The development of functional connectivity along the hippocampal long-axis in infants

doi:10.21203/rs.3.rs-5079807/v1

Download PDF

Article

The development of functional connectivity along the hippocampal long-axis in infants

https://doi.org/10.21203/rs.3.rs-5079807/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

We do not remember events experienced as infants. Infancy is a critical period of development for the memory system, yet we know little about the functional neural changes that occur during this time. In adults, hippocampal-neocortical coupling is needed to establish long-term memories, and differs along the anteroposterior axis. We investigated hippocampal-neocortical functional connectivity along the long-axis at rest in 212 infants. We found that functional differentiation of the anterior and posterior hippocampus occurs very early on (<6 months old). We also identified numerous cortical regions where connectivity with the hippocampus was changing with age. A clustering analysis revealed that anteroposterior hippocampal connectivity was changing with cortical regions associated with memory, but also with canonical networks associated with salience and attention. These findings raise the possibility that infantile amnesia is in part a disorder of immature functional interaction between memory, attention, and salience systems that engender memory formation in adults.

Biological sciences/Neuroscience/Development of the nervous system

Biological sciences/Neuroscience/Learning and memory/Hippocampus

Biological sciences/Neuroscience/Learning and memory/Consolidation

Biological sciences/Neuroscience/Learning and memory/Forgetting

Most adults cannot recall events experienced before the age of 3. While infants can encode and retrieve long-term memories for some period of time, they are forgotten at an accelerated rate compared to adults¹. This phenomenon has been termed ‘infantile amnesia’, and several hypotheses have been put forth as to its impetus. Principle among them is the idea that the memory system is immature, and unable to form long-lasting episodic representations^2,3. Indeed, the development of the hippocampus is quite protracted in humans, with robust hippocampal growth occurring across the first year post-gestation and continuing through childhood^4,5. Matching initial learning and retention in rodents across ages does not mitigate infantile amnesia, suggesting that this phenomenon is unlikely to be a disorder of deficient encoding^6,7. There is, however, evidence that infantile amnesia is to some extent an issue of retrieval or consolidation³. While it is difficult to test the maturity of consolidation mechanisms in humans, there is some evidence that the maintenance mechanisms for long-term potentiation are immature during the infantile amnesia period in rodents⁸. Optogenetic reactivation and even contextual reminders of experiences encoded during the infantile amnesia period can rescue a memory from forgetting^9,10, suggesting that the encoded representation exists in some form but is inaccessible.

In healthy human adults, episodic memories do not reside exclusively in the hippocampus, rather, the hippocampus is thought to act as an index to mnemonic representations stored in the neocortex^11,12. Consequently, hippocampal-neocortical coupling is conceived as critical for the consolidation and retrieval of episodic memories. In recent years, it has become clear that the topological organization of connectivity with the neocortex differs along the long-axis of the hippocampus in adults. Specifically, the anterior hippocampus is relatively more functionally and structurally connected to ventral medial prefrontal cortex (PFC), and anterior temporal areas, while the posterior hippocampus is more connected to dorsal medial PFC, posterior ventral temporal cortex, occipital cortex, and much of the precuneus^13–17. Accordingly, these hippocampal-neocortical networks along the long-axis are thought to support memories that are qualitatively different.

One prominent idea is that the anterior hippocampus and associated cortical regions construct the gist of past experiences while the posterior system elaborates on the details of episodic memory^18–21. It’s been found that infant rodents are better able to generalize behavior between contexts compared to older rodents due to rapid forgetting of detailed stimulus attributes, suggesting that details are forgotten faster than the gist in infants. In humans, while infants prior to the age of 3 do indeed form and retrieve memories, there is evidence that the quality of their memories differs from that of adults. In particular, human infants can retain elements of their past experience but appear unable to retrieve memories that capture arbitrary associations, which is the foundation of detailed episodic remembering^22,23. While it is yet unclear how these qualitative differences in infant memory map on to maturity of the hippocampus along its long-axis, the subfields of the hippocampus develop differently in the postnatal period. The monosynaptic pathway of the hippocampus, which involves CA1 and the subiculum, develops early and is thought to be able to support rudimentary memory capacity in infants such as statistical learning^24–26. The development of the trisynaptic path, involving the dentate gyrus and CA3, is protracted, perhaps delaying the emergence of more sophisticated episodic memory functioning^25,26. Notably, different proportions of hippocampal subfields that define these pathways exist in the anterior and posterior hippocampus. Specifically, the anterior hippocampus contains relatively more CA1, while the posterior contains relatively more dentate gyrus²⁷. Together, these studies raise the possibility that the development of hippocampal-neocortical memory systems proceeds differently along the long-axis of the hippocampus. However, we know very little about the maturity of hippocampal-neocortical connectivity in human infants, particularly along the long axis.

