You are on page 1of 12

Neuropsychologia 80 (2016) 3546

Contents lists available at ScienceDirect

Neuropsychologia
journal homepage: www.elsevier.com/locate/neuropsychologia

Default network activation during episodic and semantic memory


retrieval: A selective meta-analytic comparison
Hongkeun Kim
Department of Rehabilitation Psychology, Daegu University, Gyeongsangbuk-Do 38453, South Korea

art ic l e i nf o a b s t r a c t

Article history: It remains unclear whether and to what extent the default network subregions involved in episodic
Received 3 August 2015 memory (EM) and semantic memory (SM) processes overlap or are separated from one another. This
Received in revised form study addresses this issue through a controlled meta-analysis of functional neuroimaging studies in-
7 November 2015
volving healthy participants. Various EM and SM task paradigms differ widely in the extent of default
Accepted 9 November 2015
Available online 10 November 2015
network involvement. Therefore, the issue at hand cannot be properly addressed without some control
for this factor. In this regard, this study employs a two-stage analysis: a preliminary meta-analysis to
Keywords: select EM and SM task paradigms that recruit relatively extensive default network regions and a main
Fmri analysis to compare the selected task paradigms. Based on a within-EM comparison, the default network
Default mode network
contributed more to recollection/familiarity effects than to old/new effects, and based on a within-SM
Episodic memory
comparison, it contributed more to word/pseudoword effects than to semantic/phonological effects.
Semantic memory
Meta-analysis According to a direct comparison of recollection/familiarity and word/pseudoword effects, each involving
a range of default network regions, there were more overlaps than separations in default network
subregions involved in these two effects. More specically, overlaps included the bilateral posterior
cingulate/retrosplenial cortex, left inferior parietal lobule, and left anteromedial prefrontal regions,
whereas separations included only the hippocampal formation and the parahippocampal cortex region,
which was unique to recollection/familiarity effects. These results indicate that EM and SM retrieval
processes involving strong memory signals recruit extensive and largely overlapping default network
regions and differ mainly in distinct contributions of hippocampus and parahippocampal regions to EM
retrieval.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction reportedly specic to EM or SM retrieval.


Such mixed results likely reect the fact that different types of
1.1. The purpose of the study EM tasks and related activation contrasts recruit different sets of
brain regions and that this also holds for different types of SM
The distinction between episodic memory (EM) and semantic tasks and associated contrasts. For example, in both EM and SM,
memory (SM) has sparked a massive number of experimental and some tasks involve a relatively high demand for controlled pro-
lesion studies (Tulving, 1972). Here a critical issue concerns whe- cessing, whereas others, a low demand. If EM and SM retrieval
ther and to what extent EM and SM processes involve similar or tasks selected for a comparison entail a high demand for con-
differential neural substrates. Although this issue has been in- trolled processing, common regions for the two tasks may include
vestigated by many functional neuroimaging studies (e.g., Bur- fronto-parietal control regions such as the lateral prefrontal cortex
ianova and Grady, 2007; Dzel et al., 1999; Maguire and Mum- (PFC). However, if both tasks involve a relatively low demand for
mery, 1999; Nyberg et al., 2003; Rajah andMcIntosh, 2005; Sha- controlled processing, then the PFC and other regions are less
pira-Lichter et al., 2013; Wiggs et al., 1998), it remains con- likely to show some common activity. In addition, the PFC and
troversial because of highly mixed results across studies. Previous other regions may emerge as specic to either EM or SM retrieval
studies have indicated the common involvement of various frontal,
if the demand for controlled processing is high in one task but low
temporal, parietal, occipital, and subcortical subregions and de-
in the other. Although many studies have addressed these and
monstrated the existence of similar diversity for regions
other potentially biasing factors, a fully adequate control scheme
has not always been practical because of the complexity of each
E-mail address: hongkn1@gmail.com task's cognitive processes. Nonetheless, a meaningful and

http://dx.doi.org/10.1016/j.neuropsychologia.2015.11.006
0028-3932/& 2015 Elsevier Ltd. All rights reserved.
36 H. Kim / Neuropsychologia 80 (2016) 3546

theoretically important comparison between EM and SM pro- demonstrated in recent meta-analyses (Kim, 2013; McDermott
cesses requires the careful selection of task paradigms and the et al., 2009; Spaniol et al., 2009). Recognition memory may be
discrimination of commonalities/differences from intrinsic pro- based on either vivid recall of an item and associated contexts
cesses and those from extrinsic factors. (recollection) or a feeling of oldness in the absence of any con-
Recent studies have indicated that both EM and SM retrieval textual information (familiarity). The contrast of recollection with
processes are related to a large-scale, intrinsic network often re- familiarity has been widely used to probe the EM retrieval process
ferred to as the default network (Raichle et al., 2001), as reviewed involving autonoetic consciousness and other related properties
in more detail in the next section. However, it remains largely (Gardiner, 2001). A meta-analysis of this recollection/familiarity
unclear whether and to what extent the default network sub- contrast indicated a more consistent relationship with default
regions involved in EM and SM processes overlap or are separated network regions (Kim, 2010), providing clearer evidence linking
from one another. To address this issue, this study provides a se- laboratory-based EM retrieval to the default network. However, no
lective and controlled meta-analysis of relevant functional neu- meta-analysis has directly compared recollection/familiarity and
roimaging studies involving healthy participants. old/new effects in the extent of default network involvement.
More generally, many laboratory-based EM task types activate at
1.2. The default network and memory retrieval least some default network regions, but they also show relatively
large differences in the scope of default network involvement. In
Recent advances in the eld can provide a useful framework for this regard, there is a need to determine the extent to which
a comparison between EM and SM processes and suggest that various EM retrieval paradigms differ in default network involve-
both involve the default network. Major components of the default ment and why.
network include the anteromedial PFC extending laterally to the Binder et al.'s (2009) meta-analysis has recently played a key
dorsal and ventral PFC, the posterior cingulate cortex extending role in disseminating the notion that SM retrieval is associated
ventrally to the retrosplenial cortex, the inferior parietal lobule, with the default network. Binder et al. restricted their meta-ana-
the lateral temporal cortex, and the medial temporal lobe, in- lysis to studies using verbal stimuli, arguing that much of human
cluding the hippocampal formation (see pink regions in Fig. 1). knowledge is represented symbolically through language and in-
Functional roles of the default network are not fully understood, dicating their desire to maintain a clear focus on the processing of
but there is strong evidence that it plays a critical role in any in- concepts rather than percepts. Their analysis considered about 30
ternally oriented mental activity such as recall, future thinking, types of SM task paradigms simultaneously, many of which were
social cognition, stimulus-independent thought, and other spon- used in only one or two studies. The most common paradigm in
taneous cognition (AndrewsHanna et al., 2014). their analysis, referred to here as the word/pseudoword paradigm,
Evidence that EM retrieval is associated with the default net- involves the presentation of words and pseudowords in target and
work comes from studies using the autobiographical memory baseline conditions, respectively, and lexical decisions. The second
paradigm (Buckner and Carroll, 2007; Kim, 2012). Until recently, most common paradigm is the semantic/phonological paradigm,
however, similar evidence has been lacking in studies using la- which involves semantic and phonological processing tasks for
boratory-based EM tasks. For example, one of the most proto- words in target and baseline conditions, respectively. Their meta-
typical task paradigms in using neuroimaging in the analysis of analysis involving these and other paradigms indicated a clear
laboratory-based EM retrieval is to compare the correct recogni- overlap with default network regions. However, some caution is
tion of studied items (hit) with the correct non-recognition of necessary because a consistent relationship with default network
novel items (correct rejection). This old/new (i.e., old 4new) effect regions may be the property of not all but only some paradigms
shows only a moderate overlap with default network regions, as included in the meta-analysis. For example, some studies that

dPFC IPS PCC/RSC


pmPFC
midPFC IPL dPCU

vPFC amPFC
aINS pMTG MTL
LTC

Default network Fronto-parietal control network


Fig. 1. Estimates of the default-mode network (shown in pink) and the fronto-parietal cognitive control network (green). Estimates were produced using data from Yeo et al.
(2011), which can be downloaded from the Surface Management System Database (SumsDB; http://sumsdb.wustl.edu/sums). Data were from a clustering analysis of resting-
state functional MRI signals involving 1175 uniformly spaced cortical regions and 1000 subjects. aINS, anterior insula; amPFC, anteromedial prefrontal cortex; dPCU, dorsal
precuneus; dPFC, dorsal prefrontal cortex; IPL, inferior parietal lobule; IPS, intraparietal sulcus; LTC, lateral temporal cortex; midPFC, middle prefrontal cortex; MTL, medial
temporal lobe; PCC, posterior cingulate cortex; pmPFC, posteromedial prefrontal cortex; pMTG, posterior middle temporal gyrus; RSC, retrosplenial cortex; vPFC, ventral
prefrontal cortex. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)
H. Kim / Neuropsychologia 80 (2016) 3546 37

