Hippocampal activation and connectivity in the aging brain

Abstract

The hippocampus and underlying cortices are highly susceptible to pathologic change with increasing age. Using an associative face-scene (Face-Place) encoding task designed to target these regions, we investigated activation and connectivity patterns in cognitively normal older adults. Functional MRI scans were collected in 210 older participants (mean age = 76.4 yrs) in the Baltimore Longitudinal Study of Aging (BLSA). Brain activation patterns were examined during encoding of novel Face-Place pairs. Functional connectivity of the hippocampus was also examined during encoding, with seed regions placed along the longitudinal axis in the head, body and tail of the structure. In the temporal lobe, task activation patterns included coverage of the hippocampus and underlying ventral temporal cortices. Extensive activation was also seen in frontal, parietal and occipital lobes of the brain. Functional connectivity analyses during overall encoding showed that the head of the hippocampus was connected to frontal and anterior/middle temporal regions, the body with frontal, widespread temporal and occipital regions, and the tail with posterior temporal and occipital cortical regions. Connectivity limited to encoding of subsequently remembered stimuli showed a similar pattern for the hippocampal body, but differing patterns for the head and tail regions. These results show that the Face-Place task produces activation along the occipitotemporal visual pathway including medial temporal areas. The connectivity results also show that patterns of functional connectivity vary throughout the anterior-posterior extent of the hippocampus during memory encoding. As these patterns include regions vulnerable to pathologic change in early stages of Alzheimer’s disease, continued longitudinal assessment of these individuals can provide valuable information regarding changes in brain-behavior relationships that may occur with advancing age and the onset of cognitive decline.

Introduction

Historically, the hippocampus was thought to be one of the key seats of memory in the brain (Andersen et al. 2007; Scoville and Milner 1957). Over the years, further study of the temporal lobe has shown that ventromedial temporal cortical areas are also critically important for our ability to initially process incoming information to be remembered, and to later recall or recognize previously learned material. These cortical regions are particularly suited to process visual information as they are components of the ventral visual pathway (Ungerleider and Mishkin 1982), and are thought to be involved in initial stimulus feature processing, contextual binding and discerning familiarity (Eichenbaum et al. 2012; Nadel and Peterson 2013). Thus, the medial temporal lobe memory system is comprised of the hippocampus and underlying ventromedial parahippocampal, perirhinal and entorhinal cortices, and these regions collectively work together to perform the processes required to remember facts and events.

These regions are important not only for the processes they perform, but also because they are sites of early pathologic change in aging and Alzheimer’s disease (AD). Autopsy studies have shown that abnormal accumulation of tau protein in the form of neurofibrillary tangles generally begins in the entorhinal cortex and then progresses to the hippocampus and medial temporal cortex (Braak and Braak 1995; Braak and Braak 1997). Amyloid deposition in the form of diffuse plaques also begins early in the temporal pole, affecting the entorhinal cortex and progressing to the hippocampus and medial temporal cortex (Braak and Braak 1991; Thal et al. 2002). To understand the processes leading to cognitive decline and impairment in aging and preclinical stages of AD, it is important to study the progression of functional change not only within the hippocampus, but in other temporal lobe regions as well.

The hippocampus itself has an additional feature to consider when assessing its role in cognitive decline. The hippocampus, composed of CA1–4 subfields, in conjunction with the dentate gyrus and subiculum, runs along the floor of the inferior horn of the lateral ventricle and is approximately 7 cm long in adults (Narasinga Rao et al. 2012). While prior studies have examined contributions of the various hippocampal subfields to memory (Carr et al. 2010; Kesner and Rolls 2015), it is also important to consider the role of the hippocampus in relation to its longitudinal axis. In studies of hippocampal volume change across the lifespan, there is evidence of differential volume loss along the longitudinal axis with age in both anterior (Ta et al. 2012) and middle (Malykhin et al. 2017) regions. There is also evidence of functional specialization along the longitudinal axis, with differential specialization based on the type of information being processed (e.g. emotional and motivational information in anterior regions, and neutral, spatial and visual information in posterior regions) and the information processes themselves (e.g. encoding and associative processes in anterior regions, and retrieval in posterior regions) (Collin et al. 2015; Poppenk et al. 2013; Shafer and Dolcos 2014; Ta et al. 2012). Finally, there is autopsy evidence of a posterior-to-anterior pathologic gradient in both aging and Alzheimer’s disease (AD). When compared with the anterior hippocampus, the posterior region exhibits greater neurofibrillary tangle accumulation and a lower cell number in the aging brain, and these findings are more pronounced in those with AD (Ball 1977).

Many studies have examined hippocampal function during episodic memory encoding in humans. Functional neuroimaging studies have investigated activation during the encoding of objects (Johnson et al. 2006; Mandzia et al. 2009), words (Daselaar et al. 2003; Kircher et al. 2008), faces (Dennis et al. 2008; Grady et al. 1995), shapes (Gron et al. 2003; Strange et al. 2005), and complex visual scenes (Binder et al. 2005; Park et al. 2003; Stern et al. 1996). Associative encoding tasks have also been used to investigate activation during face/name associations (Celone et al. 2006; Dickerson et al. 2005), and object/scene associations (Awipi and Davachi 2008). These studies all find hippocampal activation during episodic encoding, yet the regional extent of activation of the hippocampus varies according to the type of stimulus used. For example, more anterior activation is observed for faces, and more middle and posterior activation is observed with complex scenes. These regional patterns appear to be preserved in normal aging, although studies have also shown that hippocampal activation levels during encoding often increase in early mild cognitive impairment (MCI) and decrease with the onset of (AD) (Celone et al. 2006; Dickerson et al. 2005) relative to cognitively normal individuals. Thus, the extent of activation of the hippocampus and underlying cortices in these studies is often limited to a portion of hippocampus and does not extend along the longitudinal axis.

Here, we have designed an activation task based on the Face-Name task that was developed by Sperling and colleagues (Sperling et al. 2001). The Face-Name task is an associative memory task designed to activate anterior regions of the hippocampus and is used extensively in studies of aging (Miller et al. 2008; Putcha et al. 2011; Rentz et al. 2011) and cognitive impairment (Atri et al. 2011; Jurick et al. 2017; Pihlajamaki et al. 2011). In our Face-Place task, we substituted novel scenes for names. Novel scenes were chosen because scene tasks were among the first to demonstrate hippocampal activation with functional MRI (Stern et al. 1996), complex visual scenes are known to activate the occipitotemporal visual pathway including ventral temporal cortical regions (Constable et al. 2000; Fransson et al. 2001; Kohler et al. 2002), and novel complex scenes are known to preferentially activate more posterior aspects of the hippocampus (Binder et al. 2005; Menon et al. 2005; Menon et al. 2000). Based on these previous findings, we theorized that the use of novel scenes and faces in an associative memory encoding task would produce activation along the anterior-posterior axis of the hippocampus, as well as activation of the underlying ventromedial temporal cortices.

