The hippocampus is critically involved in episodic memory, yet relatively little is known about how the development of this structure contributes to the development of episodic memory during middle to late childhood. Previous research has inconsistently reported associations between hippocampal volume and episodic memory performance during this period. We argue that this inconsistency may be due to assessing the hippocampus as a whole, and propose to examine associations separately for subregions along the longitudinal axis of the hippocampus. In the present study, we examined age-related differences in volumes of the hippocampal head, body, and tail, and collected episodic memory measures in children ages 8–11 years and young adults (N = 62). We found that adults had a smaller right hippocampal head, larger hippocampal body bilaterally, and smaller right hippocampal tail compared with children. In adults, but not in children, better episodic memory performance was associated with smaller right hippocampal head and larger hippocampal body. In children, but not in adults, better episodic memory was associated with larger left hippocampal tail. Overall, the results suggest that protracted development of hippocampal subregions contribute to age-related differences in episodic memory.
Episodic memory, the capacity to remember past events along with details about the context in which they occurred (e.g., where the event happened, when it occurred, who was there), is a vital cognitive capacity. For example, it supports our ability to form a detailed autobiographical history (Bauer et al. 2007) and helps us understand our past and pre-experience the future (Cooper et al. 2011). Behavioral research indicates age-related improvements extending into adolescence on tasks that require episodic memory (e.g., Billingsley et al. 2002; Brainerd et al. 2004; Ghetti and Angelini 2008). Changes in neural structure and function supporting these improvements are largely uncharted. The emphasis in behavioral research on experimental manipulations affecting encoding and retrieval strategies (Schneider et al. 2004; Ghetti 2008; Schwenck et al. 2009) and other controlled processes (Gathercole 1998; Roebers 2002; Shing et al. 2008) led to the view that the development of cortical regions responsible for these processes may be the primary force driving the development of episodic memory beyond early childhood.
Previous research indeed provides evidence that continued functional development of prefrontal regions contributes to improvement in episodic memory (e.g., Ofen et al. 2007). Furthermore, associations during childhood between decreased cortical thickness in prefrontal regions and measures of memory have been reported (e.g., Sowell et al. 2001). Overall late developing, prefrontally mediated control processes clearly contribute to improvement in episodic memory; this fact, however, does not rule out age-related improvement in binding processes, supported by hippocampal regions, as an additional factor. A wealth of adult neuroimaging research demonstrates that the hippocampus is critically involved in binding features of an event, including features of the context, into a mental representation that integrates them; the resulting “bound” representation is central to the ability to form and retain episodic memories (Diana et al. 2007; Eichenbaum et al. 2007). Thus, protracted development of the hippocampus may contribute to age-related improvements in episodic memory. Until recently, it was thought that the hippocampus completed its development early in postnatal life, due to evidence of important development during infancy (e.g., Seress and Ribak 1995; Gilmore et al. 2012), with the exception of the dentate gyrus of the hippocampus which has been found to continue to develop into the preschool years (Seress 2007). The purpose of the present investigation was to examine whether age-related differences in hippocampal structure contribute to the development of episodic memory in middle and late childhood.
Support for protracted development of the hippocampus as a source of age-related improvements in episodic memory could come from evidence of structural hippocampal changes (i.e., volume) that extend into adolescence. However, results to date are mixed. In a recent investigation with 8- to 19-year olds, Ostby et al. (2009) found that total hippocampal volume showed some increases with age. Other investigations (e.g., Giedd et al. 1996; Yurgelun-Todd et al. 2003) found little to no relation between hippocampal volume and age.
However, there is some evidence that the age-related variation in hippocampal volume may be lost by evaluating the structure as a whole. Indeed, a longitudinal report, with a sample of participants ranging in age from 4 to 25 years, found that age–volume associations differed along the longitudinal axis of the hippocampus: with age, the anterior regions of the hippocampus decreased in volume, more posterior regions increased in volume, although the very posterior section of the hippocampal tail were mostly age invariant or showed a decrease on the left side (Gogtay et al. 2006; see also Insausti et al. 2010). Gogtay et al. (2006) did not collect behavioral memory measures, preventing the examination of volume–behavior associations.
These results suggest the hypothesis that, in adults, the associations between hippocampal volume and episodic memory might correspond to the direction of volumetric change during development; in other words, better episodic memory in adults may be associated with a smaller hippocampal head and larger posterior hippocampal regions, with the exception of the hippocampal tail which has not been shown to increase in volume. Consistent with this hypothesis, a recent study conducted with adults showed that smaller anterior hippocampus and larger posterior hippocampus are associated with better performance on episodic memory tasks (Poppenk and Moscovitch 2011); this study did not distinguish, however, between hippocampal body and tail within the posterior hippocampus. Thus, we might expect that this region is associated differently to episodic memory compared with the hippocampal body.
