Delineating self-referential processing from episodic memory retrieval: Common and dissociable networks
Introduction
Recently, self-referential processing (SRP) has received increasing interest in neuroimaging studies (Gillihan and Farah, 2005, Legrand and Ruby, 2009, Northoff and Bermpohl, 2004, Northoff et al., 2006). The heterogeneous nature of SRP, entailing a complex set of operations, poses significant challenges in identifying its specific neural correlates. Consistently, self-referential (relative to control) tasks induce increases in BOLD signals (hereafter, “activations”) in the ventral (VMPFC) and dorsal medial prefrontal cortex (DMPFC) (extending into the anterior cingulate cortex (ACC)) as well as the medial and lateral parietal cortex (extending into temporal areas). Because this self-network is activated across different sensory modalities and cognitive domains (e.g., spatial, facial, emotional, social) (Northoff et al., 2006), it could be assumed that distinct subregions within this network may correspond to specific processes involved in SRP. Investigators have recently started to disentangle these processes. For instance, they have studied SRP in relation to emotion processing (Moran et al., 2006, Northoff et al., 2009, Phan et al., 2004), theory of mind (Vogeley et al., 2001), inferential processing (Legrand and Ruby, 2009), reward processing (de Greck et al., 2008), realness (Summerfield et al., 2009) and sexual arousal (Heinzel et al., 2006).
Here we examine SRP in relation to episodic memory retrieval (EMR). There appears to be a theoretical consensus that SRP generally involves EMR; according to some authors, SRP and EMR are even intrinsically related (Conway and Pleydell-Pearce, 2000, Gardiner, 2001, James, 1892). Similar to EMR, SRP depends on the individual's life history and involves the recollection of past experiences. On the other hand, EMR seems to implicate reference to the self, as the retrieved episodic information is unique to an individual and is tied to a specific personal context (Craik et al., 1999, Ingvar, 1985, Tulving, 1983). Behaviorally, the link between SRP and EMR is reflected in the so-called self-reference effect of memory: encoding with reference to self yields superior memory performance relative to semantic or other-referential encoding (Rogers et al., 1977, Symons and Johnson, 1997). Further support for this link comes from neuroimaging investigations: EMR studies report activations in brain regions that are also identified by SRP tasks, again including the anterior and medial prefrontal cortex, as well as the medial and lateral parietal cortex (Donaldson et al., 2001, Fletcher et al., 1996, Henson et al., 2005, Konishi et al., 2000, McDermott et al., 2000) (for reviews see Cavanna and Trimble, 2006, Legrand and Ruby, 2009). Because these brain areas also show high neural activity during so-called rest conditions (i.e., conditions without externally focused tasks), both SRP and EMR have been considered components of the brain's default-mode network (Buckner et al., 2008).
Despite these commonalities, several lines of evidence indicate that SRP and EMR can be well distinguished: First, it is phenomenologically evident that besides retrieval, SRP requires reference to one's own person, i.e., reference to the person's self-concept (concerning physical and psychological traits), value system, motives, and internal goals among others (Zysset et al., 2002). Episodic memory processes, on the other hand, also concern the retrieval of events that are characterized by low self-relevance. Second, a series of priming experiments using trait adjectives as probes showed no impact of self-description (i.e., a form of SRP) on EMR and vice versa, indicating that one process does not automatically invoke the other (Kihlstrom and Klein, 1997). Third, some case studies in patients with retrograde amnesia following traumatic brain injury report a loss of EMR in combination with preserved self-description (Klein et al., 1996, Tulving, 1993), suggesting that both processes may, at least partially, be represented independently. Fourth, functional connectivity analyses of fMRI (functional magnetic resonance imaging) data suggest functional separation within the default-mode network, revealing two distinct subsystems, namely a medial temporal lobe subsystem associated with EMR and a medial prefrontal subsystem associated with SRP; both systems seem to converge on the medial and lateral parietal cortex (Buckner et al., 2008). Finally, functional neuroimaging studies suggest that functional specialization may exist within the medial parietal cortex: Cavanna and Trimble (2006) proposed a dissociation within the precuneus into an anterior region, involved in SRP and a posterior region, subserving EMR. Fig. 1 illustrates this dissociation based on studies reviewed by Cavanna and Trimble (2006).