Using data collected from older participants from the Baltimore Longitudinal Study of Aging (BLSA), we investigated cross-sectional brain activation patterns resulting from the Face-Place encoding task. We assessed the overall encoding pattern resulting from a comparison of novel Face-Place encoding relative to a repeated, previous learned Face-Place pair, and the more memory-specific pattern limited to encoding trials that were subsequently remembered in the recognition phase of the task. We further investigated functional connectivity patterns during encoding of the three main hippocampal subdivisions (head, body and tail) in cognitively normal older adults. Our main goal was to determine if this task produced activation along the entire anterior-posterior axis of the hippocampus. We also investigated whether activation and functional connectivity patterns involved regions known to be susceptible to disruption or dysfunction with age-related dementia (Dickerson and Eichenbaum 2010; Jagust 2013), and could therefore serve as a targeted tool for continued assessment of regional brain function with advancing age and the onset of cognitive decline.

Methods

Participants

Data from 210 older participants (99 males; mean age at baseline 76.45 (8.63 SD)) in the neuroimaging substudy (Resnick et al. 2000) of the BLSA (Shock et al. 1984) met criteria for inclusion in these analyses (Table 1). 301 participants completed the Face-Place task, however, 91 individuals were excluded based on a clinical diagnosis of cognitive impairment (n = 12), significant health conditions that could affect brain structure or function (i.e. stroke, closed head injury, brain surgery, malignant cancer, meningiomas and cysts with brain tissue displacement, seizure and bipolar disorders; n = 27); fMRI signal abnormalities (frontal pole signal dropout (n = 6); medial frontal and/or midline signal dropout due to falx cerebri calcifications (n = 24)); poor normalization due to image preprocessing brain extraction errors (n = 3); significant scan motion with < 70% viable fMRI scan frames (n = 13); and significant brain atrophy with ventricular volume > 2 SD from the sample mean (n = 6). The remaining 210 participants are considered cognitively normal agers.

Table 1.

Participant demographics

Demographic data
Participants (n)	210
Age (Mean (SD))	76.45 (8.63)
Age range	60.30–98.80
Sex (male)	99
Education (Mean (SD))	17.05 (2.53)
MMSE (Mean (SD))*	28.43 (1.27)

^*Mini-Mental State Examination (MMSE) is the score out of a possible 30 points

This study was approved by the local Institutional Review Board. All participants provided written informed consent prior to the assessment.

MRI scanning

Scanning was performed on a Philips Achieva 3 T scanner. Participants underwent a structural MPRAGE T1-weighted scan using the following parameters: TR = 6.8 s, TE = 3.2 ms, 8° flip angle, 256 × 256 matrix, 1 × 1 mm² voxel size, 1.2 mm slice thickness, and 170 sagittal slices. A double echo T2-weighted scan was also collected for coregistration: TR = 3 s; TE = 8 ms, 90° flip angle, 240 × 210 matrix, voxel size, 0.94 × 0.94mm² voxel size, 3 mm slice thickness, 50 axial slices. Functional EPI scans were acquired with a TR = 2 s, TE = 30 ms, 75° flip angle, 128 × 128 matrix, 3mm³ voxel acquisition with a 1 mm slice gap, and 37 axial slices.

fMRI task design

The Face-Place (FP) task was administered in two parts. Encoding of FP pairs was performed during scanning, and a recognition phase was administered outside the scanner following the scan session. The task phases were designed and displayed using E-Prime 2.0 software (Psychology Software Tools, Inc.).

A mixed block/event design was used for the encoding phase, and 2 encoding runs were administered consecutively during the session. During each run, participants were shown 4 blocks of novel stimuli, and 2 blocks of repeated stimuli. Each block was 40 seconds in duration, and the blocks were separated by 25 seconds of a fixation condition. The repeated blocks were the first and last blocks of the run, and a 5 second fixation period started and ended the run for use as the implicit baseline in the analysis models. During each block, 7 stimuli were presented, with a display time of 4 seconds each. The stimuli were presented with a randomized, jittered ISI of 1–4 seconds. Each run lasted 375 seconds (6 minutes, 15 seconds) and a total of 188 brain volumes were collected (Fig. 1).

Fig. 1. Face-Place Task.

Top row illustrates the stimuli used in the encoding and recognition phases of the task. During the encoding phase, participants are shown a face at the bottom of a complex visual scene and asked to make a yes/no subjective decision as to whether the face is a good match for the place. The face-place stimuli are shown in blocks of either novel face-place pairs (Novel) or a single previously learned repeated face-place pair (Repeated). During recognition, participants are shown the scene with the original face seen during encoding, along with a new distractor face. They must indicate which face (left or right) was previously paired with that scene. Bottom row illustrates the fMRI run design. A mixed block/event design was used for the encoding phase. Each stimulus block was 40 seconds in duration, and the blocks were separated by 25 seconds of a fixation condition. The repeated blocks were the first and last blocks of the run, and a 5 second fixation period started and ended the run for use as the implicit baseline in the analysis models. During each block, 7 stimuli were presented (red lines), with a display time of 4 seconds each. The stimuli were presented with a randomized, jittered ISI of 1–4 seconds. Two encoding runs were administered.

During the encoding phase, participants were shown novel face-place pairs and a single repeated face-place pair. In the novel condition, a novel face was shown at the bottom of a novel complex visual scene (Fig. 1) and participants were asked to make a yes/no subjective decision as to whether the face was a good match for the place (i.e. decide if they could imagine seeing that person in that place), and instructed to remember each pair for later in the recognition phase of the task. In the repeated condition, the same FP pair was repeatedly shown, and participants were instructed to indicate their answer each time it appeared. The participants responded using response buttons placed in their right and left hands; a right button press indicated ‘yes,’ and a left button press indicated ‘no’. Participants were trained on the encoding phase before scanning and were exposed to the repeated FP stimulus pair at that time. If required, vision was corrected during scanning with MR-compatible glasses.

