Segregating cognitive functions within hippocampal formation: A quantitative meta‐analysis on spatial navigation and episodic memory

Abstract

The most important cognitive domains where hippocampal formation is crucially involved are navigation and memory. Some evidence suggests that different hippocampal subregions mediate these domains. However, a quantitative meta‐analysis on neuroimaging studies of spatial navigation versus memory is lacking. By means of activation likelihood estimation (ALE), we investigate concurrence of brain regions activated during spatial navigation encoding and retrieval as well as during episodic memory encoding and retrieval tasks in humans. During encoding in spatial navigation, activity was located in more posterior regions of the hippocampal formation, whereas episodic memory encoding was located in more anterior regions. Retrieval in spatial navigation was more strongly lateralized to the right compared to episodic memory retrieval. Within studies on spatial navigation retrieval, immediate recall was located more posterior and delayed recall more anterior. Overlap between concurrence of activation in spatial navigation and episodic memory was rather limited in comparison to uniquely involved regions. This argues in favor of two distinct networks, one for spatial navigation the other for episodic memory within the hippocampal formation. Hum Brain Mapp 35:1129–1142, 2014. © 2013 Wiley Periodicals, Inc.

INTRODUCTION

The hippocampus and its surrounding structures are one of the most researched structures of the brain. However, it is an issue of debate whether the hippocampal system is specialized in spatial information processing or whether it plays a more general role in higher brain function such as memory and learning. Previous research on rodents and monkeys has supported the former view. In recordings of single neurons in freely moving rodents, neurons have been identified that are active predominately when the animal passes through a particular area in space [Derdikman and Moser, 2010; Moser et al., 2008; O'Keefe and Dostrovsky, 1971; Wilson and McNaughton, 2000]. These cells are active both in light and dark, suggesting that a single modality such as vision is not responsible for their positional firing [Quirk et al., 1990]. Consequently, these neurons have been termed place cells and a theory has been developed suggesting that the hippocampus acts as a cognitive map containing a representation of spatial orientation. In line with these findings, a recent fMRI study on humans revealed that place‐related information could be decoded particularly well from the bilateral hippocampi using multivoxel pattern analysis [Hassabis et al., 2009; Rodriguez, 2010aa,b]. Although the cognitive map theory suggests that episodic, but not semantic, memory is mediated by the hippocampus [Burgess et al., 2002; Moscovitch et al., 2006], the cognitive map has not been conceptually restricted to spatial representations. For instance, it has been proposed that relationships among multiple stimuli as well as contingencies and configurations may also be encoded within the hippocampus [Cohen and Eichenbaum, 1991; Eichenbaum and Cohen, 1988].

On the other hand, the famous case study of patient H.M. [Scoville and Milner, 2012] suggests that the hippocampus is of crucial importance for memory formation. More recent observations of patients, who sustained bilateral damage to the hippocampus early in life, indicate that such damage leads to deficits in memory for episodes and events, rather than for semantic material [Vargha‐Khadem et al., 1997]. Episodic memory has been defined as the ability to remember personally experienced episodes in a spatial and temporal context [Tulving, 1998]. How the consolidation of memory is accomplished has been extensively discussed in the literature. It has been debated whether memory storage initially requires hippocampal linking of dispersed neocortical storage sites, but over time this need seems to dissipate, and the hippocampal component is rendered unnecessary [Squire and Alvarez, 2006]. This so‐called standard consolidation theory assumes that as the consolidation process proceeds, the employment of other extrahippocampal structures sustain the permanent memory trace and mediate its retrieval. However, the so‐called multiple trace theory [Nadel and Moscovitch, 1997; Nadel et al., 2000] posits that each and every time information is presented, it is neurally encoded in a unique memory trace and all memory traces incorporated over time are combined into a multiple‐trace representation. In this view, hippocampal ensembles are always involved in storage and retrieval of episodic information, but the semantic gist of information can be established in neocortex without hippocampal contribution.

In rodents, experimental paradigms based on spatial navigation are often used as an operationalization of memory. Whereas in humans, these two lines of experimental research on the role of hippocampus have been pursued more or less independently. However, on a theoretical level, several unifying theories of hippocampal spatial and memory function have been proposed on a conceptually coarse scale. Burgess [2002] supposed that retrieval of information from long‐term storage requires the imposition of a particular viewpoint and therefore harbors spatial processing. Others have emphasized the commonality in the process of self‐projection [Buckner and Carroll, 2007]. In a similar line, it has been suggested that by definition episodic memories include information about time and place and therefore contextual information and scene construction is a necessary prerequisite for episodic memory [Hassabis and Maguire, 2007; Hasselmo et al., 2010; Kentros, 2006; Leutgeb et al., 2005].

