli usually belong either to the category of famous faces, known mostly through the media, or to the category of personally known faces of relatives, friends and acquaintances. The former have been used in neuroimaging studies to explore the cerebral correlates of semantic memory as identification of these faces involves retrieval of person–identity information stored in long-term memory (Bernard et al. 2004; Haist et al. 2001; Leveroni et al. 2000). Concerning faces of personal acquaintance, such as family members, friends and colleagues, Ida Gobbini et al. (2004) demonstrated that they induced stronger cerebral activations than famous faces. The authors suggested that processing personally known faces might be spontaneously associated, beyond the facial recognition, with knowledge about a person's personality, episodic memories and an emotional response to that person. Next of kin faces were typically used to investigate brain areas responsible for producing different affective states since they belong to one's own life history and they are emotionally indistinguishable to an external observer (for example, Gundel et al. 2003; Leibenluft et al. 2004; Nitschke et al. 2004). Studying romantic and maternal love, i.e., complex affective states, Bartels and Zeki (2000 2004) showed a specific network of brain areas when an individual views the face of “someone who elicits a unique and characteristic set of emotions” (Bartels and Zeki, 2000; p. 3833). To investigate the emotional aspect of autobiographical memory, we used the same stimuli to warrant the affective-laden feature of the recollections. Indeed, emotion is a crucial component of autobiographical memory and one of the most complicated (Greenberg and Rubin, 2003). Several neuroimaging studies have explored the influence of emotion on the cerebral network underlying personal recollections using different experimental designs and obtaining, not surprisingly, a variety of results. Thus, some studies compared brain networks sustaining positive and negative memories selected prior to the imaging experiment (Markowitsch et al. 2003; Piefke et al. 2003), while other reports investigated how emotion modulated cerebral activations (Addis et al., 2004, emotional intensity; Maguire and Frith, 2003a, emotional valence and intensity). In general, the medial prefrontal cortex, considered as one of the key structures in the literature on emotion (Phan et al., 2002), has been almost invariably found to be activated in autobiographical studies “as memories carry emotional meaning” (Gilboa, 2004, p. 1346). On the contrary, the activation of amygdala was not directly correlated with emotional intensity or the valence of autobiographical recollections (Addis et al. 2004; Maguire and Frith 2003a). Although the neural correlates of emotional memory have been investigated within the context of autobiographical memory, none of the neuroimaging studies, to our knowledge, has documented spontaneous remembering of emotionally laden recollections with no previous testing.#
The aim of the present study was to investigate the neural correlates of emotion induced by spontaneous (i.e., close to real life) remembering of personal episodes. To address this issue, we used highly self-relevant stimuli, namely, photographs of relatives' and friends' faces avoiding refreshment of the memory traces before the neuroimaging experiment.#
Results
Behavioural data
We obtained 81.4% (SD = 4.8) of recollections of specific autobiographical events for the autobiographical memory condition and 88.8% (SD = 3.58) of correct identification of famous faces. The corresponding reaction times were 5.13 s (SD = 2.50 s) and 1.35 s (SD = 0.48 s), respectively. The total duration of the autobiographical condition was 215.25 s (SD = 113.89 s) and that of the control condition was 59.70 s (SD = 21.23 s). When we discarded the personal recollections associated with the longest reaction times in order to match the two conditions for duration, the total duration of the experimental condition was reduced to 70.39 s (SD = 25.15 s).#
The emotional scale yielded an average result corresponding to the “slightly emotional” interval (mean = 2.25, SD = 0.79). At one end of the scale, one subject did not report any emotion, all her recollections being rated 1. At the other end, only one participant provided ratings between 3 and 5. The post-scanning subjects' responses on the emotional scale did not allow us to compare recollections rated as high emotional (rated 4, 5) with those that are rated low or non-emotional (rated 1, 2) since 58.87% of memories were low or non-emotional (see Table 1).#
The mean age of recollections across participants was 7.67 years (range: 1 week–40 years).#
fMRI data
