Focal hippocampal subfield atrophy in cognitively impaired relative to cognitively unimpaired older adults is associated with memory ability Authors: Abi Heller‐Wight, David G Ellis, Anna Wilhelm, Caroline Nester Rooney, Nadia Pare, Nicole Sergeyev, Laura Rabin, David E Warren Published: 2026 ABSTRACT AbstractINTRODUCTIONWe assessed whole hippocampal and subfield volumes, in association with memory ability, in a sample including cognitively unimpaired (CU) older adults and those in early syndromic stages of pathological aging, here subjective cognitive decline (SCD) and mild cognitive impairment (MCI).METHODSParticipants completed a battery of neuropsychological tests and a 3T magnetic resonance imaging brain study that included high‐resolution T2‐weighted images of hippocampus and its subfields: subiculum; cornu ammonis 1, 2, and 3 (CA1, CA2, CA3); and dentate gyrus (DG).RESULTSSignificant differences were found in the left whole hippocampus, left subiculum, CA1, and DG volumes in the MCI group compared to the CU or SCD groups. Least absolute shrinkage and selection operator regression techniques revealed that left subiculum and left CA1 volume significantly predicted memory ability above and beyond whole hippocampal and other subfield volumes.DISCUSSIONWe found clinical differences in memory ability in older adults were associated with focal differences in specific hippocampal subfield volumes.Keywords: aging, brain imaging, hippocampal subfields, hippocampus, memory, mild cognitive impairment, volumetrics BACKGROUND Both healthy and pathological aging are marked by changes in cognitive function and associated brain structures. For example, memory loss is the most common age‐related cognitive complaint, and memory deficits are strongly associated with focal brain changes including atrophy of brain structures that support memory abilities. Even in healthy aging, age‐related atrophy can be observed in regions such as the medial temporal lobe (MTL), particularly the hippocampus, 1 with correlated changes in memory ability. Meanwhile, more pronounced memory loss and hippocampal atrophy may be related to underlying neuropathological changes 2 , 3 including but not limited to neurodegenerative disorders such as Alzheimer's disease (AD). Thus, atrophy of the hippocampus, whether due to healthy or pathological aging, is closely linked to changes in memory. Outstanding questions regarding hippocampal volume differences and memory function include whether a more granular approach to characterizing hippocampal volume, perhaps at the level of its component subfields, might refine the field's understanding of associations between regional brain volume and memory ability. The hippocampus is composed of several subfields with unique cytoarchitecture and well‐described interconnectivity 4 , 5 that collectively support critical memory processes. 6 Growing evidence highlights the importance of characterizing changes in hippocampal subfield volumes in older adults with or without neurological disease using cross‐sectional 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 and longitudinal 19 , 20 , 21 study designs. Critically, the studies using robust, reliable subfield segmentation techniques have found evidence of lower volume in patient populations experiencing cognitive declines in specific subfields, most frequently in the subiculum, cornu ammonis 1 (CA1), and dentate gyrus (DG). 9 , 19 , 22 , 23 Reductions in whole hippocampal volumes are associated with other disease processes as well (e.g., slower but robust hippocampal volumetric reductions in Parkinson's disease), making the characterization of focal hippocampal subfield changes important for developing a more sensitive magnetic resonance imaging (MRI) signature of AD‐related pathological change in the hippocampus that may contribute to memory decline. 24 , 25 , 26 Studies focused on distinguishing healthy aging trajectories from early syndromic stages of pathological aging typically include cognitively unimpaired (CU) individuals and those with mild cognitive impairment (MCI). Less attention has been focused on individuals with subjective cognitive decline (SCD), who report persistent cognitive concerns about memory and other cognitive abilities without objective evidence of clinically significant cognitive impairment. SCD has been studied as a potential precursor to MCI, and as such would represent a putative pre‐clinical stage in which microstructural changes in the brain may already be occurring in the absence of objectively measurable cognitive symptoms. 