Our hypothesis and these results are also consistent with studies examining associations between hippocampal volumes and expertise with spatial representation, another function ascribed to the hippocampus. Maguire et al. (2000) showed that taxi drivers with extended experience driving through the streets of London, compared with control participants who do not drive taxis, exhibit decreased anterior hippocampal volumes and increased posterior hippocampal volumes. Similar patterns were confirmed when the control group included bus drivers with comparable driving experience and levels of stress, but who drive fixed routes (Maguire et al. 2006; see also Moser and Moser 1998, for evidence of dissociable functions of anterior vs. posterior hippocampus in the rodent literature).
Overall, previous research indicates that associations between hippocampal volume and episodic memory may vary along the longitudinal axis of this structure. The apparent change along this axis during development further suggests the hypothesis that associations between subregional hippocampal volume and episodic memory performance differ as a function of age.
Relations between hippocampal volume and episodic memory in childhood are debated in the literature (for a review, see Van Petten 2004); indeed, the results emerging from only a handful of studies have failed to provide a consistent picture about these associations. While nonsignificant associations between hippocampal volume and memory in childhood have been reported (e.g., Yurgelun-Todd et al. 2003), there is evidence of negative associations between overall volume and behavior (Sowell et al. 2001). More recently, nonsignificant associations between volume and behavior have been reported in a larger study assessing 8- to 19-year-olds' performance in the Rey Complex figure task after a 30-min and 1-week delay; interestingly, however, retention from the 30-min to the 1-week delay was positively correlated with hippocampal volume (Ostby et al. 2012). Whereas the authors reasonably interpreted this association as an indication that hippocampal volume tracked delayed consolidation processes, these results might also suggest that examining the hippocampus as a whole may fail to capture associations with memory performance at any given point in time.
Overall, the findings across studies assessing childhood have revealed an inconsistent pattern of results which raise the questions of whether previous research used measures that were sensitive enough to reflect true volume/behavior associations, and whether the absence of a direct comparison with an adult group prevented from observing differences in the nature of these associations. As detailed in the next paragraphs, in the present study, we provided a more sensitive test for the investigation of the age-related differences in associations between hippocampal volume and episodic memory by 1) assessing relations between regional volumes and episodic memory performance which might have been obscured by assessing the hippocampus as a whole, 2) focusing on one core behavioral component of episodic memory, namely, the ability to retain information about item-context associations, and 3) comparing associations between children and adults within the same study.
The Present Study
The central goal of the present investigation was to examine age-related differences in the relation between regional hippocampal volumes and episodic memory performance. We predict that in adults, smaller hippocampal head and larger hippocampal body will be associated with better episodic memory performance; whether associations would be evident in the tail is unclear: based on Gogtay et al. (2006), we might expect a negligible or negative correlation corresponding to the direction of volumetric change; however, based on Poppenk and Moscovitch (2011), who collapsed across hippocampal body and tail, we might expect a positive correlation.
While these associations for head and body are expected to be present in adults, they may not have emerged as clearly in children, as suggested by the inconsistent pattern of findings reported by previous volumetric studies examining childhood (Sowell et al. 2001; Yorgelun-Todd et al. 2003; Ostby et al. 2012). From this perspective, weaker associations may be expected in children compared with adults (for a discussion of a developmental perspective on volume/behavior associations, see Van Patten 2004). Conceptually consistent with this hypothesis are the findings from an fMRI study of encoding processes which showed adult-like activity profiles in anterior hippocampus in 14-year olds, but not in children 8–11 years of age whose activity pattern was less differentiated (Ghetti et al. 2010). Furthermore, age-related differences in hippocampal function were found during a retrieval task requiring participants to discriminate between true and false memories, with adults exhibiting a pattern of activity which discriminated among trial types in the experimental design to a greater extent than children ages 8–12 years (Paz-Alonso et al. 2008).
Although the small body of evidence reviewed thus far points to weaker associations in children, alternative hypotheses can be advanced. In a recent fMRI study with a subset of the present sample (DeMaster and Ghetti 2012), we investigated developmental differences in the recruitment of hippocampus and cortical regions during one of the episodic memory retrieval tasks utilized here. We found that children, but not adults, recruited the hippocampal tail bilaterally more strongly when they remembered an episodic detail accurately compared with when they did so inaccurately. In contrast, activity in the head and anterior body of the hippocampus was associated with successful episodic retrieval in adults; in children activity in these more anterior regions did not reliably discriminate between accurate and inaccurate episodic retrieval. The examination of whether the role of the hippocampal tail in children extended to volumetric measures further motivated the present investigation.