Face stimuli were selected from research-based databases of faces with neutral expressions (Park Aging Mind Laboratory Face Database (Minear and Park 2004); Psychological Image Collection at Stirling (PICS); Massachusetts Institute of Technology and the Center for Biological and Computational Learning). Diverse faces were chosen to offset the cultural bias that exists in facial recognition (O’Bryant and McCaffrey 2006). Scene stimuli were collected from websites of scenic images available for public use (Photobucket.com, Morguefile.com), and included both indoor and outdoor scenes from cities, suburbs and the countryside. The scenes, which did not include written language or other people in the image, were randomly paired with the faces.

In the recognition phase, each scene shown during encoding was presented with 2 faces at the bottom: the original face seen during encoding and a novel face. The participants were asked to select the face that was previously shown with that place, by indicating if their chosen face was displayed on the right or left side of the screen. The recognition phase of the task was self-paced, with the next stimulus shown once a response was made, or after a maximum of 4 seconds.

fMRI data analysis

Preprocessing

Slice timing, realignment, and within subject co-registration between functional and structural scans were performed using SPM12 (Wellcome Department of Cognitive Neurology, London, England) in MATLAB 2016a (The MathWorks Inc. Natick, Massachusetts). For each subject, the EPI scans were slice time corrected and realigned to the first volume collected. Functional scans were co-registered to the MPRAGE by registering the mean EPI image to their T2-weighted image, and the T2-weighted image to their MPAGE. These transformations were concatenated and applied to the realigned EPI scans. Skull stripping of the MPRAGE images was performed using FSL 5.0.9 (Jenkinson et al. 2012), and normalization to MNI152 space was performed using ANTS 2.1.0 (Avants et al. 2008; Klein et al. 2009) with an intermediate registration step to a BLSA template made from 100 randomly selected participants between 60 and 80 years old. The transformations from subject MPRAGE-to-BLSA template and BLSA template-to-MNI152 were concatenated and applied to the MPRAGE co-registered EPI images. Lastly, the images were smoothed using a 6 mm³ kernel. Motion outliers were identified using the FSL motion outlier tool (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLMotionOutliers) and a framewise displacement (fd) threshold ≥ 0.5 mm on the raw functional data, and spatial root mean square of the data after temporal differencing (DVARs) with a threshold of DVARs ≥75th percentile plus 1.5 times the interquartile range (P75 + 1.5*IQR) on the normalized, smoothed data. Frames identified as outliers were statistically removed (weighted zero in the GLM) from the first-level analysis. Participants with > 70% of viable frames were included in the analyses (n = 210). There was no relationship between the number of outlier frames identified and age, r = .086, p = 0.213, 95% CI [−0. 05,0.219].

Activation

The encoding task regressors were based on performance during the recognition phase of the task. ‘Hits’ were defined as novel encoding pairs where the correct face was selected during recognition, and ‘misses’ were defined as novel encoding pairs where the incorrect face was selected during recognition. The data were examined in two ways: activation resulting across all encoding trials regardless of recognition outcome (All Encoding), and activation limited to trials where the correct face was chosen during the recognition phase (Correct Encoding). The data were analyzed using event-related design timing parameters in SPM12. In the first level single-subject analyses, the magnitude and temporal derivatives for 4 task regressors were included in the model (Novel hits, Novel misses, Novel no response, Repeated), 6 motion regressors (3 translation+3 rotation), and regressors indicating frames to exclude based on the motion analysis. Contrast maps were created to represent the All Encoding task activation pattern, controlling for basic visual perception, motor responses, and yes/no decision making (Novel hits + Novel misses + Novel no response trials > Repeated trials). For the Correct Encoding analysis, contrast maps were created to represent activation patterns limited to trials subsequently remembered during the recognition (Novel hits only > Repeated).

To examine All Encoding and Correct Encoding, a second level group t-test was used, adjusting for age and sex. Contrasts of interests were corrected using voxel-wise FWE correction of p<0.05 in SPM12 which corresponds to a t-value>5.01, and >50 voxels spatial extent. Including education (years) and hippocampal volume in the model were also examined. Using the second level multiple regression model, age effects on the activation patterns were examined, controlling for sex. By using a mask of the activated regions, age was correlated with the regional activation pattern to determine voxels that showed a relationship between age (years) and activation levels (significance threshold p<0.005, >50 voxels spatial extent).

Hippocampal segmentation and head, body, and tail ROIs

The whole hippocampus was defined based on a robust automatic HC segmentation approach that uses several registration algorithms and atlases to provide consensus labelling via a label fusion technique. The MUSE segmentation uses 35 atlases from the OASIS dataset where each atlas has been semi-manually segmented into ROIs by an expert using the Neuromorphometircs, Inc. protocol (Doshi et al. 2016). For each individual’s MPRAGE image, 11 atlases that required the least amount of warping were selected and used for label fusion. Once the HC was segmented for each individual in their native space, the head, body, and tail masks were created by first removing the first and last 3 slices to eliminate areas at the extreme ends where increased uncertainty exist. Next, moving anterior-to-posterior in the y-plane, the first 12 mm were assigned as head, the middle 12 mm as body and last 12 mm as tail. This created rectangular ROIs that were 12 mm in length with the dorsal/ventral and medial/lateral extent defined by each individual’s segmentation boundaries. These ROIs were then registered to normalized space by applying the MPRAGE to MNI transformations acquired during preprocessing of the functional data (see Supplemental Figure 1 for an example).

Volume data for the whole hippocampus was extracted from the MUSE segmentation for each participant. Volume data for the head, body and tail regions were obtained by extracting the volume for each seed region ROI mask in subject native space (sum of all voxels within the mask multiplied by the voxel size (1mm³)). All volumes were ICV corrected.

Functional connectivity

To determine functional connectivity of the hippocampus during encoding, differences in voxel-wise functional connectivity for hippocampal head, body, and tail were then examined by implementing Psychophysiological Interactions (PPI) in SPM12. The average time series within each normalized head, body, tail ROI mask was extracted and multiplied with a task regressor representing either the all Novel > Repeated contrast (All Encoding) or the Novel hits > Repeated contrast (Correct Encoding) to generate the PPI interaction term. General task and physiological correlation effects were controlled for by entering the main effects of task and the mean ROI time series as co-variates of no-interest when examining task-dependent connectivity for the ROIs of interest. When using this approach, the contrasts of interest (the PPI interaction term) identified areas where the relationship between the hippocampal seed time series and the time series of other voxels in the brain significantly differed as a function of task condition. Thus, the resultant interaction term identifies voxels that show greater coordinated communication with the hippocampus during the encoding of novel stimuli relative to the repeated stimulus. Statistical significance for both All Encoding and Correct Encoding models was assessed using a threshold of p < 0.005, and spatial extent of > 50 voxels.