17 , 27 , 28 Some imaging studies of SCD have reported significant decreases in whole hippocampal volume 29 , 30 , 31 , 32 , 33 while others have not. 34 Despite the variability of SCD‐related volumetric findings to date, this potentially transitional stage preceding clinically significant cognitive impairment could provide insight into which patients are at greatest risk of progression to MCI and dementia. Understanding how changes in hippocampal subfield volumes are associated with early syndromic stages of pathological aging, including the associated changes in cognitive abilities such as memory, could potentially support earlier and more accurate diagnosis, enhanced treatment, and improved outcomes. Using high‐resolution brain imaging, this study aimed to investigate whether memory performance and hippocampal volumes (whole and subfield) varied among aging adults classified as CU, SCD, or MCI. We predicted that, compared to CU, individuals with MCI who had impaired memory ability would have decreased hippocampal volumes, including subfields. Meanwhile, predictions for SCD required nuanced consideration: to the extent that hippocampal volumes, including subfields, were associated with objectively measured memory ability, we predicted that the SCD group would have similar subfield volumes versus the CU group and potentially greater subfield volumes than the MCI group; to the extent that hippocampal volumes, including subfields, were instead associated with subjective insight regarding cognitive impairment, we predicted that the SCD group would more closely resemble the MCI group. Regarding the subfields most likely to differ between groups, based on prior work, we predicted that any differences observed would most likely be evident in the subiculum, CA1, and DG, the three largest subfields with the most frequent findings of age‐related and pathological differences. RESEARCH IN CONTEXT Systematic Review: Memory decline is the most common cognitive concern in older adults. Prior work has linked age‐related memory decline in healthy and pathologic aging to volumetric differences in the hippocampus and its subfields. We tested associations of memory ability and hippocampal subfield volumes in older adults across a continuum including cognitively unimpaired (CU), subjective cognitive decline (SCD), and mild cognitive impairment (MCI). Systematic Review: Memory decline is the most common cognitive concern in older adults. Prior work has linked age‐related memory decline in healthy and pathologic aging to volumetric differences in the hippocampus and its subfields. We tested associations of memory ability and hippocampal subfield volumes in older adults across a continuum including cognitively unimpaired (CU), subjective cognitive decline (SCD), and mild cognitive impairment (MCI). Interpretation: We observed volumetric decreases in the left hippocampus, subiculum, cornu ammonis 1 (CA1), and dentate gyrus in cases of objective (MCI) but not subjective (SCD) memory impairment. Subfield volumes robustly predicted memory performance across the sample, underscoring the value of subfield‐level measures in aging. Our findings suggest clinical differences in memory ability can be associated with focal differences in specific hippocampal subfields, specifically CA1 and subiculum. Interpretation: We observed volumetric decreases in the left hippocampus, subiculum, cornu ammonis 1 (CA1), and dentate gyrus in cases of objective (MCI) but not subjective (SCD) memory impairment. Subfield volumes robustly predicted memory performance across the sample, underscoring the value of subfield‐level measures in aging. Our findings suggest clinical differences in memory ability can be associated with focal differences in specific hippocampal subfields, specifically CA1 and subiculum. Future Directions: Longitudinal monitoring of subfield volumes, particularly in at‐risk groups such as those with SCD, may provide a sensitive marker of cognitive decline and disease progression. Future Directions: Longitudinal monitoring of subfield volumes, particularly in at‐risk groups such as those with SCD, may provide a sensitive marker of cognitive decline and disease progression. METHODS Participants All participants were recruited in Omaha, Nebraska as part of a National Institutes of Health–funded study measuring practical judgment in older adults (R15 AG084348). Enrollees ranged across healthy and syndromic pathological populations. The study protocol was reviewed and approved (IRB #789‐19‐EP) by the institutional review board of the University of Nebraska Medical Center (UNMC). Participants were between 60 and 90 years old and provided informed consent to participate in the study. Exclusion criteria included left handedness, diagnosed neurological disease (e.g., focal stroke, Parkinson's disease, epilepsy, multiple sclerosis), history of head trauma with loss of consciousness > 5 minutes, neurodevelopmental disorder, diagnosed learning disability, substance abuse disorder, and psychiatric disorder. Participants with a history of medication use that could affect cognition or any contraindications for MRI were excluded from the study. Additionally, a close associate of each participant provided informant reports; informants