While examining these associations, we had the opportunity to achieve 2 additional goals. First, we sought to replicate age-related differences in regional volumes of the hippocampus reported by Gogtay et al. (2006), which remains the only empirical demonstration of a developmental dissociation in volume changes along the longitudinal axis of the hippocampus. Our study was conducted using a method that is different from that used in Gogtay et al. (2006) thereby affording the opportunity to replicate these previous results using a different approach. It is predicted that our child group will show a larger hippocampal head, a smaller hippocampal body, and smaller or similar hippocampal tail compared with adults. Second, we sought to replicate the weak or null associations between volume and episodic memory performance when the hippocampus is considered as a whole.
We examined episodic memory by combining 2 tasks aimed at measuring memory for item–context associations in children aged 8–11 years and adults. We selected these age groups based on behavioral research showing robust developmental differences between this group and adults in episodic memory measures (Brainerd et al. 2004; Ghetti and Angelini 2008; Shing et al. 2008) and based on research with fMRI showing the most sizeable differences from adults in the functional profile in the hippocampus (DeMaster and Ghetti 2012; Ghetti et al. 2010; Paz-Alonso et al. 2008; for evidence of developmental differences in hippocampal activation profile, see also Maril et al. 2010).
Overall, although there is no current consensus on the exact nature of processes supported by anterior versus posterior portions of the hippocampus, a growing literature suggests that functional differentiations may exist along the longitudinal axis of the hippocampus and that these differentiations may change with age. As the first investigation to examine how volumetric differences in subregions within the hippocampus in children and adults may contribute to age-related differences in episodic recollection, this study furthers our understanding of the development of episodic memory.
Materials and Methods
Sixty-two children and adults participated in this study. The child group consisted of 37 (54% female) 8- to 11-year olds (M = 9.65, SD = 1.32). The adult group consisted of 25 (48% female) 18- to 26-year-olds (M = 20.33, SD = 1.98). We excluded data from 3 child participants due to excessive head motion during imaging, and 2 adult participants due to scores 2 standard deviations below the mean on the Wechsler Abbreviated Scale of Intelligence (Wechsler, 1999). Participants were pre-screened for left hand dominance, color-blindness or other vision impairments, special education placement, history of head trauma, neurological or psychiatric conditions, and previous history of attention or learning disorders. Standard procedures were followed for excluding individuals who should not participate in MRI research (e.g., metal in the body, heart pace makers, insulin pumps, etc.). All sessions took place at the University of California, Davis Imaging Research Center. Children were financially compensated for their participation whereas adults received course credits.
Participants' consent was first obtained from parents in the case of child participants or directly from adult participants. In addition to parental consent, child participants were read a letter of information explaining the study and verbal consent was obtained.
Participants took part in 2 separate memory tasks both evaluating recollection of an item and the context in which the item was originally presented. The 2 tasks (a color task and a spatial task) occurred in 2 sessions, separated by 5–8 days; the order of the tasks was counterbalanced across participants. The tasks were adapted from several neuroimaging studies with adults that used similar paradigms to show that retrieval of item and detail information represents a process that is reliably found to be highly hippocampally dependent (Cansino et al. 2002; for a review, see Diana et al. 2007). Thus, we developed these tasks to measure age-related improvement in hippocampally mediated binding processes. Throughout their participation, we ensured children were comfortable and we provided opportunity for breaks.
In the encoding phase, participants viewed drawings of items presented in black ink. Either a red or green border surrounded each drawing. Participants were asked to remember what the drawing was or what it looked like and to remember the color of the border surrounding the drawing. Participants were also asked to press 1 of 2 buttons on a keypad corresponding to the color of the border surrounding the drawing (green or red). Participants practiced the encoding task on 12 trials. Following practice, participants viewed and responded to 100 unambiguous line drawings each presented for 2 s with a 2 s intertrial interval. All drawings have been used in previous studies and are normalized for visual complexity and familiarity (Cycowicz et al. 1997). To ensure encoding, an experimenter was seated behind the participant and monitored that the individual was attending to the pictures and pressing the button. Following the encoding phase, participants were given a mandatory break of 20 min, before beginning the retrieval phase.
In the retrieval phase, participants viewed old and new drawings presented in only black ink. Participants were asked to respond by selecting 1 of 3 options on a response pad: 1) The border was green; 2) the border was red; 3) the item was new. The task therefore emphasized the retrieval of contextual details. Participants practiced the retrieval task using the 12 drawings previously viewed in the encoding practice phase mixed with 9 drawings that were new. Following practice, participants viewed the 100 drawings that were previously studied and 75 new drawings. The retrieval task was performed in the scanner; functional data for a subset of the current sample are reported in DeMaster and Ghetti (2012).