The significant activations for the comparison autobiographical memory vs. control condition were all found when a model based on the canonical haemodynamic response function alone was applied; no further significant activations were observed with two derivatives.#
The contrasts of autobiographical retrieval vs. semantic retrieval yielded a distributed set of bilateral but predominantly left-sided brain activations (see Fig. 1) that involved lateral (dorsolateral BA 8, 9, 46 and ventrolateral cortex BA 47) and medial prefrontal regions (including anterior cingulate cortex BA 24, 32), the temporo–parieto–occipital junction (BA 39, 40, 19), the precuneus (BA 7), the fusiform gyrus (BA 20), the subcortical areas (head of the caudate nucleus, putamen and thalamus), the region overlapping the periaqueductal grey (PAG) of the midbrain and the cerebellum. Bilateral but predominantly right-sided activation was observed in the posterior cingulate cortex (BA 30, 31) (see Table 2).#
Likewise, although the MTL activation was bilateral, the local maximum for this structure was right-lateralised (R > L) (right: 20, −32, −20; Z score = 4.49, k = 99; left: −22, −36, −20; Z score = 3.74, k = 24) (Fig. 2). Since the left MTL activation was included within a large cluster comprising cerebellum, midbrain and fusiform gyrus (k = 341), it was detected by a small volume correction approach (a sphere of 5 mm radius centred on −20, −32, −20).#
Discussion
We have documented the influence of nonverbal self-relevant stimuli, presented with no refreshment of the memory trace, on emotional processing, hemispheric lateralisation and MTL involvement during retrieval of personal recollections.#
Our central finding is the predominantly right-sided increased activation of the MTL during the retrieval of personal lifespan events. The left-sided activation of the MTL and the prefrontal regions has been reported in the majority of autobiographical memory neuroimaging reports (for reviews, see Gilboa 2004; Maguire 2001), whereas few studies have demonstrated preferentially right-sided hippocampal involvement. Fink et al. (1996) suggested that activations in the right MTL might reflect the emotional character of recalled autobiographical memories. Piolino et al. (2004) pointed out the time for retrieval (45 s) that permits recollection of detailed memories. In the same vein, Gilboa et al. (2004) suggested that the right hippocampus takes longer than the left to reach a peak response (from 6 to 8 s). Using a design where the retrieval time was fixed by the subjects themselves, we observed right parahippocampal activation during autobiographical recollections that lasted for about 5 s (see Results). It appears, therefore, that the right MTL activation might reflect some other factors besides the time allowed for retrieval. In line with our own study, the four recent works (Gilboa et al. 2004; Levine et al. 2004; Piolino et al. 2004; Tsukiura et al. 2002) that revealed right hippocampal or parahippocampal activations all avoided refreshment of the memory trace prior to scanning. This points to the possibility that remembering past events not previously ‘reactivated’ might be more readily emotionally marked, hence the right-sided MTL activations.#
Concerning the left-lateralised frontal activation frequently reported in previous neuroimaging studies of autobiographical memory (Gilboa, 2004) as well as in the present work, the dorsolateral prefrontal cortex (DLPFC) was interpreted as sustaining the anterior control processes that accompany effortful reconstruction of recollections whereas the medial prefrontal cortex (MPFC) has an essential role in the processing of self-relevant information.#
The self, in Tulving's theory, is one of the three central components of autobiographical memory, together with autonoetic awareness and subjectively sensed time (Tulving, 2002). Recently, Northoff and Bermpohl (2004) discussed the relationship between self-referential processing and the cortical midline structures, which include four regions, each involved in different processes: the orbital and medial prefrontal cortex seems to be implicated in the representation of self-referential stimuli which are monitored in the supragenual anterior cingulate cortex and evaluated in the dorsomedial prefrontal cortex; the posterior cingulate cortex (PCC) plays a role in “the integration of these stimuli in the emotional and autobiographical context of one's own person” (p. 104). The MPFC together with the PCC have very frequently been reported in the autobiographical neuroimaging literature (for example, Maguire and Frith, 2003b) as well as in studies of ‘theory of mind’ (e.g., den Ouden et al. 2005; Frith and Frith 2003). This is not surprising as both remembering past events and mentalising require self-referential processing. In addition, the already mentioned midline regions (MPFC and PCC) are also implicated in processing emotional information (for a review, see Northoff and Bermpohl, 2004). This is particularly the case for stimuli that are processed as emotional on the basis of their self-relatedness, as determined by the emotional memories that an individual has associated with them (Bechara et al. 2003; Phan et al. 2004), as in the current study.#