Attempts to unite functions associated with activity in hippocampus on a theoretical level stand in contrast to evidence suggesting different localization of episodic memory and spatial navigation within the hippocampal formation. A multi‐electrode recording study in the hippocampus of rats has demonstrated an anatomically separated and highly specialized division of labor between regions that were active during spatial aspects of a task and regions that responded to nonspatial aspects of a task [Hampson et al., 1999]. In humans [Maguire et al., 2000, 2006] as well as in primates [Moser and Moser, 1998] an anterior–posterior distinction (also referred to as rostral‐caudal distinction and equivalent to a ventral‐dorsal distinction in rodents) within the hippocampus has been proposed, with the posterior part being more strongly involved in spatial navigation. Another frequently suggested division of labor across species is the lateralization of hippocampal involvement with navigation dominating in the right and memory in the left hemisphere [Burgess, 2002; Postma et al., 2008]. Likewise, patients with damage to the right temporal lobe frequently exhibit selective deficits in memory of the location of objects while memory of the identity of objects themselves is preserved [Smigh and Milner, 1982].

Within the scope of this study, we set out to systematically test the proposed posterior location and the right hemispheric lateralization of spatial navigation within the hippocampus. To explore the unique and overlapping regions involved in spatial navigation and episodic memory within the hippocampal formation, we conducted a quantitative meta‐analysis on neuroimaging studies in humans. The used activation likelihood estimation (ALE) approach [Eickhoff et al., 2009; Laird et al., 2005; Turkeltaub et al., 2002] enabled the identification of concordance of activated voxels across numerous studies on spatial navigation. It also allowed comparisons to be made with respect to the location of concordance within studies on episodic memory.

MATERIALS AND METHODS

Selection of Studies

Studies were selected using a systematic search process. For the episodic memory analyses, we used the coordinate database Brainmap Sleuth (http://brainmap.org/sleuth/index.html) because it contains neuroimaging coordinates classified as memory tasks. We used the search terms [Diagnosis=Normals] AND [Behavioral Domain=Cognition.Memory] OR [Class=Encoding]. Moreover, we excluded studies in which the stimulus material was explicitly spatial or scenic in nature including autobiographical memory. Since, to our knowledge, no database of neuroimaging coordinates contains classification terms such as navigation or spatial memory, we performed a literature search manually. Peer‐reviewed articles published in English until March 2011 were selected from the search results of two separate literature databases (Pubmed, ISI Web of Knowledge). Keyword searches were conducted for the spatial navigation analyses using the following terms: (1) “neuroimaging” <OR> “fMRI” <OR> “PET,” and (2) “spatial navigation” <OR> “navigation” <OR> “route.”

From the resulting papers we selected those that presented contrasts reflecting brain activity during spatial navigation or episodic memory in comparison to a control condition. The reference lists of the selected papers were searched for additional studies that fit these criteria. We included all studies of which we were able to obtain MNI or Talairach coordinates [Talairach and Tournoux, 2011] of contrast. We only included coordinates resulting from analyses computed across the whole brain and not restricted by partial coverage, regions of interest or small volume correction. Because both fMRI and PET have been used to identify the neural correlates of navigation and memory, we included data from studies using either method despite the fact that they have a different physiological basis. Our rationale was to provide an all‐embracing overview of the attempts to identify the neural correlates of spatial navigation and episodic memory. In total a number of 29 spatial navigation studies were included, of which 11 were included in an analysis exploring encoding in spatial navigation. Twenty‐one studies exploring retrieval in spatial navigation were also included. The spatial navigation encoding analysis comprised 185 foci of altogether 237 participants, the spatial navigation retrieval analysis 353 foci of altogether 295 participants (Table 1). In the episodic memory encoding analysis, we included 19 studies comprising 263 foci of 306 participants (Table 2), in the episodic memory retrieval analysis 35 studies with 387 foci of altogether 593 participants were considered (Table 3). In the tables, we refer to “baseline” whenever the contrast was computed as an implicit baseline.

Table 1.