Taken together, there is evidence for both functional overlap and dissociation between SRP and EMR. Because evidence provided so far is either indirect or relies on data from different imaging studies, the exact extent to which SRP and EMR depend on common and distinct brain regions remains unclear. The aim of the present study is to address this issue by combining standard tasks for both processes in one functional MRI (fMRI) experiment. Healthy volunteers were asked to perform both a self-referential and an episodic memory task in relation to each of 160 pictorial stimuli taken from the International Affective Picture System (IAPS) (Lang et al., 2005). The task demand thus remained the same, whereas self-relatedness and EMR varied across pictures. This allowed us to avoid confounds introduced by differences in task-related cues and task demands. For fMRI data analysis, picture trials were classified based on the participants' responses in the self-referential and episodic memory tasks, resulting in a 2 × 2 factorial design with the factors self-relatedness (self-referential, non-self-referential) and EMR (retrieved, non-retrieved). This allowed us to identify (1) the effect of SRP (self-referential > non-self-referential stimuli), (2) the effect of EMR (retrieved > non-retrieved stimuli), (3) common activations for SRP and EMR (conjunction of the two former contrasts), (4) SRP × EMR interaction effects, (5) activations specific for SRP relative to EMR (self-referential/non-retrieved > non-self-referential/retrieved), and (6) activations specific for EMR relative to SRP (retrieved/non-self-referential > non-retrieved/self-referential).
Based on the studies reported above, we hypothesized that both processes, SRP and EMR, would recruit a common neural network including (1) the medial parietal cortex, (2) the anterior and medial prefrontal cortex, and (3) the lateral parietal cortex. Besides overlapping activations, we also predicted an anterior–posterior gradient within the medial parietal cortex with SRP stronger activating the anterior division and EMR stronger activating the posterior division.
Section snippets
Subjects
Twenty-nine right-handed healthy volunteers (14 men, 15 women, aged 30–50 years, mean ± standard deviation (SD) = 39 ± 5.5 years, mean IQ = 118 ± 13.5 measured with the Mehrfachwahl-Wortschatz-Intelligenztest (MWT-B) (Lehrl, 2005)), gave written informed consent to participate in the experiment. Exclusion criteria were left-handedness, current limiting general medical conditions, current neurological disorder and history of psychiatric axis I or II disorders in the subjects (assessed with SCID I and II
Trial distribution and familiarity task performance
Pictures presented during scanning were classified based on the participants' responses in the SRP and EMR tasks as well as the oldness of the pictures (Fig. 2). Analysis of the resulting number of trials per condition yielded a main effect for the factors SRP [F(1,28) = 11.977, p = 0.002] and familiarity [F(1,28) = 13.016, p = 0.001] with more pictures judged as non-self-referential and familiar, respectively [mean number of trials ± SD rated as self-referential: 67 ± 4; non-self-referential: 93 ± 20;
Discussion
The present data suggest common and dissociable networks for SRP and EMR. Three main findings concern (1) the medial parietal cortex (PCC, precuneus), (2) the prefrontal cortex, and (3) the lateral parietal cortex (Fig. 7). More specifically, self-referential stimuli specifically activate the PCC/anterior precuneus, the ventral and dorsal medial prefrontal cortex (extending into the ACC), and an inferior division of the inferior parietal lobule extending into the superior and middle temporal
Conflict of interest
The authors of the study have no conflicts of interest to declare.
Acknowledgments
We would like to thank Matt Walker for helpful discussion of the paradigm of the study.
This work was supported by grants from the German Federal Ministry of Education (BMBF-01GWSO61 to F.B., BMBF-01GV0612 to A.S., M.S., and A.W., BMBF-01GS08148 to A.H., BMBF-01GS08159 to A.H.); German Research Foundation (DFG-SFB 779-A6 to G.N.); Canada Research Chair (CRC to G.N.); and EJLB-Michael Smith Foundation (to G.N.).
References (77)
- et al.
Self-projection and the brain
Trends Cogn. Sci.
(2007) - et al.
Functional connectivity with anterior cingulate and orbitofrontal cortices during decision-making
Brain Res. Cogn. Brain Res.
(2005) - et al.
Amygdala tractography predicts functional connectivity and learning during feedback-guided decision-making
NeuroImage
(2008) - et al.
Memory orientation and success: separable neurocognitive components underlying episodic recognition
Neuropsychologia
(2003) - et al.
Dissociating state and item components of recognition memory using fMRI
NeuroImage
(2001) - et al.
The mind's eye—precuneus activation in memory-related imagery
NeuroImage
(1995) - et al.
Psychophysiological and modulatory interactions in neuroimaging
NeuroImage
(1997) - et al.
Conjunction revisited
NeuroImage
(2005) Autobiographical and episodic memory—one and the same? Evidence from prefrontal activation in neuroimaging studies
Neuropsychologia
(2004)- et al.
Is our self based on reward? Self-relatedness recruits neural activity in the reward system
NeuroImage
(2008)