In the present study, we investigate the development of hippocampal functional connectivity with the neocortex along the hippocampal long-axis across the first two postnatal years of life in a large cohort of human infants. We hypothesized that we would observe increases in functional connectivity with age between the hippocampus and neocortical regions associated with episodic memory with age, and that connectivity along the long-axis of the hippocampus would become more differentiated with development as hippocampal-cortical networks are strengthened and refined.

We employed a seed-to-voxel analysis to measure functional connectivity of the anterior and posterior hippocampus to the whole brain. We ran a linear multivariate model to test for differences in connectivity as a function of long-axis (anterior/posterior), age bin (0-6mo/7-12mo/13-18mo/19-25mo), and hemisphere (left/right), while controlling for whole brain mean temporal signal-to-noise (tSNR) and mean motion. To better understand the effects of age bin and long-axis with development, we also contrasted hippocampal connectivity across successive age bins, and contrasted connectivity of the anterior with that of the posterior hippocampus (collapsed across hemispheres) within each age bin using paired t-tests (all FDR-corrected to q < 0.01, clusterized to k = 10 or more voxels).

Hippocampal coupling with memory-related cortex increases across the first 6–12 months of age

First, we present the main effect of age bin, collapsed across long-axis (Fig. 1). We observed large increases in hippocampal connectivity with memory-related regions including the medial PFC, medial parietal, angular gyri, posterior inferior temporal cortex, anterior temporal cortex, and superior and middle frontal gyri at 7–12 months of age compared to 0–6 months, as well as decreases in connectivity between the hippocampus and post-central sulcus. There were fewer changes in hippocampal connectivity between 13–18 month and 7–12 month age bins, although coupling between the hippocampus and medial PFC continued to increase as connectivity with occipital and left middle temporal gyrus decreased. There were no reliable changes in connectivity between 19–25 month and 13–18 month age bins. Thus, most extensive changes in hippocampal-cortical connectivity occurred across the first year, and were primarily characterized by increases in coupling with memory-related cortex.

Functional connectivity along the long-axis of the hippocampus is differentiated in even the youngest infants

Next, we tested for differences in connectivity along the hippocampal long-axis. Connectivity of the hippocampus in adults varies along the long-axis, with the anterior hippocampus showing stronger functional connectivity to ventral medial PFC and anterior temporal lobes, while the posterior hippocampus is more connected to dorsal medial PFC, posterior lateral and ventral temporal cortex, occipital cortex, and much of the precuneus^13–17. To determine whether the same is true of developing infants, we plotted the main effect of long-axis as well as t-tests comparing the anterior > posterior contrast for each age bin (Fig. 2). We observed differentiation of connectivity along the long-axis in all age bins, including the youngest infants aged 0–6 months. Specifically, we found that the anterior hippocampus was more functionally connected to medial PFC and the anterior temporal lobes, while the posterior hippocampus was more broadly connected to medial parietal cortex, cingulate cortex, posterior lateral and inferior temporal cortex, motor cortex, occipital cortex, and the middle frontal gyrus.

While there were evident similarities in connectivity across the age bins, there were also some notable changes with development. We next sought to identify differences in long-axis connectivity as a function of age bin. To answer this question, we turned to the interaction between long-axis and age bin using the aforementioned model. We found 44 neocortical clusters where the interaction between long-axis and age bin was significant (k = 10, p = 0.001, q < 0.01, these neocortical clusters can be viewed on the surface of the brain in Fig. 3, ignoring the color of the clusters presently), spanning the dorsal medial PFC, lateral PFC, premotor cortex, anterior insula, posterior parietal cortex, superior parietal lobule, medial temporal lobe cortex, inferior temporal cortex, superior temporal gyrus, and occipital cortex. Thus, while some degree of anteroposterior differentiation was present in even the youngest infants, there were several cortical areas where connectivity was still developing with the hippocampus along its long axis.

We also found some neocortical clusters wherein hemisphere interacted with our variables of interest (long-axis and age bin). We report these results in Supplementary Figs. 2–4. In addition, as data in this experiment were collected at two different sites, we wanted to make sure that site was not driving observed effects. Out of the 212 infants in this sample, 189 had site information documented at the time of the scan. None of the predictors significantly interacted with site, even at a liberal threshold (k = 1, p < 0.05 uncorrected).