Several recent neuroimaging studies have investigated the influence of emotion on the cerebral network underlying autobiographical recollections. These studies used verbal stimuli previously collected from the participants and discussed preferentially the relationship between emotion and the amygdala (for example, Addis et al. 2004; Markowitsch et al. 2000). Our subjects' responses on the emotion scale precluded any attempt to study emotionality by comparing brain activations sustaining intense with slightly emotional memories (see Results). Nevertheless, despite the low emotionality rating, we observed the involvement of cerebral regions that are related to emotion processing, namely, the MPFC, the anterior cingulate cortex (ACC), the orbitofrontal region and subcortical structures (Phan et al., 2002). As mentioned above, Bartels and Zeki (2004), studying attachment, reported increased activations in the lateral orbitofrontal cortex, the ACC and the PAG when mothers were presented with photographs of the face of their own child. These regions were interpreted as reflecting maternal behaviour and pleasant emotions associated with maternal love. It appears, therefore, that the PAG and ventrolateral prefrontal activations revealed in the present study may sustain the emotional aspects not consciously involved in autobiographical recollections cued with faces of next of kin.#
We reported striatal activations, namely, in the caudate nucleus and putamen, and, bearing in mind that the photographs included those of partner and child, we interpret this observation in accordance with Bartels and Zeki (2004). These authors suggested that particular subregions of the reward circuitry in the human striatum reveal a general network specialised in the mediation of romantic and maternal attachment.#
The absence of increased amygdala activation in our study is not surprising in view of a recent meta-analysis study that states that “very few recall-driven emotion activation studies engaged the amygdala (at 7% frequency)” (Phan et al., 2002, p. 343). This conclusion is also in accordance with recent autobiographical memory studies that failed to find direct correlation between amygdala activity and ratings of emotional valence and/or intensity (Addis et al. 2004; Maguire and Frith 2003a).#
Turning back to the important issue of the apparent contradiction between our behavioural and neuroimaging data, we ask why our participants reported low emotional intensity where we observed increased activation in regions associated with emotion. Several reasons could account for such a discrepancy. Firstly, according to Davidson (2003), the verbal self-report on emotion should be interpreted with caution since it is not always a reliable indication of the presence or the absence of an emotional state: he asserted that “much of the affect that we generate is likely to be non-conscious” and that there are “regulatory processes in emotion that occur automatically” (p. 131). Secondly, the increased cognitive effort necessary for retrieval of memories not previously refreshed could divert the attentional resources from the emotional value of recollections (Phan et al., 2004). Thirdly, the fMRI environment, namely, limited space, closeness of the head coil, strong acoustic noise and restriction of movement, could act as a psychological stressor (Raz et al., 2005) and thus render the participant unaware of the emotional re-experiencing of a personal event. Our participants did indeed report that the fMRI condition had inhibited their ability to fully re-live the original emotion during recollections.#
In future studies, it would be of interest to use behavioural indices such as the electro-galvanic skin response in order to probe as objectively as possible the emotional states accompanying the retrieval of personal memories. Another goal for future research should be to distinguish a possible early emotion induced by mere face presentation from emotion evoked following recall of a life event.#
In conclusion, the present study provides evidence that the autobiographical recollections cued with highly self-related stimuli presented directly during the scanning session involved the right MTL, cortical midline structures and subcortical circuits, known to sustain the self-generated emotion, even though no emotion was explicitly acknowledged.#