contrasted a semantic processing task with a phonological pro- network involvement. According to this hypothesis, EM and SM
cessing task have indicated only modest default network in- task paradigms involving only modest default network contribu-
volvement (e.g., Emmorey et al., 2013; Price et al., 1997). Therefore, tions may be associated with relatively weak memory signals in
as in EM retrieval, there is a need to determine the extent to which target conditions and/or relatively strong memory signals in re-
various SM retrieval paradigms differ in default network involve- ference conditions. Some evidence providing support for this hy-
ment and why. pothesis is presented later. For now, such EM and SM task para-
Conscious memory retrieval, both episodic and semantic, is digms are less useful for addressing the question at hand in this
intrinsically associated with attention to internal representations. study (i.e., whether and to what extent the default network sub-
Therefore, the involvement of the default network in EM and SM regions involved in EM and SM processes overlap or are separated
retrieval is consistent with the notion that the default network from one another).
supports internally oriented mentation. Here a related issue of Various EM and SM task paradigms differ widely in the extent
great importance concerns whether and to what extent these two of default network involvement. Therefore, the issue at hand
processes involve similar or differential default network sub- cannot be properly addressed without some control for this factor.
regions. Although several studies have compared neural substrates A better comparison would require that both EM and the SM task
of these processes, these questions remain largely underexplored paradigms selected for a comparison involve default network re-
because most studies have been designed without a clear con- gions to a relatively large extent. If this requirement is met in only
ceptualization of the default network and its role in EM and SM one task, then the comparison is likely to be biased (i.e., yield
retrieval. spurious differences), and if it is not met in either condition, then
EM and SM processes differ in many aspects, including their the comparison cannot identify the motivating factor. To address
phenomenological experience, self-references versus cognitive this issue, the present meta-analysis was conducted in two stages.
references, and simultaneous retrieval and integration of what, The rst (preliminary) stage involved two within-memory com-
where, and when information unique to EM. However, both parisons: a comparison of two EM task paradigms (recollection/
EM and SM represent consciously retrieved long-term cognitive familiarity and old/new) to determine which one would engage
memory (i.e., declarative memory). In addition, their operations the default network in a more consistent manner and a compar-
are highly intertwined with each other (Greenberg and Verfaellie, ison of two SM task paradigms (word/pseudoword and semantic/
2010; Irish and Piguet, 2013; Klein, 2013), suggesting some broadly phonological) for a similar objective. Although the two task
shared neural substrates. In encoding, one's new SM forms from paradigms in each subject area are not exhaustive of all reported
repeated episodic experiences, and the semantic processing of in the literature, they do represent those that have been most
some to-be-remembered material facilitates the storage of new widely used and thus are considered highly relevant in that area.
EM. In retrieval, EM retrieval necessarily includes semantic re- The second (main) stage involved a comparison between the EM
presentations in that recalling I visited a museum yesterday task paradigm showing the more consistent involvement of the
necessitates the retrieval of the concepts visit, museum, and default network in the rst-stage comparison and the SM task
yesterday (Binder et al., 2009), and SM retrieval can be aug- paradigm showing the same. Fortunately, as discussed in the Re-
mented by calling on relevant episodic information (Ryan et al., sults section, EM and SM task paradigms were relatively equal in
2008). Many memory types are collectively called personal se- the extent of default network involvement, rendering their com-
mantics (Renoult et al., 2012) and have some features of both EM parison balanced in that regard.
and SM, indicating that the two types of memory have less clear
boundaries than initially conceived. For example, memory for re-
2. Materials and methods
peated personal events (e.g., making morning coffee in the kitch-
en), or repisodic memory (Neisser, 1981), does not fully qualify
2.1. Data collection
as episodic but frequently includes specic spatial and contextual
information. Therefore, a general prediction is that the default
The meta-analysis focused on the following four task paradigms: recollection/
network subregions involved in the two memory processes show familiarity, old/new, word/pseudoword, and semantic/phonological. Functional
no complete overlap or separation from one another but a mixture neuroimaging studies involving these task paradigms were identied through
of overlaps and separations. multiple searches of the Pubmed and Google Scholar databases. The search terms
included fMRI, PET, episodic memory, recollection, recognition memory,
semantic memory, lexical decision, pseudowords, and semantic processing.
1.3. A controlled meta-analytic comparison
Reference lists of recent meta-analyses (Binder et al., 2009; Kim, 2013, 2015) were
also considered. Then search results were ltered to include only those studies that
Many within-study comparisons of EM and SM process studies (1) included healthy participants, (2) performed whole-brain analyses, and (3) re-
have held items constant in the two conditions and required ported activation foci in the standard reference space (i.e., Talairach or Montreal
Neurological Institute [MNI] coordinates). Ultimately, the meta-analysis involved a
subjects to process the same items either episodically (e.g., old/
total of 105 studies (articles) collectively involving 1792 participants (see Sup-
new judgments) or semantically (e.g., living/nonliving judgments) porting information References for a list of studies included in the meta-analysis).
(Barba et al., 1998; Burianova and Grady, 2007; Dzel et al., 1999; The imaging modality was functional magnetic resonance imaging in 99 studies
Hantke et al., 2013; St.-Laurent et al., 2011; Wiggs et al., 1998). and positron emission tomography in the rest. The sensory modality for the sti-
Although ndings vary widely across studies, EM and/or SM tasks mulus presentation was visual in 99 studies and auditory in the rest.
have generally been found to involve default network regions to
some limited extent, and therefore default network regions rarely 2.2. Task paradigms

gure prominently in either common or specic effects. Although


Table 1 lists the numbers of experiments, peak foci, and participants for each of
these ndings may seem puzzling because of the presumed pivotal
the four task paradigms (see Supporting information Tables S1S4 for a detailed
role of the default network in mnemonic processing, they em- description of studies included in each category). Although most studies have in-
phasize the notion that not all conventional memory tasks and cluded only one relevant experiment of the same task paradigm, some reported
associated activation contrasts consistently tap cognitive processes more. In the case of the latter, if multiple experiments involved entirely separate
supported by the default network. Here the proposed hypothesis is sets of trials, all were included. Otherwise, only one was selected for inclusion to
prevent these studies from overly inuencing results. The experiment with the
that not all but only task paradigms that involve both strong largest number of associated peak coordinates was selected because this approach
memory signals in target conditions and weak (or no) memory was considered to likely enhance the sensitivity of the meta-analysis.
signals in reference conditions entail a high level of default Recollection/familiarity contrasts involved a comparison of recollection
38 H. Kim / Neuropsychologia 80 (2016) 3546