Spatial navigation studies included in the meta‐analysis

Study	Modality	n	Foci	Contrast	Encoding	Retrieval	Retrieval delay
Avila et al. [2006]	fMRI	12	11	Mental navigation to landmarks in hometown > covertly counting numbers		X	Long
Baumann et al. [2010]	fMRI	17	18, 24	Invisible object in abstract virtual maze as target > visible object as target	X	X	Short
Brown et al. [2010]	fMRI	22	19	Navigation in maze overlapping > non‐overlapping with previously learned virtual maze		X	Short
Ghaem et al. [1997]	PET	5	14	Mental navigation of a walk learned before > resting condition		X	Short
Hartley et al. [2003]	fMRI	16	8, 1	Way finding > trail following, memorized route finding > way finding	X	X	Short
Iaria et al. [2003]	fMRI	14	17	Navigation in virtual maze > visible object as target	X
Iaria et al. [2007]	fMRI	16	16, 9	Map learning in virtual maze > trail following, Map retrieval in virtual maze > trail following	X	X	Short
Iaria et al. [2008]	fMRI	10	19	Navigation in familiar virtual maze > navigation in familiarization phase		X	Short
Igloi et al. [2010]	fMRI	19	22	Navigation in virtual maze > trials in environment without landmarks		X	Short
Ino et al. [2002]	fMRI	16	10	Mental navigation in home town > control task: counting backwards from 3digit number		X	Long
Jordan et al. [2004]	fMRI	10	15	Navigation in virtual maze > trail following		X	Short
Latini‐Corazzini et al. [2010]	fMRI	16	17	Snapshot of virtual environment, indicate direction to follow route > decide on orientation of a house relative to body midline		X	Short
Maguire et al. [1997]	PET	11	11	Mental navigation of taxi drivers in hometown > number repetition		X	Long
Maguire et al. [1998]	PET	10	4	Navigation in virtual town > trail following		X	Short
Marsh et al. [2010]	fMRI	25	9	Navigation in virtual maze > trail following	X
Mayes et al. [2004]	fMRI	9	3	Personal memory of route episodes > personal memory of static episodes		X	Long
Mellet et al. [2000]	PET	5	11	Mental navigation from one landmark to the other > resting condition		X	Short
Moffat et al. [2006]	fMRI	30	25	Navigation in virtual environment > trail following (young participants)	X
Moffat et al. [2006]	fMRI	21	21	Navigation in virtual environment > trail following (old participants)	X
Ohnishi et al. [2006]	fMRI	56	22	Passive navigation trough virtual maze > passive movement on straight path	X
Orban et al. [2006]	fMRI	24	47	Navigation in virtual environment > baseline		X	Short
Pine et al. [2002]	fMRI	20	27	Navigation in virtual environment > trail following		X	Short
Rauchs et al. [2008]	fMRI	16	28	Navigation in virtual environment, find alternative way if original is blocked > navigation in the same environment		X	Short
Rodriguez et al. [2010]	fMRI	11	13	Navigation retrieval in virtual environment > navigation encoding		X	Short
Rosenbaum et al. [2004]	fMRI	10	7	Mental navigation in virtual environment, find alternative way if original is blocked > judge whether one of two landmarks contains more vowels		X	Long
Shelton et al. [2002]	fMRI	12	18	Navigation in virtual environment > fixation	X
Spiers et al. [2006]	fMRI	20	31	Navigation in virtual hometown and mentally planning routes > not thinking while navigating		X	Long
Weniger et al. [2010]	fMRI	19	17	Navigation in virtual maze > baseline	X
Wolbers et al. [2005]	fMRI	11	13	Navigation in virtual environment > navigation on one straight road with landmarks	X

Table 2.

Encoding contrasts in non‐spatial memory studies included in the meta‐analysis

Study	Modality	n	Foci	Stimuli	Contrast
Achim et al. [2005]	fMRI	18	26	Animals, objects	Encoding of items and associations > pair of abstract images
Beauregard et al. [1998]	PET	13	4	Words	Subliminal incidental encoding > numbers on screen
Braver et al. [2001]	fMRI	28	2	Words, faces	Encoding > baseline
Dannhauser et al. [2008]	fMRI	10	2	Words	Encoding > reading
Dupont et al. [2002]	fMRI	10	17	Words	Encoding > baseline
Halsband et al. [2002]	PET	10	17	Word pairs	Encoding > nonsense words
Halsband [2006]	PET	7	4	Word pairs	Encoding > nonsense words
Ino et al. [2004]	fMRI	39	29	Word pairs	Encoding > repeating numbers
Jager et al. [2007]	fMRI	40	6	Photos	Associative learning > classification of photos
Kapur et al. [1996]	PET	12	5	Word pairs	Encoding > reading
Kelley et al. [1998]	fMRI	5	55	Words, objects, faces	Encoding > fixation
Meltzer et al. [2005]	fMRI	12	5	Word pairs	Encoding novel > encoding familiar
Mottaghy et al. [1999]	fMRI	6	11	Word pairs	Encoding > nonsense words
Neuner et al. [2007]	fMRI	15	15	Colored shapes	Encoding > fixation
Pihlajamäki et al. [2003]	fMRI	12	15	Animals, objects	Encoding > tracking
Ragland et al. [2001]	PET	23	7	Words	Encoding > finger tapping
Savage et al. [2001]	PET	8	6	Words	Encoding > fixation
Sperling et al. [2001]	fMRI	8	15	Face‐name pairs	Encoding > fixation
Sperling et al. [2003]	fMRI	30	22	Face‐name pairs	Encoding > fixation