were typically spouses, siblings, or friends. One hundred older adults enrolled. Four participants did not complete all study procedures, and two participants had non‐usable MRI data. Therefore, the sample size for analysis was N = 94. Study design Neuropsychological assessments Participants completed a variety of neuropsychological tests assessing cognitive domains including memory, attention, and executive function. For detailed descriptions of all neuropsychological assessments and classification of individuals into CU, SCD, and MCI, please see the supporting information. MRI data acquisition and processing All MRI data were collected from participants in Omaha, Nebraska, at the UNMC's Core for Advanced Magnetic Resonance Imaging (CAMRI, RRID: SCR_022469). After completing MRI screening, participants completed a non‐contrast brain MRI scan with a Siemens 3T Prisma instrument using a 32‐channel head coil. Participants were provided with ear protection, positioned comfortably in the scanner, and asked to hold still. Real‐time audio–video feeds were used to monitor participants during scans. The protocol for the MRI study was adapted from the Lifespan Human Connectome Project. 35 Structural MRI collected as part of the protocol included a whole‐brain T1‐weighted (T1w) magnetization‐prepared rapid gradient echo (MPRAGE) sequence (echo time [TE] = 2.2 ms, repetition time [TR] = 2400 ms, slice thickness = 0.8 mm, field of view [FOV] = 256 mm, 0.8 × 0.8 mm in‐plane resolution, slices = 208, flip angle = 8°, acquisition time = 398 seconds), a whole‐brain T2‐weighted (T2w) sampling perfection with application optimized contrasts using different flip angle evolution (SPACE) sequence (TE = 563 ms, TR = 3200, slice thickness = 0.8 mm, FOV = 256 mm, 0.8 × 0.8 mm in‐plane resolution, slices = 208, acquisition time = 357 seconds), and a T2w turbo spin echo (TSE) sequence with high in‐plane resolution (0.39 × 0.39 mm) and 2 mm coronal–oblique slices oriented approximately perpendicular to the long axis of the hippocampus (TE = 66 ms, TR = 7790 ms, 0.39 × 0.39 mm in‐plane resolution, slice thickness = 2.0 mm, slices = 30, flip angle = 145°, FOV = 170 mm, acquisition time = 211 seconds). MRI quality control (QC) procedures are detailed in the supporting information. After QC review, the whole brain including the hippocampus was automatically segmented with FreeSurfer software (v6.0.0) 36 using the recon‐all workflow. FreeSurfer's subcortical segmentation processing produced the volume of the whole left and right hippocampi. Meanwhile, hippocampal subfields were automatically segmented using Automated Segmentation of Hippocampal Subfields (ASHS) software (Figure 1). 37 This required registration of the T1w whole‐brain image with the T2w slab image of the hippocampus using the ASHS “greedy” tool, followed by bias correction, and then segmentation into subfields using the University of Pennsylvania Aging Atlas. 22 The resulting segmentations represented volumetric measures of subiculum, CA1, CA2, CA3, and DG. Hippocampal subfield segmentation. Subfield extent is denoted by color (see key). A, Representative examples of ASHS segmentation of T2‐weighted TSE slab MRI data in each of the three study groups: CU, SCD, and MCI. MRI are presented in radiological convention. B, Representative example of 3D rendering of hippocampal subfields generated from ASHS output. AHS, Automated Segmentation of Hippocampal Subfields; CA, cornu ammonis; CU, cognitively unimpaired; DG, dentate gyrus; MCI, mild cognitive impairment; MRI, magnetic resonance imaging; SCD, subjective cognitive decline; TSE, turbo spin echo. Statistical analysis Statistical analysis used R software (R Core Team, 2023). When available, demographically corrected (age, sex, education) scores on neuropsychological assessments were utilized for analysis. Chi‐squared tests (categorical variables) and analysis of variance (ANOVA, continuous variables) were used to test differences in demographic data between groups (Table S1 in supporting information). Whole hippocampal and subfield volumes were corrected for total intracranial volume (ICV) using a published regression approach 38 to account for differences in cranial volume between participants. ICV‐corrected whole hippocampal and subfield volumes were used in all models. Differences between groups were compared using analysis of covariance (ANCOVA) models estimating marginal means using the rstatix::emmeans_test function while adjusting for covariates to evaluate differences in scores among CU, SCD, and MCI groups while correcting for age, sex, and education attainment (Table 1). Cell means were corrected for significant covariates before post hoc testing was completed, and false discovery rate (FDR) correction was applied to control for multiple comparisons. All models with assumption violations were tested with non‐parametric permutation ANCOVA models