Table 1 uncertainly being less (for further details, see Eickhoff et al. (2009)). The modeled
A summary of four different task paradigms in the meta-analysis. probability distributions across all reported foci were combined within a given
study and then across all studies, which ultimately produced voxelwise ALE values
Memory type Task paradigms Experiments Foci Participants that estimated the activation likelihood of each voxel across studies. The statistical
signicance of ALE scores was determined in reference to the analytically derived
EM retrieval Recollection 4familiarity 27 400 529 null distribution of the ALE statistic. A cluster-level familywise error-corrected
Old 4new 42 571 746 threshold was set at p o .05, and a cluster-forming threshold was set at the voxel
level at p o.005.
SM retrieval Word4pseudoword 20 211 338 To apply the ALE algorithm to a subtraction analysis of two ALE maps (Eickhoff
Semantic 4phonological 17 120 179 et al., 2011), all studies contributing to either ALE map were pooled, the combined
group was randomly split into two sets of studies having the same size as the
EM, episodic memory; SM, semantic memory. original two groups, and the difference in ALE scores between the two sets of
studies was calculated. This process was repeated 10,000 times to estimate the null
distribution of a chance difference between the two ALE maps. This estimated
responses with familiarity responses in a recognition memory test. There were 27 chance distribution was used to determine the statistical signicance of observed
experiments of this type. In 17 of these experiments, the stimulus type of retrieval differences in ALE scores between the two ALE maps. The threshold was set to a
cue was verbal (words), and in 10, it was pictorial (e.g., object pictures and face voxel level at p o .005 and inclusively masked by respective main effects with a
photos). The response method featured a standard remember/know procedure cluster-level familywise error-corrected threshold at p o .05. A conjunction analysis
(Gardiner, 2001) in 13 experiments, condence ratings in 7, and a combination of of the two ALE maps was conducted based on the intersection between the two
remember/know and condence ratings in 7. Condence-based experiments were maps, each with a cluster-level familywise error-corrected threshold set at p o.05.
included because, (1) although there is no one-to-one direct correspondence, on The possibility of false-positive ndings in subtraction and conjunction analyses
average, high- and low-condence recognition align more with remember and was further constrained by applying a spatial extent threshold exceeding 500 mm3.
know recognition, respectively (Dunn, 2004) and (2) studies have shown that Results of the meta-analysis were visualized through the projection of thre-
experiments based on condence ratings yield activation patterns largely com- sholded ALE maps onto an inated population-average landmark surface (PALS) by
parable to those based on the remember/know procedure (Daselaar et al., 2006; using CARET (Van Essen, 2005). Yeo et al. (2011) recently parcellated the cerebral
Kim and Cabeza, 2009). cortex region into seven intrinsic networks based on a clustering analysis of rest-
Old/new contrasts involved a comparison of hit responses with correct rejec- ing-state BOLD signals involving 1175 uniformly spaced cortex regions and 1000
tion responses in a recognition memory test. There were 42 experiments of this subjects (see Fig. 1). Estimated boundaries of the default network in this study were
type. In 27 of these experiments, the stimulus type of retrieval cue was verbal, and projected onto the PALS as exible guidelines for evaluating whether convergence
in 15, it was pictorial. clusters were located within or outside the default network. When appropriate,
Word/pseudoword contrasts involved the presentation of words and pseudo- estimated boundaries of the fronto-parietal control network were also projected
words in target and baseline conditions, respectively, and lexical decision tasks. onto the PALS (see green regions in Fig. 1). For the visualization of results for the
There were 20 experiments of this type. A word/pseudoword experiment was in- hippocampal formation, ALE maps were overlaid onto the International Consortium
cluded only when it appropriately controlled for orthographic and phonological for Brain Mapping template by using Mango (http://ric.uthscsa.edu/mango).
processes. Therefore, the pseudoword condition always involved the presentation
of orthographically legal and pronounceable letter strings. Although the suitability
of a lexical decision task in this study may be questioned because it required a
3. Results
relatively shallow level of semantic processing, it was included because (1) the
task has been widely used in previous studies and (2) there is behavioral evidence
that the task involves access to the word meaning instead of or in addition to 3.1. EM retrieval
presemantic word codes (Balota et al., 2004).
In this study, the semantic/phonological contrasts involved semantic and Table 2 and Fig. 2 show the results of separate ALE meta-ana-
phonological processing tasks for words in target and baseline conditions, re-
spectively. The semantic processing tasks involved a two-choice response to
lyses of recollection/familiarity and old/new effects and a direct
questions such as Does the word represent a living or nonliving object? and comparison of the two effects. Both effects were predominantly
Does the word represent a concrete entity or not? and the phonological pro- lateralized to the left. First, recollection/familiarity effects over-
cessing tasks had questions such as Does the word have two syllables or not? and lapped mainly and extensively with default network regions, in-
Does the word have the target phoneme?
cluding the bilateral posterior cingulate/retrosplenial cortex, the
bilateral inferior parietal lobule, the bilateral anteromedial PFC,
2.3. Data analysis
the bilateral medial temporal lobe including the hippocampal
formation, the left ventral PFC, and the bilateral lateral temporal
Two types of meta-analyses were conducted. The rst type was a single-study
meta-analysis which separately evaluated each of the four task paradigms. The cortex, in approximate order of decreasing spatial extent (see
second type compared pairs of task paradigms. More specically, the evaluation of Fig. 2A).
differential effects involved a subtraction meta-analysis, and the evaluation of Second, old/new effects showed a more moderate overlap with
common effects involved a conjunction meta-analysis. This type of pairwise ana-
default network regions, including the bilateral inferior parietal
lysis proceeded in two stages. The rst stage involved a comparison of two EM task
paradigms (recollection/familiarity and old/new) and a comparison of two SM task
lobule, the bilateral posterior cingulate cortex, the left lateral
paradigms (word/pseudoword and semantic/phonological). The second stage in- temporal cortex, and the left parahippocampal cortex, in approx-
volved a comparison of the EM task paradigm that indicated stronger default imate order of decreasing spatial extent (Fig. 2B). However, they
network engagement in the rst-stage comparison and the SM type contrast that overlapped extensively with the fronto-parietal control network,
indicated the same. The rationale for this two-step approach is explained in the
including the bilateral middle PFC, the bilateral intraparietal sul-
Introduction section and thus is not detailed here.
cus, the bilateral anterior insula, the bilateral dorsal precuneus,
2.4. Meta-analysis techniques and the left posteromedial PFC, in approximate order of decreasing
spatial extent.
All meta-analyses were conducted using the activation likelihood estimation Finally, the regions that were associated more with recollec-
(ALE) algorithm implemented in GingerALE 2.3.3 (http://www.brainmap.org). The tion/familiarity effects than with old/new effects involved mainly
objective of the ALE algorithm is to determine brain regions showing above-chance default network regions, including the bilateral medial temporal
activation convergence across a set of independent studies (Turkeltaub et al., 2012).
lobe, the bilateral anteromedial PFC, the bilateral inferior parietal
To apply the algorithm to the present data set, the standard reference space was
determined for each study, and all activation foci reported in the MNI space were lobule, and the left retrosplenial cortex (yellow regions in Fig. 2C).
converted into Talairach coordinates by using icbm2tal (Lancaster et al., 2007). Each By contrast, reverse effects the regions that were associated
activation focus was then modeled as the center of the 3D Gaussian probability more with old/new effects than with recollection/familiarity ef-
distribution reecting the spatial uncertainty of each focus. The size of the modeled fects involved mainly fronto-parietal control regions, including
Gaussian was calculated based on an extended ALE algorithm accounting for be-
tween-subject variability and between-template/institution variability introduced
the left middle PFC, the bilateral intraparietal sulcus, the bilateral
by different spatial normalization methods. The algorithm included a step assign- dorsal precuneus, the left anterior insula, and the left poster-
ing a tighter, taller Gaussian to larger samples to reect the notion of spatial omedial PFC (cyan regions). The regions common to both task
H. Kim / Neuropsychologia 80 (2016) 3546 39

Table 2
Results of separate ALE meta-analyses of recollection/familiarity and old/new effects and a direct comparison of two effects.

Network Volume (mm3) ALE Z Talairach H Region Brodmann area

x y z

Recollection 4familiarity
DN 8200 .032 8 56 18 B Posterior cingulate cortex, RSC 23, 30, 31
6360 .033 40 70 28 L Inferior parietal lobule 39
5768 .026 6 50 2 B Anteromedial PFC 9, 10, 24, 32
4192 .027 26 20 6 R Hippocampus, PHC 35, 36
3840 .031 28 26 12 L Hippocampus, PHC 35, 36
1816 .019 28 34 4 L Ventral PFC 47
872 .014 46 66 8 R Inferior parietal lobule 39
672 .018 60 28 12 L Lateral temporal cortex 21
552 .016 54 4 14 R Lateral temporal cortex 21

Other 1536 .030 18 2 10 L Lentiform nucleus


1008 .024 4 8 2 R Caudate nucleus
744 .022 52 60 4 L Posterior temporal cortex 37

Old 4new
DN/FPCN 13,600 .059 34 66 40 L Inferior parietal lobule, intraparietal sulcus 7, 19, 39, 40
9664 .056 8 70 28 B Posterior cingulate cortex, dorsal precuneus 7, 31
5472 .038 32 68 36 R Inferior parietal lobule, intraparietal sulcus 7, 19, 39

DN 5504 .046 4 38 32 B Posterior cingulate cortex 23, 31


1328 .028 60 40 8 L Lateral temporal cortex 21
832 .021 16 30 10 L PHC 35

FPCN 14,320 .040 36 46 10 L Middle PFC 6, 9, 10, 45, 46


4032 .035 6 26 42 L Posteromedial PFC 8, 32
3288 .049 30 18 0 L Anterior insula 13
1696 .030 30 20 4 R Anterior insula 13
768 .021 28 52 12 R Middle PFC 10