The spatial task was identical to the color task with the exception that memory for location of the item was tested rather than the color of the border that originally surrounded the item. Thus, during the encoding session, participants viewed drawings presented on either the right or left side of the screen and during the retrieval session, they used the response pad to determine whether the item had been seen on the right side, on the left side, or was new. The retrieval phase of this task was also performed in the scanner.
MRI data were acquired on the UCD-IRC 3T Siemens scanner. High-resolution MP RAGE anatomical scans were acquired (TR = 2580, TE = 5.05, flip angle = 7°, FOV = 210, slice thickness = 0.66, and voxel size = 0.7 × 0.7 × 0.7). This anatomical scan, lasting 12:32 min, was acquired while participants watched cartoons.
Parcellation of the Hippocampus
Hippocampal volume was estimated with FreeSurfer version 4.5.0 (http://surfer.nmr.mgh.harvard.edu), an automated segmentation software program which identifies the hippocampal structure along with other cortical and subcortical regions. After FreeSurfer segmentation, 2 independent trained researchers further delineated the hippocampal structure into 3 distinct regions (Fig. 1). For the left and right side separately, the hippocampal head, body, and tail were identified based on anatomical landmarks as described below.
First, in the coronal view using the initial anterior hippocampus slice identified by FreeSurfer, the researchers moved caudally to identify the final slice of the hippocampal head (Fig. 1A). This slice was the one in which digitations on the dorsal edge of the hippocampus are no longer apparent and the hippocampus begins to round into a tear drop shape (Duvernoy 2005; Mai et al., 2008; see also Pfluger et al. 1999) (See Fig. 1B shown example of initial slice of hippocampal body); similar approaches have been implemented in high-resolution protocols of the medial temporal lobes (Libby et al. 2012). We consider this landmark more sensitive to the transition from hippocampal head to hippocampal body than the previously used identification of the uncal apex (Weiss et al. 2005), because parcellating at the uncal apex results in the inclusion of considerable portions of the body into the head (Duvernoy 2005; Mai et al., 2008). Importantly, we believe our landmark may be more sensitive to age-related differences based on the results of Gogtay et al. (2006) in which age-related reductions in hippocampal volume were strongest in the most anterior portions of the hippocampal head. Using the uncal apex as a landmark may thus increase the likelihood that any age-related differences between children and adults in the hippocampal head are underestimated.
To identify the initial slice of the hippocampal tail, the researchers moved caudally through the hippocampus starting with the last hippocampal head slice, and selected the slice at which the fornix separates from the hippocampus and become clearly visible (Fig. 1D) unlike in the preceding slice (Fig. 1C; Watson et al. 1992). FreeSurfer determined the final slice of the hippocampal tail. Both researchers were blind to participant age and sex.
With regard to inter-rater reliability, the selected slice for end of the hippocampal head and beginning of hippocampal tail was the same (or differed by 3 slices or less) for 97.6% and 100% of cases, respectively (collapsing across left and right sides). To calculate inter-rater reliability formally, we calculated the following intraclass correlations (Shrout and Fleiss, 1979): 1) Intraclass correlation for the hippocampal head: We correlated the slice number selected by each experimenter as the end of the hippocampal head using the first hippocampal slice identified by FreeSurfer as the first slice in our count; and 2) intraclass correlation for the hippocampal tail: We correlated the slice number selected by each experimenter as the beginning of the hippocampal tail using the first slice of the body as the first slice in our count.
The results indicated reliability of our protocol. For the hippocampal head, the intraclass coefficient was 0.90 across participants (children: 0.91; adults: 0.90; P's < 0.0001). For the tail of the hippocampus, the intraclass coefficient was 0.98 (children: 0.98; adults: 0.97, P's < 0.0001). We used the values provided by one rater for all of the analyses involving subregions of the hippocampus.
Manual Tracing Validation
Previous studies have examined associations between age and hippocampal volume (Gilmore et al. 2012) and between hippocampal volumes and memory in children using FreeSurfer (e.g., Bramen et al. 2011; Ostby et al. 2012) based on the validation provided by Tae et al. (2008). Thus, the literature provides several examples of studies employing FreeSurfer to identify the external boundaries of the hippocampus in children and relating the resulting volumes to memory. However, FreeSurfer may result in somewhat poorer precision in the boundary between hippocampus and amygdala (Morey et al. 2009). Thus, an additional and independent researcher manually traced the hippocampal head and tail using a well-validated protocol used frequently in developmental work (Schumann et al. 2004; Nordahl et al. 2012), on a randomly selected subset of the current sample (N = 12; 19% of the total sample). For the hippocampal head, the intraclass correlation across participants was 0.97 (children: 95; adults: 0.98); for the hippocampal tail, the intraclass correlation across participants was 0.95 (children: 0.92; adults: 0.96). This result provides additional confirmation that the head and the tail of the hippocampus can be reliably identified with FreeSurfer.