Table 3.

Retrieval contrasts in nonspatial memory studies included in the meta‐analysis

Study	Modality	n	Foci	Stimuli	Contrast
Braver et al. [2001]	fMRI	28	1	Words, faces	Recognition > baseline
Buckner et al. [1998]	fMRI	26	53	Words	Retrieval > fixation
Burianova et al. [2007]	fMRI	12	7	Pictures	Retrieval > scrambled pictures
Cabeza et al. [1997]	PET	24	10	Word pairs	Retrieval > encoding
Daselaar et al. [2001]	fMRI	13	9	Words	Retrieval > baseline
Daselaar et al. [2006]	fMRI	14	20	Words, non‐words	Recollection, familiarity > baseline
Dupont et al. [2002]	fMRI	10	30	Words	Retrieval > fixation
Düzel et al. [2001]	PET	11	2	Words	Retrieval > new words
Grady et al. [2001]	PET	12	18	Words, pictures	Retrieval > new words, pictures
Grosbras et al. [2001]	fMRI	10	12	Eye movement sequences	Retrieval > baseline
Halsband et al. [1998]	PET	13	7	Word pairs	Retrieval > control condition
Halsband et al. [2002]	PET	10	6	Words pairs	Retrieval > control condition
Halsband et al. [2006]	PET	7	5	Word pairs	Retrieval > control condition (Reference II)
Heckers et al. [1998]	PET	8	3	Words	Retrieval > word generation
Hofer et al. [2003]	fMRI	20	15	Words	Retrieval > baseline
Jernigan et al. [1998]	PET	8	6	Words	Retrieval > identification
Johnson et al. [2006]	fMRI	77	7	Pictures	Retrieval > new drawings
Kensinger et al. [2007]	fMRI	19	25	Pictures	Retrieval > baseline
Köhler et al. [2000]	PET	12	11	Pictures, words	Retrieval > encoding
Krause et al. [1999]	PET	12	1	Word pairs	Retrieval > control condition
Lepage et al. [2001]	PET	12	9	Visual and haptic objects	Retrieval > encoding
Mensebach et al. [2009]	fMRI	18	12	Words	Retrieval > baseline
Mottaghy et al. [1999]	fMRI	6	10	Word pairs	Retrieval > control condition (Reference II)
Neuner et al. [2007]	fMRI	15	20	Abstract objects, color, and shape	Retrieval (immediate, delayed) > motor baseline
Nyberg et al. [2000]	PET	11	3	Sentences, pictures	Retrieval > encoding
Ongur et al. [2005]	fMRI	15	17	Abstract picture pairs	Retrieval > baseline
Pihlajamaki et al. [2003]	fMRI	12	10	Picture pairs	Retrieval > tracking
Prince et al. [2005]	fMRI	14	10	Word pairs	Retrieval > encoding
Ragland et al. [2001]	PET	46	8	Words	Retrieval > finger tapping
Reber et al. [2002]	fMRI	10	12	Dot pattern	Retrieval > counting
Ries et al. [2006]	fMRI	14	7	Pictures	Retrieval > new stimuli
Sperling et al. [2003]	fMRI	30	13	Face‐name pairs	Retrieval > fixation
Squire et al. [1992]	PET	18	3	Words	Retrieval > baseline
Taylor et al. [1998]	PET	8	2	Emotional pictures	Retrieval > encoding
Wheeler et al. [2000]	fMRI	18	3	Pictures	Retrieval > fixation