Other 8824 .049 10 8 6 B Caudate nucleus, lentiform nucleus


640 .022 14 74 10 L Lingual gyrus 18

Recollection/familiarity 4 old/new
DN 3328 3.89 26 20 10 R Hippocampus, PHC 35, 36
2912 3.89 29 30 14 L Hippocampus, PHC 35, 36
1920 3.89 1 56 5 B Anteromedial PFC 10, 32
1512 3.89 47 73 15 L Inferior parietal lobule 39
896 3.89 7 55 14 L RSC 23
624 3.43 46 68 6 R Inferior parietal lobule 39

Other 1200 3.89 20 9 12 L Lentiform nucleus


736 3.89 50 62 1 L Posterior temporal cortex 37

Old/new 4recollection/familiarity
FPCN 5680 3.89 2 72 38 B Dorsal precuneus 7
5128 3.89 33 63 44 L Intraparietal sulcus 7
3992 3.72 40 53 9 L Middle PFC 10, 46
3296 3.89 26 73 42 R Intraparietal sulcus 7, 19
2336 3.72 40 10 34 L Middle PFC 9
1664 3.72 36 15 8 L Anterior insula 13
1240 3.24 3 25 38 L Posteromedial PFC 32

Other 2064 3.89 9 0 11 L Caudate nucleus

Recollection/familiarity old/new
DN 2976 .027 8 46 30 B Posterior cingulate cortex 23, 31
2184 .031 40 70 30 L Inferior parietal lobule 39

ALE, activation likelihood estimation; B, bilateral; DN, default network; FPCN, fronto-parietal control network; H, hemisphere; L, left; PFC, prefrontal cortex; PHC, para-
hippocampal cortex; R, right; RSC, retrosplenial cortex.

paradigms involved only default network regions, including the predominantly lateralized to the left. First, word/pseudoword ef-
bilateral posterior cingulate cortex (more to the left) and the left fects overlapped exclusively and extensively with default network
inferior parietal lobule (Fig. 2D). regions, including the bilateral inferior parietal lobule, the bilateral
posterior cingulate/retrosplenial cortex, the left dorsal PFC, the
3.2. SM retrieval bilateral anteromedial PFC, and the left lateral temporal cortex, in
approximate order of decreasing spatial extent (see Fig. 3A).
Table 3 and Fig. 3 show the results of separate ALE meta-ana- Second, semantic/phonological effects overlapped mainly with
lyses of word/pseudoword and semantic/phonological effects and default network regions, but the scope of this overlap was modest
a direct comparison of the two effects. Both effects were relative to that of word/pseudoword effects, including the left
40 H. Kim / Neuropsychologia 80 (2016) 3546

Recollection > Familiarity Subtraction

y=-20 y=-25 y=-20 y=-25

ALE .012 .033 Rec/Fam > Old/New Old/New > Rec/Fam

Old > New Conjunction

ALE .014 .059 Rec/Fam Old/New

Fig. 2. Above-threshold regions in an ALE meta-analysis of (A) recollection 4familiarity effects and (B) old4 new effects, (C) a subtraction analysis of two ALE results, and
(D) a conjunction analysis of two results. Red and green lines indicate estimated boundaries of the default network and the front-parietal control network, respectively. For a
detailed explanation of these lines, see Fig. 1. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

inferior parietal lobule, the left anteromedial PFC, the left ventral network subregions was addressed by a direct comparison of re-
PFC, the left dorsal PFC, the left lateral temporal cortex, and the left collection/familiarity and word/pseudoword effects. Given that
retrosplenial cortex, in approximate order of decreasing spatial both effects involved a relatively extensive range of default net-
extent (Fig. 3B). work regions but few others (see Figs. 2A and 3A), an additional
Finally, the regions that were associated more with word/ advantage of the comparison was the clear focus on the default
pseudoword effects than semantic/phonological effects involved network. The two effects were also similar in the volume of above-
exclusively default network regions, including the bilateral pos- threshold default network regions approximately 32,272 mm3
terior cingulate cortex and the bilateral inferior parietal lobule for recollection/familiarity and 36,664 mm3 for word/pseudoword
(Fig. 3C). No regions were associated with the reverse effects. The (computed from relevant data in Tables 2 and 3) ensuring a re-
regions common to both task paradigms involved only default latively balanced comparison in terms of the extent of default
network regions, including the left inferior parietal lobule and the network involvement.
left retrosplenial cortex (Fig. 3D). Table 4 and Fig. 4 show the results of this comparison. First,
those regions that were associated more with recollection/famil-
3.3. EM versus SM retrieval iarity effects than with word/pseudoword effects involved only the
bilateral hippocampal formation and (to a lesser extent) the ad-
The results reported earlier indicate greater contributions of jacent parahippocampal cortex region (see Fig. 4A). Second, no
the default network to recollection/familiarity effects than to old/ regions were associated with reverse effects. Third, the regions
new effects and to word/pseudoword effects than to semantic/ common to both task paradigms were exclusively within the de-
phonological effects. Therefore, the issue of whether and to what fault network, including the bilateral posterior cingulate cortex/
extent EM and SM processes involve similar or differential default retrosplenial cortex (more to the left), the left inferior parietal
H. Kim / Neuropsychologia 80 (2016) 3546 41

Table 3
Results of separate ALE meta-analyses of word/pseudoword and semantic/phonological effects and a direct comparison of two effects.

Network Volume (mm3) ALE Z Talairach H Region Brodmann area

x y z

Word4pseudoword
DN 11,248 .030 38 70 36 L Inferior parietal lobule 19, 39, 40
9048 .024 6 54 26 B Posterior cingulate cortex, RSC 23, 31
5360 .019 24 16 52 L Dorsal PFC 6, 8
5104 .019 48 66 16 R Inferior parietal lobule 39, 40
3560 .022 2 46 4 B Anteromedial PFC 9, 10, 32
2344 .025 60 48 8 L Lateral temporal cortex 21

Semantic 4phonological
DN 4520 .021 42 66 14 L Inferior parietal lobule 39
2112 .013 38 38 10 L Ventral PFC 47
2096 .013 8 50 38 L Anteromedial PFC, dorsal PFC 8, 9
1640 .017 10 48 6 L Anteromedial PFC 32
816 .013 54 24 8 L Lateral temporal cortex 21
736 .012 50 22 0 L Ventral PFC 45
728 .013 24 18 38 L Dorsal PFC 8
696 .010 56 44 0 L Lateral temporal cortex 21
696 .011 6 56 12 L RSC 23

Other 1480 .013 42 22 18 L Fusiform gyrus 20


672 .012 8 10 18 L Caudate nucleus

Word/pseudoword 4semantic/phonological
DN 4264 3.89 4 36 37 B Posterior cingulate cortex 23, 31
2328 3.89 38 51 19 R Inferior parietal lobule 39, 40
2152 3.54 37 73 41 L Inferior parietal lobule 39
1272 3.35 54 54 23 L Inferior parietal lobule 40

Semantic/phonological 4word/pseudoword
(no above-threshold clusters)

Word/pseudoword semantic/phonological
DN 2568 .018 42 66 16 L Inferior parietal lobule 39
576 .011 6 56 12 L RSC 23

For abbreviations, see Table 2.

lobule, and the left anteromedial PFC (Fig. 4B). In sum, both effects (Dunn, 2004; Wixted and Mickes, 2010). Along these lines, some
involved a relatively extensive range of default network regions, EM (and SM) task paradigms involving only modest default net-
and these regions overlapped to a large extent. However, the re- work contributions, including old/new, may involve relatively
cruitment of the hippocampus and parahippocampal region was weak memory signals in target conditions. The absence of any
unique to recollection/familiarity effects. hippocampal region associated with old/new effects may also re-
ect some masking by the stronger encoding-related activity of
the region for novel items than for old ones (Stark and Okado,
4. Discussion 2003). Old/new effects overlapped more with fronto-parietal
control regions, providing support for the hypothesis that famil-
4.1. EM retrieval and the default network iarity and associated low recognition condence entail a strong
demand for controlled retrieval processes (Henson et al., 1999;
Recollection/familiarity effects overlapped extensively with the Kim and Cabeza, 2009; Yonelinas et al., 2005).
default network, whereas old/new effects, modestly. According to This study's results and the ndings of previous meta-analyses
a direct comparison of the two effects, many default network re- (Kim, 2013; Spaniol et al., 2009) indicate that old/new effects are
gions were associated more with recollection/familiarity effects. only modestly associated with the default network. This merits
Robust contributions of the default network to recollection/fa- further comments in that it seemingly goes against the hypothesis
miliarity effects provide direct evidence that recollection involves that EM retrieval processes are critically related to the default
the default network more consistently than familiarity. This evi- network. First, as mentioned earlier and in conjunction with the
dence in turn suggests that the lower involvement of the default fact that old/new effects include familiarity-based recognition in
network in old/new effects is tied to the inclusion of familiarity- target conditions, memory signals during familiarity-based re-
based recognition in the target condition. As mentioned earlier, cognition may not be strong enough to activate the default net-
experiments based on condence ratings yielded activation pat- work. Second, low-condence memory judgements generally de-
terns virtually identical to those based on the remember/know mand more controlled processing resources than high-condence
procedure (Daselaar et al., 2006; Kim and Cabeza, 2009). In this ones (Henson et al., 2000), and studies have shown that the task-
regard, familiarity recognition typically involves low recognition induced deactivation of the default network increases as the task
condence and (by inference) weak memory signals. This hy- demand increases (McKiernan et al., 2003, 2006). Along this line,
pothesis resonates with the view that the remember-know pro- familiarity-based recognition may induce the default network's
cedure, as it is ordinarily used, may measure memory strength relatively strong deactivation. Finally, new responses may be
more than it distinguishes between recollection and familiarity based on a recall-to-reject process in which mismatching
42 H. Kim / Neuropsychologia 80 (2016) 3546