implemented with lmPerm in R and tested post hoc with Wilcoxon rank‐sum tests. Models using permutation testing are available in Table S2 in supporting information. Covariate‐corrected cell means are reported in Table 1, and Pearson correlations between variables of interest are included in Figure 2. TABLE 1Group differences: whole hippocampal and subfield volumes.Adjusted cell means P valuesPartial η2 CUSCDMCICU/SCDCU/MCIMCI/SCDWhole bilateral hippocampus7227 (128.0)7755 (125.0)7175 (125.0)0.550.04* 0.01* 0.19Right hippocampus3866 (71.6)3994 (69.9)3681 (70.1)0.480.170.070.13Left hippocampus3761 (65.4)3811 (63.8)3494 (64.0)0.690.01* 0.006** 0.21Right sub436 (10.8)442 (10.5)419 (10.6)0.650.380.340.10Left sub467 (10.8)467 (10.5)433 (10.5)0.940.03* 0.03* 0.16Right CA11205 (26.3)1221 (25.6)1167 (25.7)0.620.510.440.10Left CA11201 (25.2)1201 (24.6)1116 (24.7)0.910.001** 0.001** 0.19Right CA220 (0.8)20 (0.7)21 (0.7)0.680.680.680.00Left CA231 (1.1)30 (1.1)29 (1.1)0.840.640.640.02Right CA376 (3.7)78 (3.6)72 (3.7)0.680.680.690.01Left CA373 (3.5)76 (3.4)69 (3.4)0.970.880.880.01Right DG730 (16.4)687 (16.0)724 (16.0)0.770.770.140.15Left DG690 (18.3)697 (17.8)631 (17.9)0.810.02* 0.02* 0.20 Notes: Unless otherwise indicated, values are mean (± SE). All volumes are in mm3. “Adjusted cell means” refer to mean values adjusted for covariates in analysis of variance models (age, sex, and education in volume models). P values describe post hoc pairwise comparisons with multiple comparisons (FDR) correction. ns = not statistically significant after post hoc comparisons.Abbreviations: CA, cornu ammonis; CU, cognitively unimpaired; DG, dentate gyrus; FDR, false discovery rate; MCI, mild cognitive impairment; SCD, subjective cognitive decline; SE, standard error.*P < .05,** P < .01. Group differences: whole hippocampal and subfield volumes. Notes: Unless otherwise indicated, values are mean (± SE). All volumes are in mm3. “Adjusted cell means” refer to mean values adjusted for covariates in analysis of variance models (age, sex, and education in volume models). P values describe post hoc pairwise comparisons with multiple comparisons (FDR) correction. ns = not statistically significant after post hoc comparisons. Abbreviations: CA, cornu ammonis; CU, cognitively unimpaired; DG, dentate gyrus; FDR, false discovery rate; MCI, mild cognitive impairment; SCD, subjective cognitive decline; SE, standard error. *P < .05, ** P < .01. Cross‐correlation matrix of memory and hippocampal volumetric variables. The two memory indices were positively correlated, as were many of the hippocampal volumetric measures. Delayed memory index scores were positively correlated with several hippocampal volumetric measures. All P values corrected for multiple comparisons (FDR). *P < .05, **P < .01. CA, cornu ammonis; DG, dentate gyrus; FDR, false discovery rate; MI, memory index. To further assess the predictive performance of volumetric measures for delayed memory performance, we utilized cross‐validated elastic net regularization regression techniques (least absolute shrinkage and selection operator [LASSO], 20 folds) for variable selection from an initial multiple regression that included many volumetric predictors simultaneously. RESULTS As expected, memory performance differed between groups such that the MCI group had significantly lower scores on both memory indices than the CU and SCD groups (Figure 3). Descriptive statistics of demographic information and neuropsychological testing scores across groups are available in the supporting information. Regarding volumetric differences in whole hippocampus and hippocampal subfields, positive associations between hippocampal certain volumetric measures and delayed memory index (DMI) scores were observed (Figure 2). Group differences in memory index scores. The CU and SCD groups were not significantly different on either measure, while the MCI group had reduced performance relative to both groups on the immediate and delayed memory index scores. Boxplots show the mean (center line), interquartile range (box), and whiskers extending to 1.5 × interquartile range; individual points represent participants. A, Differences between groups on RBANS immediate memory index scores (*P < .01, ** P < .001). B, Differences between groups on RBANS delayed memory index scores (*P < .01, ** P < .001). Individual data points were jittered for display. CU, cognitively unimpaired; MCI, mild cognitive impairment; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; SCD, subjective cognitive decline. Group differences in whole hippocampal and subfield volumes Generally, we observed that across the whole hippocampus and within subfields, the CU and SCD groups had similar volumes, while the MCI group frequently had significantly smaller volumes. Significant between‐group differences were observed in the left whole hippocampus (F[2,88] = 11.62, P < .001). Planned