Creation of ALE Maps

The ALE method provides a voxel‐based meta‐analytic technique for neuroimaging data [Eickhoff et al., 2009; Turkeltaub et al., 2002]. By means of the software Brainmap GingerALE (http://brainmap.org/ale/) statistically significant concordance in the pattern of brain activity among several independent experiments was computed. ALE maps display regions in the brain that comprise statistically significant peak activation locations from multiple studies with a resolution of 2 mm × 2 mm × 2 mm. Coordinates reported in Talairach were converted to MNI using Lancster et al. [2007] (icbm2tal). In the approach taken by ALE, probability distributions for the foci were modeled at the center of 3D Gaussian functions, and the Gaussian distributions were summed across the entire set of experiments to generate a map of interstudy consistencies that estimated the likelihood of activation for each voxel—the ALE statistic. The false discovery rate (FDR) method was used to correct for multiple comparisons at a significance threshold of P < 0.01 and a cluster threshold of 100. Coordinates are reported in MNI space.

Conjunction Analysis

The ALE maps were imported into SPM8 (Wellcome Trust Centre for Neuroimaging, London, UK), to undertake a conjunction analysis to examine the correspondence of consistently activated regions in spatial navigation and episodic memory studies. The conjunction was determined by multiplying the resulting ALE maps. This conjunction does not constitute a statistical test but depicts regions of overlap.

RESULTS

The convergence of all coordinates reported in studies of encoding and retrieval in spatial navigation and episodic memory can be found in Table 4. The main focus of our analysis was the exploration of differential activation within the hippocampal formation in spatial navigation as compared to episodic memory without explicit spatial aspects. Figure 1 depicts the results within this region of interest.

Table 4.

Statistical concurrence observed across studies on spatial navigation and nonspatial memory

Anatomical region	Brodmann area	Coordinates (MNI)			Volume (mm3)
		x	y	z
(a) Spatial navigation encoding
Left precuneus, occipital gyrus	19/31	−28	−80	26	1280
Right precuneus	19	33	−75	36	1168
Posterior cingulate	30	−14	−59	7	768
Right precuneus	7	18	−64	60	592
Left parahippocampus/hippocampus	30	−19	−39	−6	568
Right parahippocampus/hippocampus	30	22	−39	−7	520
Left precuneus	7	−15	−62	59	512
Left precentral gyrus	6	−29	−3	58	400
Pre supplementary motor area	6	−7	11	52	128
Left occipital gyrus	19	−45	−83	14	120
Left lingual gyrus	18	−5	−82	0	112
(b) Spatial navigation retrieval
Right parahippocampal gyrus/hippocampus	35	26	−35	−11	3432
Left parahippocampal gyrus/hippocampus	37	−26	−47	−9	1944
Left posterior cingulate	30	−17	−53	15	1632
Left posterior cingulate	30	−15	−59	19	976
Left occipital gyrus	19	−34	−85	26	712
Left posterior cingulate	29	−9	−50	6	480
Right superior frontal gyrus	6	26	5	56	336
Presupplementary motor area	6	8	17	49	304
Left dorsal premotor cortex	6	−18	−2	55	192
Left lingual gyrus	18	−3	−91	−6	120
(c) Nonspatial memory encoding
Presupplementary motor area	6	−2	13	50	2424
Left inferior frontal gyrus	45	−41	27	6	2064
Left fusiform gyrus	37	−42	−64	−12	1952
Left precentral gyrus	6	−44	7	33	1160
Right precentral gyrus	6	44	8	28	1120
Left dorsolateral prefrontal cortex	9/46	−41	24	24	1048
Right middle frontal gyrus	8/9	34	42	36	992
Left parahippocampal gyrus/hippocampus	35	−19	−30	−8	976
Left dorsal premotor cortex	6	−40	2	48	912
Left occipital gyrus	18	−31	−90	10	896
Right parahippocampal gyrus/hippocampus	35	21	−27	−15	880
Right fusiform gyrus	37	40	−51	−20	744
Left precuneus	7	−27	−63	45	576
Right insula	13	43	28	11	336
Right cuneus	17	25	−92	8	120
Left cerebellum		−44	−56	−28	112
(d) Nonspatial memory retrieval
Left insula/ inferior frontal gyrus	13/47	−32	25	−1	1440
Left precuneus	19/7	−28	−63	45	696
Left cingulate gyrus	23	2	−25	30	584
Right insula	13/47	34	21	−9	568
Left posterior cingulate	29	−4	−44	21	464
Left occipital gyrus	18	−29	−92	0	376
Left cuneus	19	−31	−76	36	336
Left precentral gyrus	9	−40	5	30	328
Left parahippocampal gyrus/hippocampus	37	−33	−41	−9	320
Right cuneus	18	10	−76	34	320
Left dorsolateral prefrontal gyrus	46	−43	20	20	272
Right posterior cingulate	30	16	−58	17	240
Presupplementary motor area	6	−5	14	51	232
Left temporo‐parietal junction	39	−43	−67	29	224
Anterior cingulate gyrus	32	6	39	16	160
Anterior cingulate gyrus	32	−1	30	41	136
Medial frontal gyrus	10	23	49	−7	104

Figure 1.