For Correct Encoding, a follow-up analysis was performed to determine if the regional connectivity or correlation patterns were unique to each seed region. In this analysis, the regional contrast estimates (Beta weights) were extracted from a 6 mm spherical region centered in the local maxima of each area exhibiting significant connectivity with a seed region. To determine if correlation patterns were significantly different between the head, body and tail regions, the contrast estimates from local maxima associated with the head were compared with contrast estimates from the same regions in the body and tail analyses, contrast estimates from local maxima associated with the body were compared with contrast estimates from the same regions in the head and tail analyses, and contrast estimates from local maxima associated with the tail were compared with contrast estimates from the head and body. Paired-sample t-tests were used to determine significant differences in contrast estimates.

Results

Behavioral performance

The encoding phase was performed during scanning, and the recognition phase of the task was administered outside the scanner following the scan session. During the forced choice recognition phase, participants averaged 65.56 (12.12 SD) % correct responses (hits), 32.15 (12.26 SD) % incorrect responses (misses), and 2.33 (4.55 SD) % no responses.

fMRI task activation

Task activation patterns were generated for two contrasts. The All Encoding contrast (all Novel > Repeated) illustrates the overall activation pattern during the encoding phase of the task, and the Correct Encoding contrast (Novel hits > Repeated) illustrates the activation pattern specific to those trials subsequently remembered during recognition (significance threshold of p=0.05 FWE correction with >50 voxels spatial extent; Fig. 2).

Fig. 2. Task Activation Patterns.

Group-level activation patterns are shown for the All Encoding (all Novel > Repeated) and Correct Encoding (Novel hits > Repeated) contrasts. Top row illustrates the slices used in the activation plots. Both contrasts exhibit extensive activation of the hippocampus and underlying ventromedial temporal cortices

All Encoding

The All Encoding (all Novel > Repeated) contrast, adjusted for age and sex, resulted in widespread activation of frontal, temporal, parietal and occipital regions of the brain. In the frontal lobe, activation was seen in the inferior (BA 44/45/46), middle (BA 6/9/46/47), medial (BA 8/9/10), superior (BA 8) and orbitofrontal (BA 11) cortex, and anterior cingulate (BA 25/32) and anterior insular regions. In the temporal lobe, activation of the hippocampus and underlying cortices, including entorhinal (BA 28/34), perirhinal (BA 35), parahippocampal (BA 36), posterior fusiform (BA 37) and lingual (BA 18) gyri, was observed. The temporal pole (BA 20/38) was also activated by the task. Activation was also seen in the visual association cortices (BA 18/19) and parietal (BA 7/40) regions, including the posterior cingulate (BA 30) and precuneus (BA 7). The cerebellum was activated, and subcortical activation was seen in the thalamus, globus pallidus putamen, and brainstem (Table 2). Including education (years) or whole hippocampal volume in the model did not change the results.

Table 2.

Regional task activation

Region	Hemisphere	MNI coordinate			All encode	Correct encode
		x	y	z	t value	t value
Orbitofrontal Cortex (11)	L	−32	38	−14	10.14	12.45
Orbitofrontal Cortex (11)	R	34	38	−14	9.58	12.66
Med Frontal Gyrus (10)	L	−4	68	24	6.01	7.11
Med Frontal Gyrus (8)	L	−2	18	52	18.13	18.78
Med Frontal Gyrus (9)	L	−12	62	32	5.75	5.89
Inf Frontal Gyrus (44)	L	−36	8	30	16.74	16.92
Inf Frontal Gyrus (44)	R	50	16	30	8.66	8.90
Inf Frontal Gyrus (45)	L	−38	26	22	10.87	11.73
Inf Frontal Gyrus (45)	R	40	30	22	10.14	10.18
Inf Frontal Gyrus (46)	L	−48	34	22	12.59	12.35
Mid Frontal Gyrus (46)	R	54	38	16	10.18	10.63
Mid Frontal Gyrus (6)	L	−26	−6	52	9.41	7.84
Mid Frontal Gyrus (9)	R	56	12	42	8.08	7.69
Sup Frontal Gyrus (8)	L	−4	58	44	7.14	7.67
Ant Cingulate Cortex (25)	R	6	4	−12	7.21	8.68
Ant Cingulate Cortex (32)	R	6	22	−10	6.95	9.36
Ant Cingulate Cortex (32)	L	−10	24	32	8.47	9.23
Precentral Gyrus (4)	R	30	−4	50	9.79	9.26
Postcentral Gyrus (3)	R	56	−12	50	10.15	10.25
Insula	L	−36	20	2	9.27	8.78
Insula	R	32	26	6	10.41	10.24
Temporal Pole (20)	R	32	−8	−34	7.95	9.70
Temporal Pole (20)	L	−30	−6	−32	8.84	10.89
Temporal Pole (38)	L	−44	14	−34	7.77	9.07
Entorhinal Cortex (28)	L	−16	−8	−18	10.31	11.54
Entorhinal Cortex (34)	R	20	−4	−16	11.83	11.76
Perirhinal Cortex (35)	L	−38	−14	−26	9.13	11.60
Hippocampus	R	24	−18	−18	11.10	13.67
Hippocampus	L	−20	−28	0	8.11	7.25
Parahippocampal Gyrus (36)	R	32	−36	−20	23.84	27.41
Parahippocampal Gyrus (36)	L	−26	−36	−18	21.77	27.26
Fusiform Gyrus (37)	L	−30	−48	−14	24.10	24.89
Fusiform Gyrus (19)	R	30	−76	−14	24.58	24.37
Lingual Gyrus (18)	R	16	−90	−10	16.96	18.08
Lingual Gyrus (18)	L	−8	−74	−2	9.59	7.72
Inf Occipital Gyrus (18)	L	−42	−84	−4	18.33	20.10
Inf Occipital Gyrus (18)	R	32	−88	0	19.27	19.57
Mid Occipital Gyrus (19)	R	34	−76	20	21.99	24.34
Mid Occipital Gyrus (19)	L	−34	−84	22	23.44	24.91
Sup Occipital Gyrus (19)	L	−26	−80	40	15.47	16.89
Sup Occipital Gyrus (19)	R	22	−70	42	9.19	10.18
Post Cingulate Cortex (30)	L	−6	−52	12	15.51	16.34
Post Cingulate Cortex (30)	R	8	−52	12	15.46	16.83
Precuneus (7)	R	22	−60	50	11.90	13.04
Precuneus (7)	L	−20	−62	50	12.48	13.91
Cuneus (18)	L	−6	−82	8	8.84	7.68
Cuneus (18)	R	18	−94	10	15.96	14.14
Inf Parietal Cortex (40)	L	−42	−36	50	6.67	n/a
Sup Parietal Cortex (7)	L	−28	−50	48	10.39	12.10
Sup Parietal Cortex (7)	R	28	−46	52	8.15	8.95
Brainstem	R	8	−28	−6	15.20	18.64
Brainstem	L	−6	−28	−4	12.90	17.42
Globus Pallidus	L	−14	0	2	7.13	7.53
Globus Pallidus	R	12	4	8	7.07	6.80
Putamen	R	30	−16	0	9.34	8.26
Putamen	L	−28	−18	4	6.42	n/a
Thalamus	L	−22	−30	2	8.36	9.57
Thalamus	R	14	−12	4	8.68	8.90
Cerebellum	R	8	−74	−20	16.79	15.45
Cerebellum	L	−6	−78	−38	17.46	17.48