Word > Pseudoword Subtraction

Word/Pseudo > Sem/Pho


ALE .010 .030
Sem/Pho > Word/Pseudo (none)

Semantic > Phonological Conjunction

ALE .008 .021 Word/Pseudo Sem/Pho

Fig. 3. Above-threshold regions in an ALE meta-analysis of (A) word4pseudoword effects and (B) semantic 4 phonological effects, (C) a subtraction analysis of two ALE
results, and (D) a conjunction analysis of two results. Red lines indicate estimated boundaries of the default network (see Fig. 1). (For interpretation of the references to color
in this gure legend, the reader is referred to the web version of this article.)

information that is retrieved from memory is used to reject test very different retrieval-related processes. However, studies in-
foils that are similar to studied items (Rotello et al., 2000, p. 67). cluded in the laboratory-based analysis used mainly old/new task
In this regard, reference conditions of old/new effects may involve, paradigms. That is, if a meta-analysis compares recollection/fa-
at least in some conditions, memory signals that activate the de- miliarity to autobiographical tasks, then results may indicate more
fault network to some extent. overlaps involving default network regions. In conjunction with
McDermott et al.'s (2009) meta-analysis compared activated the fact that old/new effects, not recollection/familiarity effects,
regions in studies of laboratory-based EM retrieval versus auto- involve familiarity-based recognition in target conditions, this line
biographical memory retrieval and reported very few overlapping of reasoning indicates that laboratory-based recollection, not fa-
regions, suggesting that the two paradigms may typically probe miliarity, shares relatively extensive processes with

Table 4
Results of a direct comparison of recollection/familiarity and word/pseudoword effects.

Network Volume (mm3) ALE Z Talairach H Region Brodmann area

x y z

Recollection/familiarity 4 word/pseudoword
DN 2616 3.89 22 19 11 R Hippocampus, PHC 35
744 3.72 27 21 7 L Hippocampus

Word/pseudoword4 recollection/familiarity
(No above-threshold clusters)

Recollection/familiarity word/pseudoword
DN 4984 .025 42 68 24 L Inferior parietal lobule 39
4904 .023 10 50 28 B Posterior cingulate cortex, RSC 23, 31
2576 .020 2 48 4 L Anteromedial PFC 10, 32

For abbreviations, see Table 2.


H. Kim / Neuropsychologia 80 (2016) 3546 43

2006). If this line of reasoning is correct, then some SM (and EM)


Subtraction
task paradigms involving only modest default network contribu-
y=-20 y=-25 tions, including semantic/phonological, may involve relatively
strong memory signals in reference conditions. Many SM tasks
emphasize the retrieval of overlearned information such as the
lexicality and meaning of common words and (by inference)
strong memory signals. In such SM tasks, conclusions about which
task paradigms do or do not engage the default network to a large
extent may be dependent mainly on the strength of memory sig-
Rec/Fam > Word/Pseudo nals in the reference task.
Word/Pseudo > Rec/Fam (none) Seemingly disparate cognitive functions engage the default
network, including future thinking, remembering, conceiving the
Conjunction perspective of others (theory of mind), and navigation (Spreng
et al., 2009). A well-known theory (Buckner and Carroll, 2007)
proposes that mentally projecting oneself into an alternative si-
tuation may be a key process common to these functions. A dif-
ferent view (Hassabis and Maguire, 2007) suggests that mental
scene construction is a more fundamental common process. Yet
another view (Schacter et al., 2007) holds that these functions
commonly involve the extraction and recombination of relevant
episodic details from stored memories. On the one hand, the
lexicality decision (or word recognition) is not explicitly linked to
self-projection, mental scene construction, or constructive episo-
dic simulation, which suggests that any such process are sufcient
but not required for default network activation. On the other hand,
most (if not all) instantiations of these processes necessarily in-
clude some semantic representation. Therefore, as suggested in
Binder et al. (2009), the default network may be more funda-
Rec/Fam Word/Pseudo mentally portrayed as a semantic system.

Fig. 4. Above-threshold regions in (A) a subtraction analysis of


recollection4familiarity effects (EM retrieval) and word 4pseudoword effects (SM
4.3. EM versus SM retrieval and the default network
retrieval) and (B) a conjunction analysis of two effects. Red lines indicate estimated
boundaries of the default network (see Fig. 1). (For interpretation of the references This study's results as well as previous ndings indicate that
to color in this gure legend, the reader is referred to the web version of this various episodic and semantic task paradigms differ widely in the
article.)
extent of default network involvement. Therefore, a theoretically
more meaningful comparison of default network subregions in-
autobiographical retrieval. This is perhaps not surprising in that, volved in EM versus SM processes requires this factor to be con-
although the distinction between recollection and familiarity is trolled for to some extent. Along this line, the present study
seldom made explicit in studies of autobiographical retrieval, most compares recollection/familiarity effects (EM retrieval) and word/
cases of retrieval within the paradigm likely involve recollection. pseudoword effects (SM retrieval), each involving a range of de-
fault network regions and relatively few other regions. According
4.2. SM retrieval and the default network to the results, the two effects had more overlaps than separations
in default network subregions involved in the two effects. More
Both word/pseudoword and semantic/phonological effects in- specically, overlaps included the bilateral posterior cingulate/
volved mainly default network regions, but this involvement was retrosplenial cortex, the left inferior parietal lobule, and the left
much more extensive for word/pseudoword effects. According to a anteromedial PFC region, whereas separations included only the
direct comparison of the two effects, many default network re- hippocampal formation and (to a lesser extent) the adjacent
gions were associated more with word/pseudoword effects. Sub- parahippocampal cortex region, which was unique to recollection/
stantial contributions of the default network to word/pseudoword familiarity effects. This implies that EM and SM retrieval processes
effects likely reect access to meaning in the word, not pseudo- involving strong memory signals recruit extensive and largely
word, condition. Behavioral studies have provided strong evidence overlapping default network regions and differ mainly in distinct
that the mere presentation of words typically activates semantic contributions of hippocampus and parahippocampal regions to EM
processes even when subjects are not required to do so (Lesch and retrieval.
Pollatsek, 1993; Lucas, 2000; Shaffer and LaBerge, 1979). In this Noteworthy is the broad overlap of recollection/familiarity and
regard, the lower involvement of the default network in semantic/ word/pseudoword effects in the default network. First, the overlap
phonological effects may reect relatively automatic semantic goes beyond the unequivocal but banal observation that all EM
processing in the phonological condition and the associated partial retrieval processes include SM representations because language
subtraction of semantic activation (Price et al., 1996). Consistent requirements are held constant in target and reference conditions
with this hypothesis, previous studies contrasting a semantic task of recollection/familiarity contrasts. Second, it highlights the
involving words with a phonological task involving pseudowords within-component anatomical specicity of both effects. That is,
have indicated more consistent default network involvement both involved the inferior parietal lobule component more pos-
(Binder et al., 2008; Cappa et al., 1998; Dmonet et al., 1992). If teriorly (i.e., the angular gyrus), the posterior cingulate/retro-
relatively easy phonological tasks are used, then the modest in- splenial cortex component more ventrally, and the anteromedial
volvement of the default network in semantic/phonological effects PFC component more ventrally. Third, the broad overlap involving
may additionally reect the lesser task-induced deactivation of the the default network resonates with the recently proposed semantic
default network in reference conditions (McKiernan et al., 2003, scaffolding hypothesis, which suggests that semantic knowledge
44 H. Kim / Neuropsychologia 80 (2016) 3546