ALE meta‐analysis maps of (A) the encoding phase in spatial navigation (red) or episodic memory tasks (blue), (B) the retrieval phase in spatial navigation (red) or episodic memory tasks (blue), and (C) immediate (green) or delayed (yellow) retrieval in spatial navigation (P < 0.01, corrected for multiple comparisons). [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

During encoding, the activity in spatial navigation tasks and episodic memory tasks was partly overlapping bilaterally in the posterior (caudal) part of the hippocampus, close to the junction of body and tail of the hippocampus (−18, −34, −5; 22, −34, −7). Spatial navigation related activity extended into the posterior parahippocampal and lingual gyrus, whereas episodic memory related activity extended into the hippocampal body and adjacent parahippocampal gyrus (Fig. 1A).

During retrieval, the conjunction revealed concurrence across studies on spatial navigation and episodic memory, with overlap observed in left parahippocampal gyrus (−32, −43, 11) but not in the right hemisphere. Of note, retrieval processes in the episodic memory tasks showed neither concurrence in right hippocampus nor in right parahippocampal gyrus or neighboring brain regions such as lingual or fusiform gyrus. However, retrieval in spatial navigation showed profound concurrence of activation in right hippocampal body and tail and the adjacent parahippocampal gyrus (Fig. 1B).

When computing separate analyses on PET and fMRI studies within the episodic memory encoding and the retrieval analysis (P < 0.05) significant overlap was observed between fMRI and PET studies on episodic memory encoding. The left parahippocampal/posterior hippocampal cluster in the episodic memory retrieval analysis, on the other hand, was mainly driven by the fMRI studies included.

To explore possible dissociations between immediate and delayed (i.e. retrieval phase took place more than a day after the encoding phase) retrieval in spatial navigation, we computed two separate meta‐analyses. Interestingly, the delay until retrieval revealed an anterior–posterior distinction: immediate retrieval was mainly located in regions in the posterior (caudal) part of bilateral parahippocampal gyrus and right hippocampal tail, whereas delayed retrieval was located in more anterior (rostral) regions of right hippocampal body and the left fusiform gyrus (Fig. 1C).

DISCUSSION

Within the scope of this study, we performed several quantitative meta‐analyses to determine commonalities and differential involvement of the hippocampal formation in the encoding and retrieval phase of spatial navigation in contrast to episodic memory tasks. The key findings of this study are threefold: first, during encoding, activity in spatial navigation tasks was located in more posterior parts (bilaterally close to the junction of body and tail of the hippocampus, extending into posterior parahippocampal and lingual gyrus) compared to episodic memory task activation that was located in more anterior parts within the hippocampal formation (hippocampal body and adjacent parahippocampal gyrus). Second, during the retrieval phase, activity in spatial navigation tasks was strongly lateralized to the right hippocampal formation (right hippocampal body and tail and the adjacent parahippocampal gyrus), whereas episodic memory task‐related activity was restricted to the left hemisphere. Third, when dividing the studies on the retrieval phase of spatial navigation into those with immediate retrieval and those with delayed retrieval (more than a day after encoding), an anterior‐posterior distinction was observed. Delayed retrieval relied more strongly on anterior (right hippocampal body and adjacent bilateral parahippocampal gyrus) and immediate retrieval relied more on posterior regions of the hippocampal formation (right hippocampal tail and adjacent bilateral parahippocampal gyrus).

The presented findings on encoding are in line with previous suggestions that spatial navigation is located in more posterior portions of the hippocampus in humans [Maguire et al., 2000, 2006] as well as in primates [Moser and Moser, 1998].