Local maxima of regional activation in the All Encoding contrast, and the significance of these regionmaxima observed in the Correct Encoding contrast. MNI coordinates are shown and Brodmann Areas are listed in parentheses. The t value indicates the significance of the regional activation

Age effects were also examined to determine if activation within the various brain regions varied as a function of older age. For All Encoding, relatively modest areas of reduced activation were seen with greater age, with the largest region observed in the cerebellum. These regions included the cortical areas within the middle frontal gyrus (BA 9), the middle occipital gyrus (BA 18), and the cerebellum and putamen. No areas exhibited increased activation in relation to age (Table 3).

Table 3.

Age effects on brain activation

Region	Hemisphere	MNI coordinate			t value	voxels
		x	y	z
All Encoding Regions
Mid Frontal Gyrus (9)	L	−34	6	38	3.79	50
Inf Occipital Gyrus (18)	L	−36	−78	−4	4.50	314
Cerebellum	L	−22	−46	−18	4.04	154
Cerebellum	L	24	−86	−20	3.53	50
Cerebellum	L	−2	−60	−34	4.59	861
Putamen	R	28	8	−2	4.10	95
Correct Encoding Regions
Temporal Pole (38)	L	−36	26	−23	3.14	57
Hippocampus	R	24	−28	−4	3.43	55
Inf Occipital Gyrus (18)	L	−40	−86	−2	419	172
Cerebellum	R	6	−60	−30	3.79	612
Cerebellum	L	−24	−46	−18	4.20	246

Regions where older age was correlated with decreased activation levels in the All Encoding and Correct Encoding contrasts. The t value indicates the significance of the age-activity correlation. MNI coordinates are shown and Brodmann Areas are listed in parentheses

Correct Encoding

Activation of frontal, temporal and occipital lobes remained when assessing the pattern resulting from the Correct Encoding (Novel hits > Repeated) contrast, adjusting for age and sex. Of the regions observed in the All Encoding contrast, only the inferior parietal cortex (BA 40) and the putamen were not observed in the Correct Encoding contrast. The results are shown in Table 2 and Fig. 2. Including education (years) or whole hippocampal volume in the model did not change the results.

Age effects were also examined. For Correct Encoding, lower activation with advancing age was seen in the temporal pole and right hippocampus (Table 3).

Overall, the activation results show markedly similar activation patterns for the All Encoding and Correct Encoding contrasts, and both patterns involve large regions of the frontal, temporal and occipital lobes of the brain. Both contrasts also yield relatively complete involvement of the hippocampus and underlying ventral cortical regions, even when controlling for hippocampal volume. Age-related reductions in activation were seen in middle frontal, occipital and cerebellar regions during All Encoding, and in temporal pole and hippocampus during Correct Encoding.

fMRI functional connectivity

To determine functionally connected regions of the hippocampus during encoding, connectivity patterns were determined for the head, body and tail of the hippocampus for the All Encoding and Correct Encoding contrasts. Each hemisphere was assessed separately. The connectivity patterns are shown in Fig. 3.

Fig. 3. Hippocampal Connectivity Patterns.

The Top Rows show functionally connected areas of the three bilateral hippocampal seed regions (p < 0.005, 50 voxels) for the All and Correct Encoding contrasts mapped together on a single brain. These group-level results are shown on sagittal slices beginning in the left hemisphere and progressing to the right. The Bottom Rows illustrate an example seed placement in a single participant and group-level glass brain projections highlighting the combined left and right hemisphere regional connectivity territories of the tail, body and head of the hippocampus when all functionally connected regions are collapsed into a single 2-dimensional plane. The colored shadings are visual guides illustrating the general pattern of regional cortical and subcortical areas of connectivity associated with each seed region. Specific hemispheric connectivity details for each seed region are listed in Tables 4 and 5

All Encoding

The right and left hippocampal heads were functionally connected to temporal and cerebellar regions. Both heads showed connectivity with the temporal pole (BA 38) and inferior temporal gyrus (BA 20). The left head showed additional connectivity with the middle (BA 9) and inferior frontal gyrus (BA 44), and the right head showed additional connectivity with the medial frontal cortex (BA 6). All correlations were significant at p < 0.001, cluster size range 50–220 voxels (Table 4).

Table 4.