Intracranial Volume Calculation
To account for variation in hippocampal volumes resulting from age differences in total brain volume, we calculated intracranial volume (ICV) using a set of automated procedures provided by the FMRIB Software Library version 4.1.9 (FSL; www.fmrib.ox.ac.uk/fsl; Zhang et al. 2001). We used the automated FSL brain extraction tool to isolate intracranial tissue from the skull and then calculated the ICV as the sum of gray and white brain matter including the inner and outer cerebrospinal fluid spaces. All hippocampal volumes were corrected for ICV using the method reported in Raz et al. (2005). Specifically, an adjustment was performed on each volume using a formula based on the analysis of variance correction: Volume (adj) = Vol(rawi) − b × (ICV(i) − mean ICV), where Volume (adj) is the adjusted volume for the individual, Vol(rawi) is the unadjusted volume for the individual, b is the slope of the regression of the ROI volume (Vol(rawi)) on ICV, ICV(i) is the ICV for that individual and Mean ICV is the sample mean of ICV. As a preliminary check, we conducted a regression of regional volume on ICV, age group, and ICV × age group. Since the interaction of ICV × age group was not significant for any volume (P's > 0.15), children and adults were treated as one sample for the purpose of the adjustment (i.e., mean ICV was calculated using the mean of the entire sample). All following descriptive statistics, including figures, and analyses refer to ICV-adjusted regional volumes.
The goal of this research is to evaluate whether age-related differences in hippocampal structure contribute to the developmental improvement in episodic recollection. To achieve this goal, our results are presented in 3 sections. First, we evaluate age-related differences in volume of the hippocampal head, body, and tail. Second, we report age-related differences in behavioral performance on our measures of episodic memory. Finally, for the child and adult groups separately, we evaluate associations between variations in hippocampal subregion volumes and our measure of episodic retrieval.
Age-Related Differences in Hippocampal Volume
We first examined the correlation between age and total hippocampal volume and found a significant positive correlation in total hippocampus (sum of left and right hippocampi), r60 = 0.30, P = 0.02, in the left hippocampus, r60 = 0.33, P = 0.01, and a nearly significant positive correlation in the right hippocampus, r60 = 0.24, P = 0.06. We predicted that age-related difference in hippocampal volume would be evident in hippocampal subregions. Given the positive correlations of age with overall hippocampus even after adjusting for ICV, we considered it necessary to control for total hippocampal volume when evaluating age-related differences in its subregions. Thus, we first conducted a 2 (age group: children vs. adults) × 3 (region: head, body, and tail) × 2 (hemisphere: right, left) ANCOVA controlling for total hippocampal volume. A main effect of hemisphere (F1,59 = 7.02, P < 0.02), region (F2,118 = 4.86, P < 0.01), a region × age interaction (F2,118 = 10.34, P < 0.0001), and a hemisphere × region interaction (F2,118 = 3.67, P < 0.05) were found. In order to follow-up these interaction effects, we conducted ANCOVAs for right and left side separately, controlling for respective hippocampal volume (i.e., right hippocampal volume or left hippocampal volume). See Figure 2. For the right hippocampus, main effect of region (F2,118 = 6.98, P < 0.01), and a significant region × age interaction (F2,118 = 10.16, P < 0.0001) were found, such that the hippocampal head and tail were larger in children compared with adults, P's < 0.05, and the body was significantly smaller in children compared with adults, P < 0.001. For the left hippocampus, the region × age group ANCOVA revealed no significant main effect of region (F2,118 = 2.22, P = 0.11) and a significant region × age interaction (F2,118 = 5.15, P < 0.01), such that the hippocampal body was smaller in children compared with adults, P < 0.01; the hippocampal head was marginally larger in children compared with adults, P = 0.058; and there was no difference between groups in the left tail, P = 0.13. Taken together, the results show age-related differences along the anterior-posterior hippocampal axis with age-related decreases in hippocampal head volume in the right but less so in the left, age-related increases in hippocampal body volume bilaterally, and age-related decreases in right hippocampal tail were also apparent.