comparisons testing revealed that individuals in the MCI group had a significantly smaller left hippocampus on average compared to the SCD and CU groups (Table 1, Figure 4). Regarding left‐lateralized subfields, subiculum (F[2,88] = 8.10, P < .001), CA1 (F[2,88] = 8.30, P < .001), and DG (F[2,88] = 11.46, P < .001) had between‐group significant differences such that participants with MCI had less volume compared to SCD or CU groups (Table 1, Figure 4); there were no such differences in CA2 or CA3. No other significant differences between groups were observed in the right whole hippocampus or right hippocampal subfields. Volumetric measures of whole hippocampal and subfield volumes by group. The MCI group had significantly smaller volumes than the CU and SCD groups in left whole hippocampus, subiculum (Sub), CA1, and DG. Boxplots show the mean (center line), interquartile range (box), and whiskers extending to 1.5 × interquartile range; individual points represent participants. (A) Bilateral hippocampus. (B) Sub = subiculum (C) CA1 = cornu ammonis 1. (D) CA2 = cornu ammonis 2. (E) CA3 = cornu ammonis 3, (F) DG = dentate gyrus. Volumes were corrected for intracranial volume. Contrasts were corrected for multiple comparisons with an FDR correction. *P < .05, **P < .01. CU, cognitively unimpaired; FDR, false discovery rate; MCI, mild cognitive impairment; SCD, subjective cognitive decline. Hippocampal subfield volumes and memory index scores across groups We observed that several measures of whole hippocampal or subfield volume were associated with delayed, but not immediate, measures of memory performance (Figure 2). Regression analysis revealed that left whole hippocampal volume (F[3,90] = 2.98, P < .05, R2 = 0.09) and right whole hippocampal volume (F[3,90] = 2.80, P < .05, R2 = 0.08) were significant predictors of DMI performance when controlling for sex and years of education. Left subiculum (F[3,90] = 5.19, P < .01, R2 = 0.15), right subiculum (F[3,90] = 2.53, P < .05, R2 = 0.08), left CA1 (F[3,90] = 7.35, P < .001, R2 = 0.19), right CA1 (F[3,90] = 3.58, P < .01, R2 = 0.11), left DG (F[3,90] = 5.32, P < .01, R2 = 0.15), and right DG (F[3,90] = 4.52, P < .05, R2 = 0.13) subfield volumes significantly predicted DMI performance scores across the whole sample. Meanwhile, neither whole hippocampal nor subfield volumes significantly predicted immediate memory index (IMI) scores across the entire sample. These regression analyses indicated that, controlling for participant variables, greater volume in the right/left whole hippocampus, subiculum, CA1, and DG were significantly and positively associated with performance on the DMI, but not the IMI. Hippocampal subfield volumes and memory index scores within groups Within‐group correlations of volumetric measures and memory scores were also calculated (Figures S1–S3 in supporting information). Similar to the correlation pattern across the entire sample, many subfield volumes were correlated. A Jennrich test revealed significant differences between all groups. However, post hoc tests to compare individual correlations were not significant after adjusting for multiple comparisons. We interpreted this finding to reflect patterns of correlation between groups for the volumetric and cognitive variables that differed in a diffuse, not concentrated, manner. Within each group, linear regression models utilizing hippocampal and subfield volumes to predict memory index scores while controlling for sex and education level were not significant. The reduced size of individual groups versus the entire sample likely reduced statistical power for these tests. Exploratory analysis of robust hippocampal volumetric predictors of memory performance Follow‐up exploratory analysis of the association between hippocampal volumetric measures and memory performance provided evidence that two hippocampal subfields had the most robust positive associations with DMI: left CA1 and left subiculum. Having observed strong positive associations between DMI performance and several hippocampal volumetric measures (whole and subfield) between groups, we sought to refine our interpretation of these findings in a manner that would address the many strong positive correlations between hippocampal subfield volumes (Figure 2). For this purpose, we used an exploratory LASSO regression approach to select optimal volumetric predictors for DMI. Predictors in the initial model included all whole hippocampal and subfield volumes (left and right for each) as well as (non‐penalized) covariates for sex and education; the outcome was DMI. The LASSO regression model that included the DMI as the outcome produced two non‐zero coefficients (λ = 2.72, mean square error = 225.1): (1) left CA1 volume (β = 4.69), and (2) left subiculum volume (β = 2.24). Thus, the LASSO regression model supported rigorous testing of the associations between highly correlated