Hippocampal connectivity during All Encoding

Region	Hemisphere	MNI coordinate			t value	voxels
		x	y	z
R Head Seed
Med Frontal Cortex (6)	L	−6	14	60	3.99	51
Temporal Pole (38)	R	30	−4	−48	4.28	64
Temporal Pole (38)x	L	−40	22	−38	3.92	64
Inf Temporal Gyrus (20)	L	−34	−8	−46	4.23	50
Fusiform Gyrus (18)x	R	6	−94	−16	3.72	52
Cerebellum	R	30	−40	−54	3.93	143
L Head Seed
Mid Frontal Gyrus (9)	R	44	44	24	3.39	67
Inf Frontal Gyrus (44)	R	50	4	22	3.40	55
Temporal Pole (38)	L	−48	16	−38	3.64	88
Inf Temporal Gyrus (20)	R	30	−4	−48	3.91	72
Cerebellum	R	20	−76	−44	3.47	53
Cerebellum	L	−18	−74	−46	3.66	214
Cerebellum	R	14	−60	−48	3.42	56
R Body Seed
Orbitofrontal Gyrus (11)	R	26	24	−22	3.62	196
Med Frontal Gyrus (6)	L	−6	14	60	4.52	394+
Med Frontal Gyrus (6)	R	8	26	58	4.12	394+
Sup Frontal Gyrus (6)	R	50	0	58	3.98	120
Sup Frontal Gyrus (6)	R	32	30	54	3.42	89
Temporal Pole (38)	L	−42	14	−42	4.49	199
Parahippocampal Gyrus (36)	L	−38	−20	−16	2.98	68
Inf Temporal Gyrus (20)	R	36	−6	−42	3.47	82
Inf Temporal Gyrus (20)x	L	−24	−12	−40	3.60	57
Fusiform Gyrus (18)	L	−10	−102	−18	3.51	56
Inf Occipital Gyrus (18)	R	36	−6	−42	3.47	82
Cerebellum	R	58	−40	−28	3.26	57
Cerebellum	L	−20	−72	−34	3.40	88
Cerebellum	L	−44	−60	−54	3.78	121
Putamen	R	22	8	12	4.18	111
L Body Seed
Mid Frontal Gyrus (9)	R	46	8	34	3.16	64
Inf Temporal Gyrus (37)	R	44	−62	−8	3.52	110
Fusiform Gyrus (18)	L	−26	−86	−14	3.62	823
Lingual (18)	R	4	−70	−6	3.36	54
Mid Occipital Gyrus (19)	R	36	−86	8	3.76	151
Mid Occipital Gyrus (19)	R	48	−76	−8	3.33	75
Inf Occipital Gyrus (18)	L	−40	−78	−6	3.69	823
Cerebellum	L	−10	−74	−34	3.85	103
Putamen	R	22	8	12	3.80	64
R Tail Seed
Nonsignificant
L Tail Seed
Fusiform Gyrus (18)	L	−38	−92	−12	3.60	60
Mid Occipital Gyrus (19)	L	−32	−88	4	3.17	102
Putamen	R	24	10	10	3.88	66

Regions showing functional connectivity with the hippocampal seed regions in the All Encoding contrast. MNI coordinates are shown and Brodmann Areas are listed in parentheses.

⁺indicates regions contained within the same cluster.

^xregions that did not survive hippocampal seed region volume correction

Both the right and left bodies showed connectivity with regions in the frontal, temporal and occipital lobes, and with the cerebellum and putamen. The right body demonstrated the largest connectivity pattern with frontal regions, including orbito- (BA 11), medial (BA 6) and superior (BA 6) frontal areas. Both bodies also showed connectivity with inferior temporal (BA 20 with the right, BA 37 with the left) and fusiform (BA 18) regions, and the right body showed additional connectivity with the temporal pole (BA 38) and parahippocampal gyrus (BA 36) of the left hemisphere. The left body showed the largest connectivity pattern with occipital regions, including middle (BA 19) and inferior (BA 18) association areas. Both bodies exhibited connectivity with the cerebellum and with the right putamen. All correlations were significant at p ≤0.001, cluster size range 58–1661 voxels.

The hippocampal tails showed the fewest functionally connected regions, with only the left tail pattern reaching significance. The left tail showed connectivity with fusiform (BA 18, 20) and middle occipital (BA 19) regions, and with the cerebellum. All correlations were significant at p ≤0.001, cluster size range 53–108 voxels.

When controlling for volume in the analyses, the right head connectivity with the right fusiform gyrus and left temporal pole, and the right body connectivity with the left inferior temporal gyrus did not survive correction for hippocampal seed region volume. Age effects on the functional connectivity patterns were also examined. No significant age-related differences in connectivity were observed for the patterns related to the head, body or tail.

Correct Encoding

The right and left hippocampal heads were functionally connected to the lingual gyrus (BA 18) and cerebellum. The right head was additionally connected to the insula and inferior temporal gyrus (BA 20). The left head was connected to the temporal pole (BA 38) and cuneus (BA 18). All correlations were significant at p ≤0.001, regional size range 52–636 voxels (Table 5).

Table 5.

Hippocampal connectivity during Correct Encoding

Region	Hemisphere	MNI coordinate			t value	voxels
		x	y	z
R Head Seed
Insula3	R	64	−38	20	4.36	71
Inf Temporal Gyrus (20)	R	30	−4	−48	3.74	57
Lingual Gyrus (18)	L	−4	−98	−4	4.06	636
Cerebellum*2,3	R	30	−38	−54	3.81	133
Cerebellum	L	−44	−70	−16	3.86	80
Brainstem	B	0	−18	−24	3.49	52
L Head Seed
Temporal Pole (38)	R	32	−4	−48	3.60	98
Lingual Gyrus (18)	R	2	−72	−6	3.82	206+
Lingual Gyrus (18)	L	−12	−78	−6	3.18	206+
Cuneus (18)	L	−10	−102	6	3.30	165
Cerebellum2,3	R	50	−42	−42	3.85	88
Cerebellum *x	R	18	−78	−46	3.68	54
R Body Seed
Med Frontal Gyrus (8) 1,3	R	6	32	56	3.80	190
Temporal Pole (38) * 1	L	−40	14	−44	4.52	237
Entorhinal Cortex (28)	R	32	−4	−48	3.31	63
Fusiform Gyrus (19)	R	18	−76	−10	4.04	877
Fusiform Gyrus (19)	L	−36	−78	−10	4.21	231
Inf Occipital Gyrus (18) *1	R	34	−84	−6	3.92	161
Cerebellum	R	22	−72	−56	3.64	97
Cerebellum	L	−30	−78	−10	4.34	231
Putamen+	R	22	10	12	3.64	59
L Body Seed
Fusiform Gyrus (19) x	R	48	−82	−10	3.46	60
Fusiform Gyrus (19)	L	−52	−78	−12	3.75	83
Lingual Gyrus (18)	L	−12	−86	0	4.19	126
Mid Occipital Gyrus (19)	R	38	−86	2	3.24	62
Mid Occipital Gyrus (19)	L	−38	−80	−8	3.57	177
Cerebellum*	L	−16	−76	−48	3.39	99
R Tail Seed
Inf Temporal Gyrus (20)	R	34	−12	−40	3.74	59
Fusiform Gyrus (19) 1	R	48	−78	−10	4.08	151
Fusiform Gyrus (19)	L	−40	−88	−14	3.37	278
Lingual Gyrus (18)	R	4	−74	−4	3.74	124
Lingual Gyrus (18)	L	−2	−96	−12	3.44	237
Inf Occipital Gyrus (18)	R	36	−84	−4	3.57	108
L Tail Seed
Entorhinal Cortex (28) 1,2	R	20	0	−28	4.13	82
Fusiform Gyrus (18) 1,2	R	8	−96	−16	4.36	115
Fusiform Gyrus (19)	L	−34	−80	−12	3.85	230
Sup Occipital Gyrus (19)	R	36	−86	34	3.18	93
Cuneus (18) 1	R	16	−104	14	3.72	80

Regions showing functional connectivity with the hippocampal seed regions in the Correct Encoding contrast. MNI coordinates are shown and Brodmann Areas are listed in parentheses.