As described earlier in the Parcellation of the Hippocampus section, we had proposed that using the uncal apex as a landmark to distinguish hippocampal head from body may increase the likelihood that any age-related differences between children and adults in the hippocampal head are underestimated. To verify whether this was the case, we segmented the hippocampal head and body volumes based on the uncal landmark method and reanalyzed the data. When we segmented the data using the uncal landmark method, our inter-rater reliability remained high (all intraclass correlations were well above 0.90 for children and adults), yet age-related differences in hippocampal head and body were no longer significant (r’s < 0.19, P’s > 0.37). To further substantiate our claim that a more conservative approach to the head may be more meaningful to understand development, we calculated the volume included between our landmark for the hippocampal head and the uncal landmark. We then adjusted this portion of the hippocampus for ICV. We focused on the right hippocampus which shows the most robust change (Fig. 2), and found that this portion was positively correlated with age, r = 0.34, P = 0.007. However, the hippocampal head volume using our current approach was negatively correlated with age, r = −0.29, P = 0.02. A similar, but attenuated pattern is present in the left hippocampus. The fact that these correlations are in opposite directions suggests that these regions are in fact heterogeneous as noted by Insausti et al. (2010), and should not be considered in the same volume for our purposes (as would be the case if we used the uncal landmark method to obtain a hippocampal head volume). Using the uncal landmark washed away any effects of age, because we combined a region that showed a positive correlation with age and a region that showed a negative correlation with age.
The primary measure of interest in both the color and spatial tasks was defined as memory for item–contextual detail association (i.e., referred to as source to simplify). For the color task, a correct source judgment reflected correctly indicating the color of the border that accompanied the drawing during the encoding task. Likewise, correct source judgment in the spatial task was defined as correctly indicating the side of the screen on which the drawing appeared during encoding. To evaluate source accuracy, we divided the total correct source judgments by the total number of correct plus incorrect source judgments for each task. An age group (children, adults) × task type (color, spatial) repeated-measures ANOVA revealed a main effect of task (F1,60 = 23.86, P < 0.001), and a main effect of age (F1,60 = 10.63, P < 0.005) on source index accuracy (see Supplementary Fig. 1). Across age groups, performance in the spatial task (M = 0.68, SD = 0.16) was higher than performance in the color task (M = 0.59, SD = 0.13), and across tasks, source accuracy of adults (M = 0.70; SD = 0.10) was higher than those of children (M = 0.60; SD = 0.13). Importantly, there was no age × task type interaction (F1,60 = 1.56, P = 0.22).
A partial correlation, controlling for age, showed that performance on these 2 tasks were reliably correlated, r59 = 0.44, P < 0.001. Thus, we averaged scores across the 2 tasks and used this average “source index,” a comprehensive measure of memory for contextual details, to examine associations with regional hippocampal volumes. It is possible that participants engaged in some guessing at the item level. However, children and adults were significantly above chance despite the inclusion of source judgments for novel items (i.e., false alarms) in the denominator which suggests that guessing did not jeopardize the interpretability of the results.
Relations Between Hippocampal Volumes and Behavioral Performance
One of the central goals of the present research was to examine whether associations between volumes of hippocampal subregions and source index differed as a function of age. Thus, we conducted correlational analyses separately for child and adult groups and used Fisher's z-tests to assess whether there were significant differences in the correlations between the 2 age groups. As age and source index were correlated in the full sample, r60 = 0.44, P < 0.001 we controlled for age by using residual scores that accounted for the effects of age in both groups. Note that the results reported below do not change when we controlled for both age and ICV in the source index.
Correlations Between Hippocampal Volumes and Episodic Memory
We first correlated source index measures with each subregion starting from the hippocampal head and proceeding caudally. For the hippocampal head, there was a significant negative correlation between source index and volume in the right hippocampus for adults, r23 = −0.44, P < 0.05, but not in children, r35 = −0.03, P = 0.87 (Fisher's z = 1.64, P = 0.05, 1-tailed). As shown in Figure 3a, smaller right hippocampal head was associated with better memory performance in adults, but not in children. Correlations between left hippocampal head and source index were not significant for either adults, r23 = 0.04, P = 0.84, or children, r35 = −0.28, P = 0.10.
For the hippocampal body, there was a significant positive correlation between right body and source index in adults, r23 = 0.48, P < 0.05, but not in children, r35 = −0.14, P = 0.43 (Fisher's z = 2.43, P < 0.01, 1-tailed). As shown in Figure 3b, larger right hippocampal body was associated with better memory performance in adults, but not in children. Correlations between left hippocampal body and memory were not significant for either adults, r23 = −0.04, P = 0.85, or children, r35 = 0.14, P = 0.41.