provides a framework or scaffolding that facilitates most (if not all) involves elements of contextual information, scene construction,
forms of episodic remembering and future thinking (Greenberg and even a sense of reliving (Rubin and Umanath, 2015) as well as
and Verfaellie, 2010; Irish and Piguet, 2013). Finally, the results an episodic retrieval strategy during SM tasks.
raise the question of what type of common process is recruited A competing theory to the hypothesis that the hippocampus
during episodic recollection and word recognition. Here a plau- plays a more important role in EM retrieval than SM retrieval is
sible hypothesis is that strong memory signals, both episodic and the consolidation model of memory (Squire et al., 2004), which
semantic, may typically capture a subject's attention away from posits that both EM and SM are relatively equally dependent on
the visual stimulus (i.e., perceptual decoupling) and drive toward the hippocampal region and that it plays a time-limited role in
internal information. This hypothesis is consistent with the evi- both types of memory. The hippocampal region supports the
dence that the default network supports internally oriented maintenance and retrieval of both EM and SM but only until
mentation (AndrewsHanna et al., 2014) and may provide some memory consolidation is complete. Proponents of this view have
testable predictions. obtained supporting evidence from both neuroimaging and lesions
As discussed in the Introduction section, previous within-study studies, including evidence that hippocampal activity during SM
comparisons of laboratory-based EM and SM tasks have typically retrieval is related to the age of memory (Manns et al., 2003;
indicated limited overlaps involving the default network. How- Smith and Squire, 2009). The consolidation model of memory can
ever, such studies have employed EM and/or SM task paradigms account for greater hippocampal involvement in recollection/fa-
involving relatively weak contributions of the default network, miliarity than word/pseudoword effects because laboratory-based
including old/new and semantic/phonological contrasts. A smaller recollection involves newly acquired memories, whereas word
set of previous studies has employed the autobiographical mem- recognition, remotely acquired memories. In this regard, it remains
ory paradigm. For example, Maguire and Frith (2003) compared a unclear whether and to what extent differential hippocampal in-
condition involving true/false judgments about autobiographical volvement reects memory type or age or both. One related con-
event statements (e.g., You were the Christmas star in the school cern is that if memory consolidation necessarily (or typically) in-
nativity play) and a condition involving true/false judgments volves a transformation of episodic memory to semantic memory
about autobiographical fact statements (e.g., Winkey and Frawley (Winocur and Moscovitch, 2011), the question of whether the
were friends at school). Such studies (e.g., Addis et al., 2004; hippocampus supports episodic memory or newly acquired
Holland et al., 2011; Levine et al., 2004; Maguire and Frith, 2003; memory becomes a denitional, not empirical, issue.
Zysset et al., 2002) have generally indicated that autobiographical According to the results, no default network (and other) region
EM and SM tasks both involve an extensive range of default net- was associated more with word/pseudoword effects than with
work regions and that these default network regions show some recollection/familiarity effects. It is difcult to argue strongly from
broad overlaps. Here the interpretive ambiguity concerns whether negative evidence, and the results may reect the specic task
and to what extent such broad overlaps reect common self-re- paradigms included in the study. However, previous within-study
ferential components or more memory-specic commonalities. comparisons of autobiographical EM retrieval and auto-
The results of the present study demonstrate that at least some biographical SM retrieval have rarely indicated any default net-
commonalities are due to memory-specic processes by showing work region being activated more by the SM condition than by the
that a similar overlap can be observed even when self-referential EM condition. In this regard, the present study as well as previous
processing is not a common component. The study also extends studies indicates dissociation involving EM but not that involving
the literature on autobiographical memory by showing that epi- SM. A single case of dissociation involving EM provides support for
sodic and semantic processing tasks at even very low levels can Tulvings (1989) hierarchical view of memory organization, which
yield similar overlaps and separations. suggests that the episodic system depends on but goes beyond
The bilateral hippocampal formation and the adjacent para- the capabilities of the semantic system. The semantic system, on
hippocampal region were largely specic to recollection/famil- the other hand, is capable of operating in the absence, or in-
iarity effects, providing support for the hypothesis that these re- dependently, of the episodic system (p. 362). As suggested in
gions play a more pivotal role in EM retrieval than in SM retrieval Barba et al. (1998), this hierarchical organization requires that
(Nadel et al., 2000; Tulving and Markowitsch, 1998; Vargha-Kha- episodic memory involves all regions contributing to semantic
dem et al., 1997). Previous within-study comparisons of auto- memory and some additional regions not contributing to semantic
biographical EM retrieval versus autobiographical SM retrieval memory.
have frequently indicated greater contributions of the hippo-
campal region to EM retrieval (Holland et al., 2011; Levine et al., 4.4. Methodological considerations and conclusions
2004; Maguire and Frith, 2003). Studies of patients with hippo-
campal damage have generally indicated some severe decit in A possible criticism of the present selective comparison is
clinical measures of EM processes but largely intact SM processes that the broader inclusion of task paradigms in each domain may
based on standardized tests (e.g., Parkin and Leng, 1993). There- yield more generalizable and possibly more useful results. If a
fore, the hippocampus may be the key neural structure under- study's aim is only the numerical integration of results across a
pinning some distinct characteristics of EM relative to SM, such as large number of studies with relatively few theoretical concerns,
the retrieval of detailed contextual information (Yonelinas, 2013), then a more inclusive approach may be more suitable. However,
scene construction (Maguire and Mullally, 2013), and a sense of given that the present study also pursues a theoretical clarication
reliving or autonoetic consciousness (Klein, 2013; Tulving and of previous ndings, there were some reasons behind this study's
Markowitsch, 1998). Although word/pseudoword effects involved choice of a more selective approach. First, a more inclusive com-
no hippocampal or other medial temporal lobe activation, pre- parison may yield less interpretable results because different task
vious studies have shown that more complex SM tasks such as paradigms within each domain are often associated with strikingly
autobiographical fact retrieval, the retrieval of public facts, and heterogeneous effects (e.g., old/new vs. recollection/familiarity). To
semantic uency can recruit some hippocampal regions (Maguire illustrate this point, if a comparison involves old/new effects (EM)
and Frith, 2003; Sheldon and Moscovitch, 2012). Therefore, the and word/pseudoword effects (SM), then results may probably
hippocampal contribution to EM versus SM may be a relative (but not very plausibly) indicate that EM processes involve mainly
difference instead of a sharp dichotomy. Hippocampal contribu- fronto-parietal control network regions, whereas SM processes,
tions to SM tasks may reect the fact that some SM retrieval mainly default network regions. Such nontrivial confounding and
H. Kim / Neuropsychologia 80 (2016) 3546 45