Experimental procedures
Subjects
Ten healthy subjects (4 males and 6 females) participated in the fMRI study, which was approved by the local ethics committee. The subjects were right-handed and native French speakers (mean age = 40.6 years; SD = 5.7; mean years of education = 18.0; SD = 2.8). As a matter of routine, subjects were screened to rule out medical or neurological problems, current medications and/or MRI contraindications. They had normal or corrected to normal vision. The subjects gave written informed consent and were paid for their participation.#
Stimuli and tasks
The experimental fMRI stimuli consisted of 590 coloured photographs (50 pictures of relatives and friends × 10 participants, plus a pool of 50 famous faces and 40 unknown faces used for the 10 subjects) processed using Adobe PhotoShop (version 7.0) to have the same size (250 × 300 pixels) and neutral background in order to guarantee similar perceptual input. Luminance and contrast were adjusted to make the images as comparable as possible. Both the personally familiar faces and the famous faces belonged in different periods of the past (from 1950 to 2004). The 50 photographs per participant of faces of next of kin, other relations and close friends were the cues for the ‘autobiographical task’. The photographs were collected with the help of the participant's family (parents, spouse and, in a few cases, children). The participant did not know which photographs were selected by the examiner and the precise task he/she would be asked to perform during the fMRI session. The collection and preparation of photographs were extremely time-consuming. The pictures were all scanned using a Hewlett Packard Scan Jet 63,000 C. The faces of people relevant for the subject were extracted from photographs and made up the pool of experimental cue stimuli (one face = one stimulus). No two stimuli were identical. Photographs of the subject were never presented to avoid confounding cerebral activations sustaining autobiographical memory with those underlying self-recognition. Our decision to show faces alone was to minimise the differences in visual stimulation.#
In line with several functional neuroimaging studies (Greenberg et al. 2005; Maguire and Mummery 1999), we chose a control condition involving semantic retrieval, namely, the identification of faces of celebrities. Fifty famous faces constituted the stimuli of the control task. The faces were those of politicians, actors, musicians and sports figures whose fame is associated with a particular period from the 1950s to the present (e.g., President J.F. Kennedy, Celine Dion). They were selected through a behavioural pilot study (n = 18 normal subjects) in which we carried out a random presentation of 70 photographs of celebrities and 30 photographs of unknown faces on a computer screen using INQUISIT millisecond software (Psychological Data Collection Software for Windows). We retained the 50 photographs of famous faces that were associated with the fastest reaction time for correct recognition scores for the fMRI experiment.#
Selection of the unknown faces was carried out as follows: half of them were collected through the behavioural study mentioned above on the basis of the reaction time for rejecting a foil. The other half consisted of the faces of other participants' relatives. Independently of the condition, all unfamiliar faces were used as foils to maintain the subject's attention during the tasks. The first 20 unknown faces were presented as distractors during the ‘control condition’ and the remaining 20 items as distractors in the ‘autobiographical condition’.#
Prior to the scanning session, subjects were given detailed instructions about the two tasks (autobiographical and control tasks). Examples of autobiographical events and a practice trial were provided in order to ensure that subjects tried to remember a specific single episode from their past rather than general or repeated events. (i) In the autobiographical task, the subjects were asked to recollect a personal episode when a face of relative or friend appeared on the screen. They were instructed to press the left button of a PC mouse (‘yes’) once retrieval of an autobiographical event was over. If they viewed the face of an unknown person (distractor) or if they failed to evoke a specific autobiographical episode, the right button (‘no’) was to be pressed. (ii) In the control task, the participants were instructed to press the left button (‘yes’) if they successfully identified the face of a famous person and the right button (‘no’) if the face was unknown (distracter). The subjects were asked to focus on the centre of the display throughout the fMRI session.#