Maturation of functional connectivity along the hippocampal long-axis reveals canonical cortical networks associated with cognition

We next sought to quantify exactly how connectivity along the long-axis was changing with development. As the neocortical clusters elicited by a significant interaction between long-axis and age bin could reflect numerous different patterns of connectivity, we took a hierarchical clustering approach to sort these clusters into superclusters sharing similar patterns of connectivity with the hippocampus across age bins. Specifically, we subtracted each neocortical cluster’s connectivity with the posterior from the anterior hippocampus to get a measure of differential connectivity along the long-axis for each age bin, and grouped neocortical clusters with similar profiles of connectivity together. Our aim with this analysis was to facilitate interpretation by grouping the 44 neocortical clusters together according to the direction of the significant interaction.

Using this method, we identified six superclusters consisting of neocortical clusters that shared similar connectivity profiles with the hippocampus across age bins. We present the agglomeration schedule used to determine the number of superclusters, as well as the cluster dendrogram in Supplementary Fig. 5.A and B. A linear mixed effects model predicting connectivity as a function of long-axis, age bin, and supercluster indicated a significant three-way interaction between the predictors (F(15,2288) = 12.40, p < 0.0001), confirming that the identified superclusters describe reliable differences in connectivity. We present the spatial extent of each supercluster on the brain in Fig. 3, as well as the connectivity profiles for each supercluster averaged across the neocortical clusters included. We note here that while we clustered the neocortical clusters together based on the average anterior-posterior difference in connectivity, we present and statistically test anterior and posterior connectivity separately across participants in Fig. 3 to facilitate interpretation of how connectivity is changing over time. The anterior-posterior difference in connectivity can be observed in Supplementary Figs. 6–11. A detailed list of neocortical clusters included in each supercluster and their coordinates can be found in Supplementary Table 1. A full list of pairwise comparisons of hippocampal connectivity within each supercluster with corresponding statistical significance can be viewed in Supplementary Tables 2–3, and is visually summarized by the black and grey bars in Supplementary Fig. 12. Three of the identified superclusters consisted of neocortical regions canonically associated with episodic memory, and three superclusters comprised canonical networks associated with other types of cognition.

Development of hippocampal connectivity with cortical memory regions

The first supercluster consisted of only the left and right entorhinal cortex, and the second consisted of the left and right parahippocampal cortex. These superclusters had opposing patterns of connectivity with development. At 0–6 months, the entorhinal supercluster showed a large difference in hippocampal connectivity with greater anterior than posterior connectivity (t(2288) = 10.94, p < 0.0001) which attenuated with age. This attenuation can be attributed to connectivity with the posterior hippocampus quickly increasing across the first 12 months (0-6mo vs 7-12mo: t(208)=-3.55, p = 0.003; 7-12mo vs 13-18mo: t(208) = 0.84, p = 0.48; 13-18mo vs 19-25mo: t(208) = 2.68, p = 0.012), and connectivity with the anterior hippocampus gradually decreasing by 19–25 months (0-6mo vs 7-12mo: t(208) = 0.01, p = 0.99; 7-12mo vs 13-18mo: t(208) = 1.5, p = 0.20, 13-18mo vs 19-25mo: t(208) = 3.09, p = 0.005). Despite this attenuation, connectivity with the anterior hippocampus remained stronger than with the posterior for all age bins (7-12mo: t(2288) = 7.85, p < 0.0001; 13-18mo: t(2288) = 7.59, p < 0.0001; 19-25mo: t(2288) = 5.54, p < 0.0001).

The parahippocampal supercluster showed the opposite pattern with a large difference in connectivity driven by greater posterior connectivity than anterior at 0–6 months (t(2288)=-11.38, p < 0.0001) which diminished with age. This diminishing effect was partially driven by anterior connectivity quickly increasing across the first 12 months (0-6mo vs 7-12mo: t(208)=-4.30, p = 0.0001; 7-12mo vs 13-18mo: t(208)=-0.17, p = 0.86; 13-18mo vs 19-25mo: t(208) = 2.78, p = 0.012), while posterior connectivity gradually attenuated by 19–25 months (0-6mo vs 7-12mo: t(208) = 1.6, p = 0.13; 7-12mo vs 13-18mo: t(208) = 0.32, p = 0.75; 13-18mo vs 19-25mo: t(208) = 3.19, p = 0.003). Posterior hippocampal connectivity remained higher than anterior to this cluster across all age bins, despite the attenuation with age (7-12mo: t(2288)=-5.32, p < 0.0001; 13-18mo: t(2288)=-5.74, p < 0.0001; 19-25mo: t(2288)=-4.62, p < 0.0001).