interpretive ambiguity may not necessarily be mitigated by in- declarative memory retrieval processes than with the view that
cluding more paradigms in the analysis because each domain is two spatially segregated and independent memory systems sup-
now represented by multiple and often markedly dissimilar port EM and SM processes. However, it should be noted that by
effects. focusing exclusively on default network regions and thus on strong
Second, the ultimate aim of this study is to determine whether memory signals, the present study ignores potential EM and SM
and the extent to which the EM and SM processes involve similar differences involving other brain regions that may support more
or differential default network subregions. This issue may be controlled and other retrieval-related processes. In this regard, an
better addressed using EM and SM task paradigms involving re- important area of future research is to compare EM and SM re-
latively extensive default network involvement, such as recollec- trieval processes in terms of not only retrieval success but also
tion/familiarity and word/pseudoword, than using those involving other components such as modes, effort, and orientation (Rugg
only modest default network involvement, such as old/new and and Wilding, 2000).
semantic/phonological. In addition, in most cases, an ideal mea-
sure of both episodic and semantic memory processes involves
high memory signals in target conditions and low or no memory Acknowledgments
signals in baseline conditions. As discussed earlier, both old/new
and semantic/phonological task paradigms are likely to fail this This research was supported by the Daegu University Research
criterion to a nontrivial extent, and their inclusion is likely to make Grant in 2014.
results less interpretable in exchange for supercially higher but
not necessarily more meaningful generalizability. Taken together,
these considerations suggest that a selective comparison involving Appendix A. Supplementary material
recollection/familiarity and word/pseudoword task paradigms but
not the other two task paradigms may provide better control for Supplementary data associated with this article can be found in
biasing factors and more interpretable and useful results. the online version at http://dx.doi.org/10.1016/j.neuropsychologia.
Of course, it is true that this study's results are generalizable 2015.11.006.
only to the two selected task paradigms and thus not necessarily
to other EM and SM task paradigms. Therefore, any generalization
based on this study's results should be made with caution. The References
foregoing discussion provides some support for the hypothesis
that recollection/familiarity and word/pseudoword task paradigms Addis, D.R., Moscovitch, M., Crawley, A.P., McAndrews, M.P., 2004. Recollective
but not old/new and semantic/phonological ones involve both qualities modulate hippocampal activation during autobiographical memory
strong memory signals in target conditions and weak memory retrieval. Hippocampus 14, 752762.
AndrewsHanna, J.R., Smallwood, J., Spreng, R.N., 2014. The default network and
signals in baseline conditions. If this argument is correct at least to selfgenerated thought: component processes, dynamic control, and clinical
some extent, then this study's results may be generalized not only relevance. Ann. N Y. Acad. Sci. 1316, 2952.
to the two selected task paradigms but also to other EM and SM Balota, D.A., Cortese, M.J., Sergent-Marshall, S.D., Spieler, D.H., Yap, M., 2004. Visual
word recognition of single-syllable words. J. Exp. Psychol. Gen. 133, 283316.
task paradigms involving strong memory signals in target condi-
Barba, G.D., Parlato, V., Jobert, A., Samson, Y., Pappata, S., 1998. Cortical networks
tions and weak memory signals in reference conditions. In this implicated in semantic and episodic memory: common or unique? Cortex 34,
regard, it is worth noting again that previous within-study com- 547561.
parisons of autobiographical EM and SM retrieval processes have Binder, J.R., Desai, R.H., Graves, W.W., Conant, L.L., 2009. Where is the semantic
system? A critical review and meta-analysis of 120 functional neuroimaging
generally indicated that they both involve an extensive range of studies. Cereb. Cortex 19, 27672796.
default network regions and that these default network regions Binder, J.R., Swanson, S.J., Hammeke, T.A., Sabsevitz, D.S., 2008. A comparison of ve
show extensive overlaps (e.g., Addis et al., 2004; Holland et al., fMRI protocols for mapping speech comprehension systems. Epilepsia 49,
19801997.
2011; Levine et al., 2004; Maguire and Frith, 2003; Zysset et al., Buckner, R., Carroll, D., 2007. Self-projection and the brain. Trends Cogn. Sci. 11,
2002). These results provide evidence that what can be inferred 4957.
from this study's results is not strictly limited to the two selected Burianova, H., Grady, C.L., 2007. Common and unique neural activations in auto-
biographical episodic and semantic retrieval. J. Cogn. Neurosci. 19, 15201534.
paradigms but also to some EM and SM task paradigms involving Cappa, S., Perani, D., Schnur, T., Tettamanti, M., Fazio, F., 1998. The effects of se-
strong memory signals. mantic category and knowledge type on lexical-semantic access: a PET study.
In conclusion, the present comparison of recollection/famil- Neuroimage 8, 350359.
Daselaar, S.M., Fleck, M.S., Cabeza, R., 2006. Triple dissociation in the medial tem-
iarity and word/pseudoword effects provides evidence that EM
poral lobes: recollection, familiarity, and novelty. J. Neurophysiol. 96,
and SM retrieval processes involving strong memory signals re- 19021911.
cruit extensive and largely overlapping default network regions Dmonet, J.F., Chollet, F., Ramsay, S., Cardebat, D., Nespoulous, J.L., Wise, R., Rascol,
and differ mainly in distinct contributions of hippocampus and A., Frackowiak, R., 1992. The anatomy of phonological and semantic processing
in normal subjects. Brain 115, 17531768.
parahippocampal regions to EM retrieval. The broad overlap pro- Dzel, E., Cabeza, R., Picton, T.W., Yonelinas, A.P., Scheich, H., Heinze, H.-J., Tulving,
vides support for the notion that SM retrieval is a logical pre- E., 1999. Task-related and item-related brain processes of memory retrieval.
requisite for EM retrieval and also for a more formal but associated Proc. Natl. Acad. Sci. USA 96, 17941799.
Dunn, J., 2004. Remember-know: a matter of condence. Psychol. Rev. 111,
view that semantic knowledge provides scaffolding for EM re- 524542.
trieval (Greenberg and Verfaellie, 2010; Irish and Piguet, 2013). Eickhoff, S.B., Bzdok, D., Laird, A.R., Roski, C., Caspers, S., Zilles, K., Fox, P.T., 2011. Co-
The greater association of the hippocampal formation with re- activation patterns distinguish cortical modules, their connectivity and func-
tional differentiation. Neuroimage 57, 938949.
collection/familiarity effects than with word/pseudoword effects is Eickhoff, S.B., Laird, A.R., Grefkes, C., Wang, L.E., Zilles, K., Fox, P.T., 2009. Coordinate-
consistent with its presumed pivotal role in EM, not SM, retrieval, based activation likelihood estimation meta-analysis of neuroimaging data: a
although it remains unclear whether it may alternatively (or ad- random-effects approach based on empirical estimates of spatial uncertainty.
Hum. Brain Mapp. 30, 29072926.
ditionally) reect a greater role of the hippocampus in the retrieval
Emmorey, K., Weisberg, J., McCullough, S., Petrich, J.A., 2013. Mapping the reading
of newly acquired memories than in the retrieval of remote and circuitry for skilled deaf readers: an fMRI study of semantic and phonological
consolidated memories. More broadly, the study provides evi- processing. Brain Lang. 126, 169180.
dence that neural correlates of EM and SM processes show more Gardiner, J.M., 2001. Episodic memory and autonoetic consciousness: a rst-person
approach. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 356, 13511361.
overlaps than separations. This evidence is more compatible with Greenberg, D.L., Verfaellie, M., 2010. Interdependence of episodic and semantic
the view that a single unied memory system supports all memory: evidence from neuropsychology. J. Int. Neuropsychol. Soc. 16,
46 H. Kim / Neuropsychologia 80 (2016) 3546

748753. visually presented words and pseudowords. Cereb. Cortex 6, 6270.