Finally, in adults, there was no significant correlation between hippocampal tail volumes and source index (left: r23 = −0.30, P = 0.14; right: r23 = −0.24, P = 0.25); and a significant positive correlation was found in children in the left hippocampal tail, r35 = 0.36, P < 0.05, but not the in right, r35 = 0.23, P = 0.17. The difference between children and adults for these correlations was significant for the left tail (Fisher's z = 2.51, P < 0.02, 2-tailed; Fig. 3c) and approached significance for the right tail (Fisher's z = 1.77, P = 0.08, 2-tailed). Though the mean source index is more representative of individual variation in episodic memory, Supplementary Table 1 is available which presents correlations between each hippocampal subregion and performance on the color and spatial tasks for children and adults.
Finally, we examined the correlations between source index and volume of the entire hippocampus. We found nonsignificant correlations in both adults (left, r23 = −0.23, P = 0.28; right, r23 = −0.22, P = 0.29) and children (left, r35 = 0.03, P = 0.87; right, r35 = −0.008, P = 0.96).
Recollecting detailed contextual information associated with previously experienced events requires binding processes that are highly dependent on the hippocampus (Eichenbaum and Cohen 2001; Diana et al. 2007). The main contribution of the present research is that it examines age-related differences in volume of hippocampal subregions and its associations with episodic memory. We found that there were age-related improvements in our memory measures, age-related differences in total hippocampal volume, age-related differences in volumes along the longitudinal axis of the hippocampus and, central to our goals, different relational patterns between behavioral and volume measures for children and adults. For the adult group, results indicate that individuals with a smaller right hippocampal head and larger right hippocampal body obtained better source memory scores, our measure of episodic memory. In contrast, in the child group, no relation was found between head or body of the hippocampus and behavioral performance; however, a relationship was found between volume of the left hippocampal tail and source memory; the association with the right hippocampal tail failed to achieve conventional levels of statistical significance but was in the same direction as the correlation in left hippocampus. For children, individuals with larger hippocampal tails obtained higher source memory scores.
We had predicted age-related decreases in volume of the hippocampal head and increases in volume of the hippocampal body and that the direction of this change would be reflected in the sign of the correlation in adults between episodic memory and regional hippocampal volume (i.e., a negative correlation in the anterior hippocampus and a positive correlation in the body of the hippocampus). The results were fully consistent with this prediction in the right hippocampus, and with the finding reported by Poppenk and Moscovitch (2011), who showed negative associations between episodic memory and volume of the hippocampal head, and positive associations with the volume of the posterior hippocampus, which did not differentiate the body from the tail. The current study identifies the hippocampal body as the part of the posterior hippocampus driving this correlation.
Our results were also consistent with the direction of volumetric differences in anterior and posterior hippocampus between adults with experience in spatial navigation (i.e., London taxi drivers) compared with individuals with less or no experience (Maguire et al. 2000, 2006). It is not clear why these correlations were only found on the right hippocampus in the current study, though this is not the first study reporting volumetric results concerning anterior–posterior distinctions that were limited to the right hemisphere (e.g., Maguire et al. 2000).
The results from Maguire and coworkers indicate that experience is a crucial mechanism through which hippocampal change is achieved (Woollett and Maguire 2011). The absence of correlations in the hippocampal head and body in children raises the question as to the extent to which maturational changes and experience, which necessarily vary as a function of age, contribute to regional volumetric change and related associations with memory.
In contrast to results in adults, no significant correlations between hippocampal head and body with episodic memory were detected in children, which is consistent with our prediction that this age group would exhibit weaker associations in these regions. In contrast, we found an association between the hippocampal tail and episodic memory, which was not detected in adults. We had proposed the prediction that the hippocampal tail may be involved in episodic retrieval in children based on our recent results showing that children recruited the hippocampal tail bilaterally, but not the rest of the hippocampus, to retrieve episodic detail (DeMaster and Ghetti 2012); thus, it is possible that this region is the first during development to capture associations with episodic memory. The absence of correlations in these regions in adults suggests that a shift might occur during development such that more anterior regions become increasingly relevant for episodic memory compared with very posterior ones. Future research should examine whether the correlation between hippocampal tail volume and episodic memory is present even in younger children. Future research should also examine within-individual change over time: only longitudinal research can demonstrate conclusively whether and the extent to which a posterior-to-anterior shift occurs and what factors contribute to its occurrence.