The third supercluster consisted of medial parietal regions, characterized by a similar pattern of greater posterior > anterior connectivity at 0–6 months (t(2288)=-5.33, p < 0.0001) which disappeared in subsequent age bins (7-12mo: t(2288)=-1.12, p = 0.26; 13-18mo: t(2288)=-0.27, p = 0.79; 19-25mo: t(2288)=-1.66, p = 0.097), due to an increase in connectivity with the anterior hippocampus across the first year (0-6mo vs 7-12mo: t(208)=-6.10, p < 0.0001; 7-12mo vs 13-18mo: t(208) = 1.08, p = 0.28; 13-18mo vs 19-25mo: t(208) = 1.37, p = 0.21). There were no differences in connectivity with the posterior hippocampus across successive age bins (0-6mo vs 7-12mo: t(208)=-1.88, p = 0.16; 7-12mo vs 13-18mo: t(208) = 1.76, p = 0.16; 13-18mo vs 19-25mo: t(208) = 0.25, p = 0.86).

To summarize, for all three memory-related superclusters we observed evidence of differentiated connectivity with the hippocampal long-axis very early in development, which diminished as connectivity with the weaker axis of the hippocampus strengthened across the first year. In medial temporal areas, connectivity gradually decreased with the predominant axis across the first two years.

Development of hippocampal connectivity with cortical networks associated with attention, salience, and social processing

Interestingly, the remaining three superclusters revealed networks typically associated with other types of cognition beyond episodic memory. One supercluster primarily consisted of the dorsal anterior cingulate and anterior insula, reminiscent of the cingulo-opercular network in adults. The connectivity profile of this supercluster was characterized by greater anterior > posterior connectivity at 0–6 months (t(2288) = 5.78, p < 0.0001) which gradually flipped with development such that by 19–25 months there was greater connectivity of this supercluster to the posterior hippocampus than the anterior (7-12mo: t(2288) = 2.67, p = 0.008; 13-18mo: t(2288) = 0.95, p = 0.34; 19-25mo: t(2288)=-4.15, p < 0.0001). This pattern was driven by an increase in connectivity of this cluster with the posterior hippocampus across the first year (0-6mo vs 7-12mo: t(208)=-2.29, p = 0.046; 7-12mo vs 13-18mo: t(208)=-1.48, p = 0.21; 13-18mo vs 19-25mo: t(208) = 0.29, p = 0.77), as well a decrease in connectivity with the anterior hippocampus between the 13–18 month and 19–25 month age bins (0-6mo vs 7-12mo: t(208) = 0.46, p = 0.81; 7-12mo vs 13-18mo: t(208)=-0.42, p = 0.81; 13-18mo vs 19-25mo: t(208) = 2.89, p = 0.02). Connectivity between the cingulo-opercular network and hippocampus was therefore gradually inverting across infancy, with a bias of stronger connectivity with the anterior hippocampus shifting to the posterior hippocampus with maturation.

We also found a supercluster reminiscent of the dorsal frontal-parietal network in adults, consisting of the superior parietal lobule, lateral prefrontal cortex, and inferior temporal gyrus. There was no reliable difference in anterior compared to posterior hippocampal connectivity with this supercluster at 0–6 months (t(2288)=-1.75, 0.08), but there was greater connectivity with the posterior hippocampus than anterior from 7–12 months onwards (7-12mo: t(2288)=-7.82, p < 0.0001; 13-18mo: t(2288)=-5.70, p < 0.0001; 19-25mo: t(2288)=-6.03, p < 0.0001), due to an increase in posterior hippocampal connectivity across the first year (0-6mo vs 7-12mo: t(208)=-3.53, p = 0.003; 7-12mo vs 13-18mo: t(208) = 1.51, p = 0.17; 13-18mo vs 19-25mo: t(208) = 0, p = 0.99). Connectivity of the anterior hippocampus with this supercluster did not change across successive age bins (0-6mo vs 7-12mo: t(208) = 1.03, p = 0.62; 7-12mo vs 13-18mo: t(208) = 0.6, p = 0.67; 13-18mo vs 19-25mo: t(208)=-0.58, p = 0.67). Therefore, while there was no differentiation in anteroposterior hippocampal connectivity with the dorsal frontal parietal network in the first 6 months of life, by 1 years old connectivity with the posterior hippocampus strengthened and remained stronger than that with the anterior hippocampus thereafter.