A model conceptualizing the anatomical connectivity of substructures within the hippocampal formation has been derived from evidence across different species [Eichenbaum and Lipton, 2008; Witter et al., 1993]. Basically, it comprises a continuation of the well‐known distinction of a ventral “what” and a dorsal “where” visual pathway [Mishkin and Ungerleider, 1982]. It assumes that the perirhinal cortex receives more input from areas along the ventral visual pathway that are considered important for object recognition, whereas the parahippocampal cortex receives input from areas of the dorsal visual stream considered important for spatial processing. Furthermore in rats and monkeys, the perirhinal cortex tends to project more strongly to the lateral enthorinal area and this in turn to the CA3 hippocampus subfield. In contrast, the parahippocampal or postrhinal cortex tends to project to the medial enthorinal area and in turn to the CA1 subfield of the hippocampus proper [Witter et al., 1993]. Functional imaging studies on humans revealed dissociations of activation in line with this two‐stream account. In one study, changes of the identity of objects activated the perirhinal cortex, whereas changes in the spatial arrangement of objects activated the parahippocampal cortex [Pihlajamäki et al., 2003]. Likewise, another study showed that merely the instruction to remember the object or the place of visually presented objects could alone differentially activate perirhinal or parahippocampal cortex [Buffalo et al., 2006]. Since the current meta‐analysis was based on fMRI and PET data, methods that rely heavily on smoothing, and additional smoothing within the ALE analysis we should not overestimate the resolution and precision of the localization of our results. However, the proposed division of the two‐stream account into the more anterior perirhinal cortex concerned with object recognition and the more posterior parahippocampal cortex concerned with spatial processing is in accordance with our anterior–posterior division of concurrence in studies on the encoding phase of episodic memory (anterior) as opposed to spatial navigation tasks (posterior).

The observed lateralization of hippocampal activation in spatial navigation during retrieval corroborates previous notions of right hemispheric dominance [Burgess, 2002; Postma et al., 2008] and substantiates it by showing significant concurrence across a broad range of studies. This finding is in accordance with studies on patients with posterior cerebral artery strokes that lead to problems with navigation (orienting) particularly when the right hemisphere was affected [Barrash et al., 2000]. Similar patterns of rightward lateralizations in navigation have been reported in rodents [Klur et al., 2009] and avians [Kahn and Bingman, 2004]. A recent article using pattern classification during an object–location incidental learning paradigm showed best classification rates within the right hippocampus [Manelis et al., 2012]. Furthermore, the findings are in line with a recent paper showing that the dorsal hippocampus (posterior in humans) is sufficient to form memory traces of spatial information, but that both dorsal and ventral hippocampus contribute to the retrieval of spatial information in rats [Loureiro et al., 2012]. Albeit in our present findings on retrieval, only an extended area within the hippocampal body and tail was activated, but not the head of the hippocampus.

In the included studies on retrieval in episodic memory, we found no convergence within the hippocampus proper, only in left parahippocampal gyrus. This absence of hippocampal involvement can be interpreted in support of the standard consolidation theory [Squire and Alvarez, 2006], assuming that other extrahippocampal structures suffice to mediate retrieval. In contrast, the multiple trace theory [Nadel and Moscovitch, 1997; Nadel et al., 2000] would have predicted hippocampal involvement during retrieval of episodic information, which we did not observe within the present meta‐analysis.

The observation that spatial navigation retrieval is located more medially in parahippocampal gyrus, whereas the episodic memory retrieval is located in a more lateral region is in accordance with a finding recently reported by Schultz et al. (2012). They found a distinction between spatial and non‐spatial content in a working memory paradigm during retrieval after distraction showing activation in the parahippocampal medial enthorinal pathway for scenes and in the parahippocampal lateral enthorinal pathway for faces.

Previous studies exploring the delay effect in retrieval processes have shown that activity in bilateral hippocampus increases with increasing time until retrieval [Huijbers et al., 2010; Talmi et al., 1988]. Brozinsky et al. (2005) reported increases in bilateral posterior hippocampus and anterior parahippocampus. This finding is not entirely in line with our observation that immediate retrieval is located in more posterior (tail) regions, and delayed retrieval is found in more anterior regions (body) of the hippocampal formation. However, the cluster of concordance of delayed retrieval studies in spatial navigation is indeed located in hippocampus proper, whereas the cluster of immediate retrieval extends into the parahippocampal gyrus. It is important to note that these previous studies focused on delays of up to 2 min, whereas the delayed condition in our meta‐analysis includes studies using delays of over 24 h making a direct comparison difficult.