volumetric predictors available in this study to determine which had the most robust association with DMI performance, namely, left CA and left subiculum. DISCUSSION The current study measured differences in whole hippocampal and hippocampal subfield volumes across CU, SCD, MCI groups, then tested whether those volumes were associated with memory ability across the sample. By including people with SCD, this study bolsters the representation of early syndromic stages of cognitive decline, while contributing to the limited body of published literature in this group. Individuals with MCI showed significantly smaller volumes in the whole left hippocampus, CA1, subiculum, and DG compared to CU and SCD groups, who did not differ. No significant between‐group differences were observed in the right hippocampus. Across the entire sample, we observed a significant relationship between memory performance and volume of bilateral whole hippocampus, as well as select subfield volumes. These findings are consistent with the perspective that syndromal cognitive changes, specifically in memory ability, are significantly associated with smaller subfield volumes. We observed strong associations between hippocampal and subfield volumes and delayed memory performance, and our exploratory analysis refined these associations and shaped our interpretation. Specifically, among the several strongly correlated hippocampal volumetric predictors of memory ability, our exploratory LASSO approach identified left CA1 and left subiculum subfield volumes as the most robust predictors of DMI. This demonstrates the potential for superior predictive value of specific subfield volumes compared to broader, whole‐hippocampal measures, for the current sample. We found these statistically robust associations compelling, but these findings should be tested for reproducibility and generalizability. Our results also suggest that contemporary hippocampal volumetric measures lack the necessary sensitivity to detect putative hippocampal volumetric atrophy in SCD. However, our findings indicate that measures of hippocampal subfield volumes may be sufficient to distinguish between subjective and objective cognitive changes in healthy and pathological aging. That is, participants with objective cognitive changes (MCI) had significantly smaller whole hippocampal and subfield volumes on average compared to participants reporting subjective decline (SCD) or normal cognition (CU). This pattern of hippocampal atrophy is not inconsistent with a biomarker for clinically normal cognition (CU, SCD) versus clinically impaired cognition (MCI). 29 , 39 Although prior studies reported mixed findings for between‐group differences in hippocampal volumetric measures, our results align with certain studies that have included an SCD group. 34 , 40 Those findings and our results are consistent with the view that hippocampal subfield atrophy is more closely associated with objective cognitive impairment than with subjective cognitive decline. Although SCD does not appear to differ from CU in the domain of hippocampal subfield volume, subjective changes in cognition remain predictive of future cognitive decline. 41 , 42 , 43 The hippocampal atrophy frequently observed along the AD continuum has been associated with amyloid and (especially) tau deposition, synaptic loss, dendritic pruning, and neuroinflammation. 3 We speculate that these same mechanisms may have contributed to the smaller hippocampus and subfield volumes we observed in the MCI group here, although we cannot confirm a causal association as this interpretation remains beyond the scope of MRI methods. Other neuroimaging measures, however, which include disease pathology (amyloid/tau), functional brain imaging (functional MRI, fluorodeoxyglucose positron emission tomography), or other structural measures (elastography) might better predict SCD‐to‐MCI conversion risk. Complementing the predictive power of SCD for AD outcomes with objective biomarkers is an important goal for future research. We observed focal changes in three hippocampal subfields in MCI, consistent with a “brain signature” of those cognitive changes. This prompts us to speculate about whether unique neurodegenerative diseases are associated with unique “signatures” of change in hippocampal subfields. While the current study capitalized on clinical syndromic staging of cognitive decline, whole‐hippocampal and subfield volumes have also been characterized in several progressive neurological diseases. Current evidence is largely consistent with a speculative account suggesting that differential patterns or “signatures” of subfield atrophy are indeed associated with diseases influencing hippocampal volume. For example, individuals with Parkinson's disease frequently exhibit cognitive decline and decreases in whole hippocampal volume. 