fMRI design
Retrieval of autobiographical memories is an effortful process characterised by variable speed in evoking memories between and within subjects. Therefore, the self-paced event-related fMRI design (Daselaar et al., 2001) was chosen as a flexible and suitable approach for autobiographical remembering during scanning. The fMRI conditions, ‘autobiographical’ and ‘control’, each comprised of ten sequences with seven faces per sequence. Independently of the condition, the seven faces forming any sequence were presented in a random order and with the same 5:2 ratio of known to unknown faces. Thus, five faces of relatives and friends and two unfamiliar faces were presented within an ‘autobiographical’ sequence and five famous faces and two unknown faces were presented within a ‘control’ sequence. The two types of sequences were presented alternately and were separated by a fixation cross that lasted 15.0 s. Each facial stimulus was shown during 1.24 s on the screen, which was blank during the interval between stimulus offset and onset of the next face. Although presentation of facial stimuli was self-paced, a time limit of 11.24 s was maintained in the event of no response in the autobiographical condition and a time limit of 3.74 s in the control condition. These time limits were based on pilot experiments showing that all the responses fell within these interval periods (Botzung et al., 2003). Once the subject's response was given or the time limit reached, the post-trial interval started. This interval varied from 2.0 to 5.0 s in the autobiographical condition and from 1.5 to 4.5 s in the control condition according to the variable time limit duration in the two conditions. The variable inter-stimuli intervals (“jitters”) as well as the random presentation of stimuli within a sequence have the advantage of reducing potential confounds, such as habituation, anticipation or set strategy effects (Aguirre and D'Esposito 1999; Rosen et al. 1998).#
Presentation, timing of stimuli and response recording were performed using INQUISIT millisecond software (see above). Faces were projected onto the centre of a screen which was viewed by subjects wearing prism glasses.#
Post-scanning debriefing
Immediately after scanning, we carried out a debriefing session. Each participant was shown again the faces of subjects' relatives and friends he/she had just viewed during the scanning session. The subjects were asked to verbalise the events they evoked during the fMRI session, and they rated each memory on a five-point scale for the intensity of emotion, ranging from 1 = no emotion to 5 = highly intense emotion. The debriefing data were used (i) to take into account only the stimuli that triggered recollections of detailed spatially and temporally specific personal episodes during the fMRI session and (ii) to compare the intense emotion ratings with low or absent emotional evocation.#
Image acquisition
Whole-brain imaging was performed on a 2 T S200 (Bruker, Karlsruhe, Germany) whole-body MRI system. Functional images were acquired in the axial plane with echo planar imaging (EPI) pulse sequence using BOLD contrast with the following parameters: TE = 40 ms, TR = 2500 ms, matrix size = 64 × 64, 32 slices per volume, slice thickness = 4 mm. Functional MRI was followed by a structural MRI session, where data were obtained using a three-dimensional T2-weighted sequence (TE = 73.8 ms, TR = 14,000 ms, matrix size = 128 × 128 × 48).#
fMRI data analysis
All fMRI data were processed and analysed using the SPM2 software (Wellcome Department of Imaging Neurosciences, London, UK; Friston et al., 1995) implemented in Matlab6 (The MathWorks, Inc.). Time-series was realigned to the first volume and then was spatially normalised to an EPI template based on the Montreal Neurological Institute (MNI) reference brain, which approximates Talairach and Tournoux's (1988) atlas space, and resampled to 2 × 2 × 2 mm. The normalised images were spatially smoothed with an isotropic 4-mm full width at half maximum (FWHM) Gaussian kernel. Data were high-pass-filtered (cut-off period 128 s) to remove low-frequency drifts. We did not apply global scaling following advice (fMRI analysis defaults for SPM2, http://www.mrc-cbu.cam.ac.uk/Imaging/Common/spm2_analysis_defaults.shtml). Statistical analysis was based on a random effects approach (Holmes and Friston, 1998). First, for each subject, the haemodynamic response to each stimulus event (face presentation plus motor response) was modelled using a canonical haemodynamic response function (Friston et al., 1998). We took into account only correct trials, i.e., ‘yes’ responses to the stimuli events that triggered recollections of autobiographical episodes and the correct identification of famous faces. Linear contrasts were constructed for each subject to compare the conditions (autobiographical retrieval vs. semantic retrieval). A random effects analysis was then performed in which the first level linear contrasts for each subject were subjected to a one-sample t test. A threshold of P < 0.001 uncorrected for multiple comparisons was employed. We retained only clusters (k) of at least 20 voxels in order to minimise the proportion of false positive activations.#