Hantke, N., Nielson, K.A., Woodard, J.L., Breting, L.M.G., Butts, A., Seidenberg, M., Raichle, M., MacLeod, A., Snyder, A., Powers, W., Gusnard, D., Shulman, G., 2001. A
Carson Smith, J., Durgerian, S., Lancaster, M., Matthews, M., 2013. Comparison default mode of brain function. Proc. Natl. Acad. Sci. USA 98, 676682.
of semantic and episodic memory BOLD fMRI activation in predicting cognitive Rajah, M.N., McIntosh, A.R., 2005. Overlap in the functional neural systems involved
decline in older adults. J. Int. Neuropsychol. Soc. 19, 1121. in semantic and episodic memory retrieval. J. Cogn. Neurosci. 17, 470482.
Hassabis, D., Maguire, E., 2007. Deconstructing episodic memory with construction. Renoult, L., Davidson, P.S., Palombo, D.J., Moscovitch, M., Levine, B., 2012. Personal
Trends Cogn. Sci. 11, 299306. semantics: at the crossroads of semantic and episodic memory. Trends Cogn.
Henson, R., Rugg, M., Shallice, T., Dolan, R., 2000. Condence in recognition memory Sci. 16, 550558.
for words: dissociating right prefrontal roles in episodic retrieval. J. Cogn. Rotello, C.M., Macmillan, N.A., van Tassel, G., 2000. Recall-to-reject in recognition:
Neurosci. 12, 913923. evidence from ROC curves. J. Mem. Lang. 43, 6788.
Henson, R.N.A., Rugg, M.D., Shallice, T., Josephs, O., Dolan, R.J., 1999. Recollection Rubin, D.C., Umanath, S., 2015. Event memory: a theory of memory for laboratory,
and familiarity in recognition memory: an event-related functional magnetic autobiographical, and ctional events. Psychol. Rev. 122, 123.
resonance imaging study. J. Neurosci. 19, 39623972. Rugg, M., Wilding, E., 2000. Retrieval processing and episodic memory. Trends
Holland, A.C., Addis, D.R., Kensinger, E.A., 2011. The neural correlates of specic Cogn. Sci. 4, 108115.
versus general autobiographical memory construction and elaboration. Neu- Ryan, L., Cox, C., Hayes, S.M., Nadel, L., 2008. Hippocampal activation during epi-
ropsychologia 49, 31643177. sodic and semantic memory retrieval: comparing category production and
Irish, M., Piguet, O., 2013. The pivotal role of semantic memory in remembering the category cued recall. Neuropsychologia 46, 21092121.
past and imagining the future. Front. Behav. Neurosci. 7, 27. Schacter, D.L., Addis, D.R., Buckner, R.L., 2007. Remembering the past to imagine the
Kim, H., 2010. Dissociating the roles of the default-mode, dorsal, and ventral net- future: the prospective brain. Nat. Rev. Neurosci. 8, 657661.
works in episodic memory retrieval. Neuroimage 50, 16481657. Shaffer, W.O., LaBerge, D., 1979. Automatic semantic processing of unattended
Kim, H., 2012. A dual-subsystem model of the brains default network: Self-refer- words. J. Verbal Learn. Verbal Behav. 18, 413426.
ential processing, memory retrieval processes, and autobiographical memory Shapira-Lichter, I., Oren, N., Jacob, Y., Gruberger, M., Hendler, T., 2013. Portraying
retrieval. Neuroimage 61, 966977. the unique contribution of the default mode network to internally driven
Kim, H., 2013. Differential neural activity in the recognition of old versus new mnemonic processes. Proc. Natl. Acad. Sci. USA 110, 49504955.
events: an activation likelihood estimation meta-analysis. Hum. Brain Mapp. Sheldon, S., Moscovitch, M., 2012. The nature and time-course of medial temporal
34, 814836. lobe contributions to semantic retrieval: an fMRI study on verbal uency.
Kim, H., 2015. Encoding and retrieval along the long axis of the hippocampus and Hippocampus 22, 14511466.
their relationships with dorsal attention and default mode networks: the Smith, C.N., Squire, L.R., 2009. Medial temporal lobe activity during retrieval of
HERNET model. Hippocampus 25, 500510. semantic memory is related to the age of the memory. J. Neurosci. 29, 930938.
Kim, H., Cabeza, R., 2009. Common and specic brain regions in high- versus low- Spaniol, J., Davidson, P., Kim, A., Han, H., Moscovitch, M., Grady, C., 2009. Event-
condence recognition memory. Brain Res. 1282, 103113. related fMRI studies of episodic encoding and retrieval: meta-analyses using
Klein, S.B., 2013. Making the case that episodic recollection is attributable to op-
activation likelihood estimation. Neuropsychologia 47, 17651779.
erations occurring at retrieval rather than to content stored in a dedicated
Spreng, R., Mar, R., Kim, A., 2009. The common neural basis of autobiographical
subsystem of long-term memory. Front. Behav. Neurosci. 7 (3), 114.
memory, prospection, navigation, theory of mind, and the default mode: a
Lancaster, J., Tordesillas-Gutierrez, D., Martinez, M., Salinas, F., Evans, A., Zilles, K.,
quantitative meta-analysis. J. Cogn. Neurosci. 21, 489510.
Mazziotta, J., Fox, P., 2007. Bias between MNI and Talairach coordinates ana-
Squire, L.R., Stark, C., Clark, R., 2004. The medial temporal lobe. Annu. Rev. Neurosci.
lyzed using the ICBM-152 brain template. Hum. Brain Mapp. 28, 11941205.
27, 279306.
Lesch, M.F., Pollatsek, A., 1993. Automatic access of semantic information by pho-
Stark, C., Okado, Y., 2003. Making memories without trying: medial temporal lobe
nological codes in visual word recognition. J. Exp. Psychol. Learn. Mem. Cogn.
activity associated with incidental memory formation during recognition. J.
19, 285294.
Neurosci. 23, 67486753.
Levine, B., Turner, G.R., Tisserand, D., Hevenor, S.J., Graham, S.J., McIntosh, A.R.,
St-Laurent, M., Abdi, H., Burianov, H., Grady, C.L., 2011. Inuence of aging on the
2004. The functional neuroanatomy of episodic and semantic autobiographical
neural correlates of autobiographical episodic and semantic memory retrieval.
remembering: a prospective functional MRI study. J. Cogn. Neurosci. 16,
J. Cogn. Neurosci. 23, 41504163.
16331646.
Tulving, E., 1972. Episodic and semantic memory. In: Tulving, E., Donaldson, W.
Lucas, M., 2000. Semantic priming without association: a meta-analytic review.
Psychon. Bull. Rev. 7, 618630. (Eds.), Organization of Memory. Academic Press, Oxford, pp. 381402.
Maguire, E.A., Frith, C.D., 2003. Aging affects the engagement of the hippocampus Tulving, E., 1989. Remembering and knowing the past. Am. Sci. 77, 361367.
during autobiographical memory retrieval. Brain 126, 15111523. Tulving, E., Markowitsch, H.J., 1998. Episodic and declarative memory: role of the
Maguire, E.A., Mullally, S.L., 2013. The hippocampus: a manifesto for change. J. Exp. hippocampus. Hippocampus 8, 198204.
Psychol. Gen. 142, 11801189. Turkeltaub, P.E., Eickhoff, S.B., Laird, A.R., Fox, M., Wiener, M., Fox, P., 2012. Mini-
Maguire, E.A., Mummery, C.J., 1999. Differential modulation of a common memory mizing withinexperiment and withingroup effects in activation likelihood
retrieval network revealed by positron emission tomography. Hippocampus 9, estimation metaanalyses. Hum. Brain Mapp. 33, 113.
5461. Van Essen, D., 2005. A population-average landmark- and surface-based (PALS)
Manns, J.R., Hopkins, R.O., Squire, L.R., 2003. Semantic memory and the human atlas of human cerebral cortex. Neuroimage 28, 635662.
hippocampus. Neuron 38, 127133. Vargha-Khadem, F., Gadian, D.G., Watkins, K.E., Connelly, A., Van Paesschen, W.,
McDermott, K.B., Szpunar, K.K., Christ, S.E., 2009. Laboratory-based and auto- Mishkin, M., 1997. Differential effects of early hippocampal pathology on epi-
biographical retrieval tasks differ substantially in their neural substrates. sodic and semantic memory. Science 277, 376380.
Neuropsychologia 47, 22902298. Wiggs, C.L., Weisberg, J., Martin, A., 1998. Neural correlates of semantic and epi-
McKiernan, K., DAngelo, B., Kaufman, J., Binder, J., 2006. Interrupting the stream of sodic memory retrieval. Neuropsychologia 37, 103118.
consciousness: an fMRI investigation. Neuroimage 29, 11851191. Winocur, G., Moscovitch, M., 2011. Memory transformation and systems con-
McKiernan, K., Kaufman, J., Kucera-Thompson, J., Binder, J., 2003. A parametric solidation. J. Int. Neuropsychol. Soc. 17, 766780.
manipulation of factors affecting task-induced deactivation in functional neu- Wixted, J.T., Mickes, L., 2010. A continuous dual-process model of remember/know
roimaging. J. Cogn. Neurosci. 15, 394408. judgments. Psychol. Rev. 117, 10251054.
Nadel, L., Samsonovich, A., Ryan, L., Moscovitch, M., 2000. Multiple trace theory of Yeo, B.T.T., Krienen, F.M., Sepulcre, J., Sabuncu, M.R., Lashkari, D., Hollinshead, M.,
human memory: computational, neuroimaging, and neuropsychological re- Roffman, J.L., Smoller, J.W., Zllei, L., Polimeni, J.R., 2011. The organization of the
sults. Hippocampus 10, 352368. human cerebral cortex estimated by intrinsic functional connectivity. J. Neu-
Neisser, U., 1981. John Deans memory: a case study. Cognition 9, 122. rophysiol. 106, 11251165.
Nyberg, L., Marklund, P., Persson, J., Cabeza, R., Forkstam, C., Petersson, K.M., Ingvar, Yonelinas, A.P., 2013. The hippocampus supports high-resolution binding in the
M., 2003. Common prefrontal activations during working memory, episodic service of perception, working memory and long-term memory. Behav. Brain
memory and semantic memory. Neuropsychologia 41, 371377. Res. 254, 3444.
Parkin, A.J., Leng, N.R.C., 1993. Neuropsychology of the Amnestic Syndrome. Yonelinas, A.P., Otten, L.J., Shaw, K.N., Rugg, M.D., 2005. Separating the brain regions
Lawrence Erlbaum Associates, Hove, UK. involved in recollection and familiarity in recognition memory. J. Neurosci. 25,
Price, C.J., Moore, C., Humphreys, G., Wise, R., 1997. Segregating semantic from 30023008.
phonological processes during reading. J. Cogn. Neurosci. 9, 727733. Zysset, S., Huber, O., Ferstl, E., Von Cramon, D., 2002. The anterior frontomedian
Price, C.J., Wise, R., Frackowiak, R., 1996. Demonstrating the implicit processing of cortex and evaluative judgment: an fMRI study. Neuroimage 15, 983991.

You might also like