These results raise the question of whether functional segregations along the longitudinal axis of the hippocampus might help explain these results. There is evidence from the adult functional neuroimaging literature that the anterior and posterior regions of the hippocampus support distinct functions during episodic retrieval (Giovanello et al. 2009). The anterior hippocampus seems to play a critical role in flexible retrieval of contextual features, whereas more posterior regions seem to support episodic retrieval when there is an exact match between encoding and retrieval conditions (Giovanello et al. 2009). Thus, it is possible that we found volume-behavior associations in posterior regions for children because they are relying on processes that attempt to reinstate previously encoded representations. Adults, on the other hand, may rely on anterior regions in which features of events are retrieved flexibly. Flexible retrieval of contextual information is a vital feature of episodic memory, thus more reliance on anterior regions could signal more mature or sophisticated function. More broadly, it is possible that the differences observed here reflect more general reorganizations along the hippocampal axis, which transcend reliance on specific memory processes. Future studies should investigate these possibilities, as differences in the relation between hippocampal structure and memory performance for the adult and child groups are highly relevant to our understanding of the development of episodic memory. Future studies should also investigate the relation between laterality and anterior/posterior distinctions. We note that our brain-behavior correlations are restricted to the left hippocampus in the tail, and the right hippocampus for the head and body. This raises the question about whether the developmental differences in anterior/posterior are constrained by differing developmental trajectories for the left versus the right hippocampus (see Gogtay et al., 2006 for some evidence that age-related differences in hippocampal volume along the longitudinal axis may be more pronounced on one side compared with the other depending on the point along the longitudinal axis).
Several of our findings replicate previous work, and thus it is unlikely that there are idiosyncrasies with the current data set. As already discussed, we replicated the findings reported by Poppenk and Moscovitch (2011) with adults. In addition, consistent with Gogtay et al. (2006), we found developmental dissociations in size of different regions in the hippocampus along the anterior-posterior axis. One inconsistency is the finding that the age-related decrease in volume in the hippocampal head was reliable in the right hippocampus, and only approached significance in the left, whereas it was found bilaterally in Gogtay et al. (2006). It is possible that methodological differences between these studies might account for this slight difference. For example, the youngest children in the present research were 8-year-olds whereas Gogtay et al. (2006) calculated developmental change relative to hippocampal volume at 4 years of age. Thus, it is possible that age-related structural change in the left hippocampal head would have been stronger if a younger sample was included here. Gogtay et al. (2006) indeed reported more delayed developmental change in the right compared with the left hippocampus. Furthermore, asymmetries in the nature and extent of associations have been previously reported, with the right hippocampus being more likely to exhibit significant associations with a host of factors including sex differences during puberty (Bramen et al. 2011), and substance abuse during adolescence (Ashtari et al. 2011).
Also, we cannot exclude that inconsistencies may depend on the method used to examine volumetric changes in hippocampal subregions. For example, in Gogtay et al. (2006) the hippocampus was divided into 5 sections from anterior to posterior; whereas we used a segmentation method that requires segmentation of the hippocampus based on internal landmarks. Thus, we can only make conceptual, not precise, comparisons of our findings to those of Gogtay et al. (2006). A second methodological difference is that Gogtay and colleagues used a dynamic mapping technique whereas we used a tracing protocol implemented on one anatomical scan using the automated segmentation program FreeSurfer (http://surfer.nmr.mgh.harvard.edu). Ideally, this program would use 2 anatomical scans to generate an averaged template from which segmentations are formed. Nevertheless, the use of FreeSurfer was validated for the hippocampus with one anatomical scan by Tae et al. (2008); furthermore, we reported new validation data with manual tracing in the current study. Regardless of the method, our research converges with Gogtay et al. (2006) in emphasizing the importance of using parcellation protocols that capture the heterogeneity in the development of hippocampal subregions.
We acknowledge that the relatively small sample included in the present study limits the conclusions that can be drawn. For example, several interesting factors, like sex differences and age by sex interactions, could not be tested to begin to understand whether developmental differences in associations vary as a function of sex. This question is relevant for understanding late hippocampal development during the transition to puberty (Bramen et al. 2011) and should be addressed in future research. An additional caveat of the present study is that these data were collected in a cross-sectional design which only allows for between group comparisons. In future work, a longitudinal investigation could measure hippocampal structure and episodic memory, and would allow for a more comprehensive assessment of age-related improvement in binding as a function of change in hippocampal structure during the childhood years.
Understanding the relation between hippocampal structure and age-related improvement in memory for contextual details adds to our knowledge of how changes in binding processes contribute to improvement in episodic recollection during childhood. Binding, a process that is highly hippocampally dependent, is the mechanism by which all the fragmented details from each event are bound together to form qualitatively rich representations of the past. By beginning to characterize the relationship between structural development of the hippocampus and memory for contextual detail, the results from this investigation contribute to a growing body of literature indicating age-related change in the subregions of the hippocampus as a contributor to improvement in episodic recollection.
The present research was supported with a James S. McDonnell Scholar Award to S.G.
Conflict of Interest: None declared.