Additionally, ‘yes’ events of each condition were modelled by convolving onset times with a canonical haemodynamic response function and its derivatives to capture possible delayed responses (Simons et al., 2005).#
In order to ensure that the autobiographical memory cerebral pattern observed was not due to the reaction time differences between the experimental and the control condition (see behavioural results), we undertook an additional statistical analysis in which the autobiographical and the semantic tasks were matched as closely as possible for duration (duration matched data set). The matching was achieved by removing from the analysis autobiographical memory events characterised by the longest reaction times, while keeping however a minimum number of 15 memory events per subject.#
Acknowledgments
We are grateful to Prof. D Grucker, our project's principal medical researcher. We also thank Dr. J Foucher, who first hinted at a “family photo test”, Ms. C Marrer for technical assistance and Ms. N Heider for administrative assistance. Finally, we would like to acknowledge the participants' family members for their most helpful cooperation. This work was supported by Cognitique-ACI (fMRI project), the French Ministry of National Education and Research (ED's grant and LM's sabbatical year) and the medical research's French foundation (AB's grant).#
Figures and Tables
Table 1
| Subject | Mean score | High emotional recollections % | Low or non-emotional recollections % | Fairly % |
| 1 | 1 | 0 | 100 | 0 |
| 2 | 2.81 | 27.9 | 51.1 | 21 |
| 3 | 1.86 | 4.8 | 73.8 | 21.4 |
| 4 | 1.63 | 7.9 | 78.9 | 13.2 |
| 5 | 2.45 | 16.7 | 57.1 | 26.2 |
| 6 | 1.84 | 18.7 | 78.1 | 3.2 |
| 7 | 2.67 | 16.3 | 37.2 | 46.5 |
| 8 | 3.15 | 28.3 | 15.2 | 56.5 |
| 9 | 1.6 | 2.3 | 88.4 | 9.3 |
| 10 | 3.51 | 48.8 | 8.9 | 42.3 |
| Mean | 2.25 | 17.17 | 58.87 | 23.96 |
| SD | 0.79 | 14.90 | 30.72 | 19.03 |
Table 2
| Cluster | Z score | Coordinates (x, y, z) | Brain region (BA) |
| 2887 | 5.37 | 16 −54 8 | R posterior cingulate (BA 30, 31) |
| 4.82 | −6 −74 48 | L precuneus (BA 7) | |
| 1333 | 5.01 | −30 6 56 | L superior/middle frontal (BA 6,8) |
| 316 | 5.01 | 48 −72 32 | R temporo–parietal–occipital junction (BA 19, 39) |
| 746 | 4.95 | −34 −88 30 | L temporo–parietal–occipital junction (BA 19, 39) |
| 549 | 4.88 | 28 10 54 | R superior/middle frontal (BA 6) |
| 60 | 4.55 | −42 −50 38 | L temporo-parietal junction (BA 39, 40) |
| 308 | 4.51 | −42 28 28 | L middle frontal (BA 9, 46) |
| 99 | 4.49 | 20 −32 −20 | R parahippocampal (BA 35, 36) |
| 209 | 4.27 | 12 18 −16 | R medial frontal (BA 25) |
| 4.17 | 10 22 0 | R caudate (Head) | |
| 244 | 4.27 | −12 10 −6 | L putamen |
| 3.85 | −10 16 0 | L caudate (Head) | |
| 341 | 4.23 | −22 −40 −20 | L cerebellum |
| 4.13 | −12 −24 −16 | L midbrain | |
| 3.90 | −30 −34 −16 | L fusiform (BA 20) | |
| 192 | 4.23 | −4 44 4 | L medial frontal/anterior cingulate (BA 10, 32) |
| 3.79 | 4 38 6 | R anterior cingulate (BA 24) | |
| 144 | 4.16 | −30 48 12 | L superior/middle frontal (BA 10, 46) |
| 142 | 4.13 | −34 14 −14 | L inferior frontal (BA 47) |
| 31 | 3.80 | −42 8 32 | L middle frontal (BA 9) |
| 66 | 3.71 | 8 10 54 | R superior frontal (BA 6) |
| 3.52 | 12 18 44 | R cingulate (BA 32) | |
| 3.41 | 14 28 44 | R medial frontal (BA 8) | |
| 47 | 3.61 | 4 −28 −14 | R midbrain |
| 3.59 | 6 −26 −2 | R thalamus | |
| Listed are cluster size, neuroanatomical regions with associated Brodmann areas (BA), Talairach's coordinates (x, y, z) and Z score; L, left; R, right. |
References
3. A.BartelsS.ZekiThe neural basis of romantic loveNeuroReport11200038293834
4. A.BartelsS.ZekiThe neural correlates of maternal and romantic loveNeuroImage21200411551166
10. H.E.den OudenU.FrithC.D.FrithS.J.BlakemoreThinking about intentionsNeuroImage162005xxx
18. D.L.GreenbergD.C.RubinThe neuropsychology of autobiographical memoryCortex392003687728
22. A.P.HolmesK.J.FristonGeneralisability, random effects and population inferenceNeuroImage71998S754
35. G.NorthoffF.BermpohlCortical midline structures and the selfTrends Cogn. Sci.82004102107
45. E.TulvingEpisodic memory, from mind to brainAnnu. Rev. Psychol.532002125