⁺indicates regions contained within the same cluster.

^*similar region observed in the All Encoding contrast.

^xregions that did not survive hippocampal seed region volume correction.

¹connectivity of this region is significantly different from the head.

²connectivity is of this region significantly different from the body.

³connectivity of this region is significantly different from the tail

Both the right and left bodies showed connectivity with fusiform (BA 19) and cerebellar regions. The right body was additionally connected to medial frontal (BA 8), temporal pole (BA 38), entorhinal (BA 28) and inferior occipital (BA 18) regions. The left body was also connected to lingual (BA 18) and middle occipital (BA 19) regions. All correlations were significant at p ≤0.001, regional size range 59–877 voxels.

Both right and left hippocampal tails showed connectivity with the fusiform gyrus (BA 19). The right tail was also connected with inferior temporal (BA 20), lingual (BA 18), and inferior occipital (BA 18) regions. The left tail was connected to entorhinal (BA 28), superior occipital (BA 19) and cuneus (BA 18) regions. All correlations were significant at p ≤0.001, regional size range 59–278 voxels.

When controlling for volume, the left head connectivity with the right cerebellum, and the left body connectivity with the right fusiform gyrus did not survive correction for hippocampal seed region volume. No significant age-related differences in connectivity were observed for the patterns related to the head, body or tail.

To determine if these connectivity patterns were unique to each seed region, a follow-up analysis was performed to determine significant differences in regional contrast estimates (Beta weights) of each local maxima across head, body and tail regions in the same hemisphere. In the right hemisphere, the insula and cerebellum correlations with the head were significantly different from the correlations with the tail (p = 0.002, p = 0.025, respectively), and the cerebellum correlation with the head was also different from the correlation with body (p = 0.026). The medial frontal correlation with the right body was significantly different from both the head (p = 0.023) and tail (p = 0.016). The temporal pole and cerebellum correlations with the body were also significantly different from the head (p = 0.051, p = 0.044, respectively). For the right tail, fusiform gyrus correlations with the tail were significantly different from the head (p = 0.032). In the left hemisphere, the right cerebellum correlation with the head was significantly different from both the body (p = 0.018) and tail (p = 0.003). The entorhinal cortex and fusiform gyrus correlations with the tail were significantly different from both the head (p = 0.034, p = 0.014, respectively) and body (p = 0.001, p = 0.002, respectively). The cuneus correlation with the tail was also significantly different from the head (p = 0.040).

Together, these results illustrate that the head, body and tail of the hippocampus exhibit different patterns of functional connectivity during memory encoding when examined individually, and several of these regional patterns are unique to the head, body or tail of the hippocampus. The patterns of connectivity also vary when examining overall encoding or encoding of subsequently remembered stimuli. Although the body of the hippocampus is connected to similar regions for both contrasts, the head and tail regions show varying patterns depending on the contrast examined. During Correct Encoding, the head and tail regions show extensive connectivity with fusiform and lingual regions, and connectivity with the entorhinal cortex that was not observed during All Encoding.

Discussion

In this study, we present initial results for a new associative memory encoding task using faces and novel scene stimuli developed for implementation as an fMRI task. We demonstrate that the task produces activation along the entire anterior-posterior axis of the hippocampus and in underlying ventral temporal cortices in cognitively normal older individuals. Activation outside the temporal lobe was also extensive and included frontal, parietal and occipital regions. This pattern of activation was similar for overall encoding and encoding of correctly remembered stimuli during recognition testing. Analysis of hippocampal connectivity showed different patterns of functional connectivity associated with the head, body and tail of the hippocampus when examined individually, suggesting that this structure has distributed networks of connected regions during associative memory encoding. Although activation patterns were similar for overall encoding and encoding of correct trials, the connectivity patterns varied between the two encoding contrasts, especially in the head and tail regions of the hippocampus.

The encoding activation pattern produced by the Face-Place encoding task involved regions along the occipitotemporal visual pathway that are associated with visual memory processes in both primates and humans (Haxby et al. 1991; Kohler et al. 1995; Ungerleider and Mishkin 1982). These regions included visual association cortices in the occipital lobe, and lingual, fusiform, parahippocampal, perirhinal and entorhinal cortices. This cortical activation pattern confirms previous studies that have shown occipitotemporal activation in visual memory paradigms (Cabeza and Nyberg 2000; Gauthier and Tarr 2016), and demonstrates extensive coverage of these regions. This pattern is likely due to activation of the fusiform face area and other anterior temporal lobe regions associated with face processing (Collins and Olson 2014), and the encoding of novel complex visual scenes which are thought to be processed in these ventral occipitotemporal regions (Menon et al. 2000; Viskontas et al. 2016).

The pattern of activation during all encoding trials was similar to the pattern seen when limited to those trials which were subsequently remembered in the recognition phase of the task. The pattern of frontal, temporal and occipital lobe involvement in correct encoding also confirms previous studies that examined successful encoding of visual stimuli which were later remembered. These studies found that the hippocampus, perirhinal cortex and parahippocampal gyrus all increase activation during successful encoding (Gron et al. 2003; Kircher et al. 2008; Kirwan et al. 2008; Shrager et al. 2008). The prefrontal cortex and occipital regions such as the fusiform gyrus, lingual gyrus and precuneus, also increase activation levels with successful encoding of visually presented stimuli (Gron et al. 2003; Kircher et al. 2008).

The head, body and tail of the hippocampus were also activated by the task. Associative memory tasks, where subjects are required to form a link or connection between two distinct stimuli, generally activate anterior regions (head) of the hippocampus (Chua et al. 2007; Rand-Giovannetti et al. 2006). Memory processes that require the encoding of single items or objects tend to activate more central or middle regions (body) of the hippocampus (Kohler et al. 2005; Strange et al. 2005). Scene encoding, especially complex scenes, activate middle and posterior regions (body and tail) (Constable et al. 2000; Dennis et al. 2008; Hayes et al. 2010), and novelty effects produce activation in more posterior hippocampal regions (body and tail) (Binder et al. 2005; Kohler et al. 2002; Liang et al. 2013). The use of novel faces, novel complex scenes, and the associative nature of our encoding task suggest that these factors contribute to the hippocampal activation pattern observed during encoding of the Face-Place stimuli.