It is difficult to relate the spatial dissociations observed in the present meta‐analysis to high‐resolution imaging studies investigating activity within hippocampal subfields, because the resolution of standard fMRI studies is low and summarizing activity across studies further enhances blurring. However, we would like to mention that the subfields CA1, CA3, and dentate gyrus have been suggested to support encoding of non‐spatial [Eldridge et al., 2005; Zeineh et al., 2003] as well as spatial associations [Suthana et al., 1992], whereas the subicular cortex has been shown to support retrieval of learned spatial [Suthana et al., 1992] as well as nonspatial associations [Eldridge et al., 2005; Zeineh et al., 2003]. Within the present meta‐analysis we mainly detected dissociations on the level of lateralization or between anterior and posterior portions of the hippocampal formation, whereas hippocampal subfields provide a more fine‐grained discrimination on coronal slices of the hippocampus. Future meta‐analyses should attempt to summarize information across high‐resolution studies that discriminate between hippocampal subfields in order to investigate whether similar dissociations can be observed.

Overall, the results of the present meta‐analyses are consistent with what has been reported about the organization of hippocampal connectivity and anatomy in the previous literature. Within the animal hippocampus literature, which has generally been accepted as consistent across rats, cats, monkeys and humans [Burwell, 2000] CA1 in the tail of hippocampus has been shown to contain a high density and selectivity of place cells coding spatial location [Jung et al., 1994; Muller et al., 1996]. Furthermore, the subicular complex, at the posterior end of the hippocampus, contains most so‐called head direction cells coding head position in space [Taube et al., 2007]. This is in line with our finding that encoding processes in spatial navigation tasks activate the hippocampal tail. Similarly the connectivity patterns of the body and the tail of hippocampus have been shown to differ, suggesting a division within the functional domain [Risold and Swanson, 1997]. Furthermore, the posterior (dorsal) CA1 as well as the dorsal parts of the subicular complex have prominent cortical projections to retrosplenial and anterior cingulate cortices in rats [Cenquizca and Swanson, 2007]. These cortical regions are involved in spatial navigation in rats [Harker and Whishaw, 2004] as well as in humans [Spiers and Maguire, 2003].

There is substantial data supporting that the posterior part of the hippocampus, namely body and tail, are involved in cognitive processing such as memory and navigation, whereas the anterior portion (namely the head) of the hippocampus modulates affective processing [Bannerman et al., 2004; Fanselow and Dong, 2010] in particular of fear and anxiety [Gray and McNaughton, 2000]. This functional subdivision is supported by an anatomical linkage between the amygdala and the head of the hippocampus [Kishi et al., 2006; Pitkanen et al., 2000]. The adjacent parahippocampal gyrus shows a similar functional subdivision [LaBar and Cabeza, 2006]. Within this study we did not find any concurrence within the head of hippocampus. This is on non‐spatial memory in line with the fact that the included studies on memory and spatial navigation did not aim at eliciting emotions.

LIMITATIONS

Although the results presented show considerable differences between brain activity associated with spatial and nonspatial memory within the hippocampal formation, one has to acknowledge that the studies included in the present meta‐analysis are heterogeneous in nature. In particular, the retrieval part of the studies included varies between overt and covert recollection as well as between familiarity and recollection judgments. Future meta‐analyses should attempt to match retrieval processes across the different retrieval domains when a sufficient number of respective studies are published.

Although the present studies are heterogeneous in nature many studies on episodic memory and spatial navigation that did not use subtraction methods, but correlations or pattern classification were not included in the present meta‐analysis. This selection may have potentially biased the results.

Another limitation might be seen in the direct comparison of encoding and retrieval phase that is confounded by factors such as novelty effects that are exclusively associated with the encoding phase and familiarity effects associated with the retrieval phase. However, the main focus of this study was the comparison between spatial navigation and episodic memory related activation in hippocampal formation. When comparing encoding or retrieval phase across domains the confounding factors are present in both.

CONCLUSION

Taken together, although there are multiple accounts that propose a unifying coarse‐scale theoretical framework of hippocampal involvement in spatial navigation and episodic memory, the neural correlates identified within the current quantitative meta‐analysis show considerable regional disparity. Encoding in spatial navigation uses more posterior regions of the hippocampal formation, whereas episodic memory encoding utilizes more anterior regions. Furthermore, spatial navigation retrieval is more strongly lateralized to the right hippocampal formation compared to episodic memory retrieval. Within spatial navigation retrieval, immediate recall requires more posterior regions of the hippocampal formation, while delayed recall is located more anterior. To summarize, the overlap of concurrence within the hippocampal formation is rather limited. This could be interpreted as arguing against unifying accounts of hippocampal function in spatial navigation and episodic memory.