24 , 25 However, patients with Parkinson's disease also exhibit smaller volumes in CA2 and CA3 compared to patients with MCI or AD. 26 Studies of MCI and AD have identified robust volumetric changes in subiculum and CA1 with negligible changes in CA2 and CA3. 17 Taken together, these patterns suggest that the neuropathology of distinct degenerative brain diseases may differentially influence focal atrophy in hippocampal subfields. If true, then memory profiles specific to these diseases would be expected to be associated with the unique roles of the affected subfields. 44 , 45 This speculative account regarding distinct signatures of hippocampal subfield atrophy and corresponding signatures of cognitive impairment would of course require refinement and validation through additional research, but we believe that our current findings provide preliminary evidence in support of this perspective by identifying differences between SCD and MCI. The results of the current study also suggested a laterality effect, with left hippocampal subfields showing a greater number of significant differences between groups (subiculum, CA1, DG) compared to the right subfields. We interpret these laterality effects from two perspectives: normative changes associated with age and disease, and modality‐ or material‐specific associations. The included task was primarily verbal in nature, which is more closely associated to left hippocampal volume and function, and which may also contribute to the left‐lateralized findings of the current study. 46 , 47 Laterality effects have been reported in other studies as well, along with evidence that the left hippocampal volumes may be differentially affected by aging and disease compared to the right. 18 , 48 The current study had some limitations. Similar to other studies published in this domain, lack of racial and ethnic diversity in the current sample was a significant limitation, and addressing this disparity is a key future direction, not only for our research group but for the field as a whole. Another limitation was the cross‐sectional design of the study, which must restrain any speculation regarding disease trajectories in the current sample. Our study also had several strengths. First, definitions of CU, SCD, and MCI were operationally defined through rigorous selection of neuropsychological instruments applied to the sample to achieve accurate classification of each participant. Using ASHS and the Penn Aging Atlas also provided significant advantages in validity over other automatic segmentation methods, and ASHS has been reported to rival manual tracing for accurate measurement of hippocampal subfield volumes in samples of older adults such as ours. 49 These strengths should serve to bolster the rigor, reproducibility, and generalizability of our findings while contributing to the field's ongoing exploration of differences between and changes to hippocampal subfield volumes throughout older adulthood. Conclusion The current study investigated differences in hippocampal subfield volumes between participants classified as CU, SCD, and MCI, using high‐resolution brain imaging. We reported evidence of volumetric decreases in left whole hippocampus, subiculum, CA1, and DG volumes in conjunction with changes in objective cognition: the MCI group presented with significantly lower memory scores and subfield volumes than the CU and SCD groups. Additionally, across the entire sample, subfield volumes were more robust predictors of memory performance than whole hippocampal volumes, highlighting the importance of measuring subfields in addition to whole hippocampal volumes in healthy and pathological aging. Our findings contribute to the foundational scientific knowledge necessary for identification of cognitive and brain signatures of early decline that could be useful in detecting prodromal disease stages before significant clinical symptoms emerge. Ultimately, this study deepens our understanding of how structural brain changes relate to memory performance across an early syndromal continuum consistent with probable AD. CONFLICT OF INTEREST STATEMENT The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information. DISCLOSURES The authors report no financial or personal relationships that could inappropriately influence (bias) the work reported in this manuscript. CONSENT STATEMENT All human subjects included in the current study provided informed consent in accordance with University of Nebraska Medical Center Institutional Review Board's policies and procedures. The protocol was approved under IRB #789‐19‐EP. Supporting information Supporting Information ACKNOWLEDGMENTS We would like to thank all the study participants and their informants. This work was supported by the National Institutes of Health (R15 AG066039 to LR, F30 AG084348 to AHW, P20 GM130447 to UNMC, R01 AG064247 to DEW), and NSF Award #2349686. REFERENCES Associated Data Supplementary Materials Supporting Information