Our results also show that the head, body and tail of the hippocampus display different regional patterns of functional connectivity during memory encoding when each region is examined individually. In overall encoding, the head was functionally connected to the temporal pole, dorsolateral prefrontal cortex and the cerebellum. The body showed the largest regional pattern of connectivity which included the temporal pole, inferior temporal cortex, parahippocampal and fusiform gyri. The body also showed connections with frontal lobe regions including orbitofrontal, superior and medial frontal regions, and with visual association areas, the cerebellum and putamen. Tail connectivity was seen with the fusiform gyrus, visual association areas and putamen. Several of these functionally connected regions were also found to be unique to a given seed region. For example, when examining differences in correlation levels between the head, body and tail of the hippocampus, the medial frontal cortex was found to be uniquely connected to the right body, and the entorhinal cortex was uniquely connected to the left tail.

During correct encoding, functionally connected regions of the hippocampal body were similar to those observed during overall encoding, yet the head and tail patterns varied. For those trials subsequently remembered, the head and tail showed extensive connectivity with lingual, fusiform and visual association areas which are early components of the ventral visual pathway. As activation of these regions are known to be related to recognition success (Gron et al. 2003; Kircher et al. 2008), our finding suggests that heightened functional connectivity of the hippocampus with early ventral visual pathway regions may also be critical for subsequent recognition ability. Together, these results suggest that there are distributed patterns of connectivity associated with the three longitudinal subdivisions of the hippocampus, yet the patterns of head and tail regions vary depending on subsequent recognition outcomes.

Using functional covariance parcellation methods, previous studies have shown that the hippocampus exhibits similar anterior-to-posterior subdivisions for both resting state and task performance conditions (Plachti et al. 2019). The functional connectivity patterns of these divisions, however, vary based on the scanning modality. For example, prefrontal regions negatively correlate with hippocampal activity at rest and positively correlate with hippocampal activity during task performance (Chase et al. 2015). This difference is likely due to the differing cognitive operations performed by the hippocampus during rest and task conditions and the subsequent recruitment of different networks based on these cognitive operations.

The majority of studies examining functional connectivity along the longitudinal axis of the hippocampus have only investigated the differences between anterior and posterior subdivisions (Poppenk et al. 2013). One meta-analysis of 7200 imaging studies (Chase et al. 2015) used an anterior-middle-posterior subdivision scheme similar to the one used here. Our results were markedly similar to the task-based findings of Chase, et. al. (2015) in that the head functionally connects with the lateral prefrontal cortex, and the body and tail connect with fusiform, occipital and putamen regions during memory encoding. By virtue of the distributed patterns of connectivity, these studies suggest that the longitudinal subdivisions of the hippocampus may play distinct roles in memory encoding (Collin et al. 2015).

Another aspect to consider in this study is the older age of the participants. Because we had a wide age range of older adults, we investigated the effects of age on brain activation and on functional connectivity. Some regions did show an effect of age on activation levels, with older age associated with lower activation levels. These included a small region within the dorsolateral prefrontal cortex, the visual association cortices, and the putamen, with the largest effects seen in the cerebellum. When examining correct encoding, the temporal pole and hippocampus also show that older age was associated with lower activation. No significant age effects were seen in the hippocampal functional connectivity. The lack of widespread age effects on activation and the absence of effects on connectivity may be due to the sample population. This sample includes highly educated, cognitively normal individuals who receive regular medical examinations during their BLSA visits. These characteristics are not necessarily representative of the general population and may limit generalizability when compared to other aging studies. The fact that the sample includes individuals up to 99 years of age who are still cognitively normal also may suggest that our participants of older ages are more resilient to age-related cognitive decline and dementia. Although age-related differences in brain activation are most commonly seen when comparing young participants to older participants (Persson and Nyberg 2006), the effects of age on temporal lobe function in our older sample deserves further study. Future studies will assess functional differences between younger and older age, and differences in brain-behavior relationships in our sample.

Investigation of the hippocampal longitudinal axis is potentially important in the study of age-related neurodegenerative disease. Although pathologic studies of the longitudinal axis of the hippocampus are sparse, one early investigation of neurofibrillary tangle accumulation suggests a posterior-to-anterior pathologic gradient occurs in AD (Ball 1977). Together with studies of the functional subdivisions of the hippocampus (Collin et al. 2015; Poppenk et al. 2013; Shafer and Dolcos 2014; Ta et al. 2012), this suggests that there may be selective functional vulnerability along the longitudinal axis with posterior regions showing greater vulnerability early in the disease progression. The pattern of subjective cognitive complaints and insight into memory ability in AD (Rabin et al. 2017; Starkstein 2014) supports this theory. In the early stages of AD, patients are aware of, and often complain of, memory problems. As the disease progresses, memory problems worsen, yet insight into their abilities lessens (Starkstein 2014; Vannini et al. 2017a). The involvement of posterior hippocampal regions in memory retrieval processes and supposed early accumulation of tangles in this region with the onset of disease, supports the early problems with retrieval. Similarly, the functional connectivity patterns observed in this study lend support to the initial awareness of memory problems. In our study of cognitively normal agers, more anterior portions (head/body) of the hippocampus are functionally connected with frontal lobe regions associated with self-awareness (Vannini et al. 2017b; Zamboni et al. 2013). If the pathologic gradient with the onset of disease does indeed travel from posterior to anterior regions of the hippocampus, it is possible that awareness of one’s ability is maintained initially, but worsens with additional accumulation of pathology and the disruption of these anterior hippocampus-frontal lobe connections. Future studies involving the assessment of the hippocampal longitudinal axis with the onset and progression of cognitive impairment could help substantiate this theory.

Together, these results show that activation along the entire longitudinal axis of the hippocampus and along the occipitotemporal visual pathway, including ventromedial temporal cortices, can be achieved in an aging population with an associative encoding task that uses faces and novel scenes. The results also show that the hippocampus is functionally connected to widespread cortical regions of the brain, the pattern of connectivity varies along the longitudinal axis of the structure, and varies as a function of the encoding condition examined. As the brain regions involved in this task are also sites of early pathologic change in age-related disease (Braak and Braak 1997; Thal et al. 2002), continued assessment of these individuals can help clarify changes in temporal lobe brain-behavior relationships that may occur with the onset of cognitive impairment.