Finally, we identified a supercluster of regions consisting of superior temporal gyrus, vmPFC, and the occipital pole, which are primarily regions that activate during social and schematic processing^29,30. This vmPFC-superior temporal supercluster was characterized by greater posterior than anterior hippocampal connectivity at 0-6mo (t(2288)=-2.63, p = 0.0086), which flipped such that anterior connectivity became greater than posterior at 7–12 (t(2288) = 5.31, p < 0.0001) and 13–18 months (t(2288) = 3.62, p = 0.0003). Anterior-posterior connectivity was not reliably different by 19–25 months (t(2288) = 1.32, p = 0.19). This pattern was driven by an inverted U connectivity profile of the anterior hippocampus across successive age bins, where connectivity peaked between 7–18 months (0-6mo vs 7-12mo: t(208)=-6.20, p < 0.0001; 7-12mo vs 13-18mo: t(208) = 0.87, p = 0.39; 13-18mo vs 19-25mo: t(208) = 2.51, p = 0.016). Posterior hippocampal connectivity did not reliably change across successive age bins (0-6mo vs 7-12mo: t(208)=-0.98, p = 0.76; 7-12mo vs 13-18mo: t(208) = 0.12, p = 0.91; 13-18mo vs 19-25mo: t(208) = 0.67, p = 0.76). It was therefore anterior hippocampal connectivity with these social regions that was changing with development across the first two years of post-natal life.

Cluster validation

We took several steps to validate the clustering solution, as detailed in the Supplementary Methods. Briefly, we first visually inspected the connectivity profiles for each neocortical cluster contributing to the superclusters to verify that the clustering solution was doing a reasonable job of sorting neocortical clusters into superclusters sharing similar profiles (Supplementary Figs. 6–11). Second, we plotted the first two dimensions of a principal components analysis to verify that the 6 identified superclusters were relatively distinct (Supplementary Fig. 5.C). Third, we calculated silhouette statistics to assess overall supercluster quality (Supplementary Fig. 5.D), which again indicated reasonable clustering solution (silhouette statistic = 0.33 on average). Finally, we verified that the clustering solution was not driven by subject-level bias in the supercluster selection, by repeating the hierarchical clustering analysis 212 times using a leave-one-subject-out cross-validation analysis and substantiating our results with “bias-free” estimates of supercluster patterns (Supplementary Fig. 13, Supplementary Tables 4 and 5, see Supplementary Methods for details).

In this study, we examined the development of hippocampal connectivity along the long-axis in the first two years of postnatal life, during which time infants forget at an accelerated rate¹.

Consistent with prior work detailing large increases in connectivity within functional networks across the first year^31–33, we observed extensive increases in connectivity between the hippocampus and memory-related cortical regions across the first 12 months. We also found evidence that an appreciable degree of anterior-posterior differentiation in connectivity was present even in the youngest infants (0–6 months), indicating that major anterior-posterior divisions in hippocampal connectivity is present very early on. Specifically, the anterior hippocampus was more strongly connected to ventral medial PFC and anterior temporal regions, while the posterior hippocampus was more connected to much of posterior medial cortex, as well as the superior parietal lobule, lateral middle prefrontal gyri, and posterior temporal-occipital cortex. This general pattern is similar to that observed in adults^{13,15–17,34,35}, and was present across all age bins examined. This finding suggests that some degree of differentiation along the long-axis is present well before resolution of the infantile amnesia period (i.e. within the first 6 months post-gestation).

This pattern of results differs from conclusions drawn by a recent study that did not observe evidence of differentiation in anterior-posterior hippocampal connectivity in infants aged 9–11 months³⁶. However, the methods employed by their analysis differed from ours. The authors investigated average connectivity of the hippocampus with functional networks that were defined a-priori based on past parcellations in adults. The present work and work from others suggest that hippocampal-cortical connectivity is still developing over the first two postnatal years, becoming more topologically similar to adults³³. It is possible that averaging connectivity across networks defined with adult parcellations was simply not sensitive enough to reveal differences in anterior-posterior connectivity, compared to our seed-voxel analysis which was not topologically constrained.

While there were commonalities in observed anterior-posterior differentiation across the age bins examined, we also detected many neocortical clusters wherein hippocampal connectivity along the long-axis was evidently changing with development, as quantified by a significant interaction between hippocampal long-axis and age-bin. When we clustered these neocortical regions together based on similarity of the direction of the interaction, we observed different patterns of developing connectivity along the hippocampal long-axis with areas of the neocortex typically associated with episodic memory and spatial processing in adults, namely entorhinal, parahippocampal, and medial parietal superclusters^37–39. Specifically, we found evidence of stronger anterior connectivity to entorhinal cortex, and greater posterior hippocampal connectivity to parahippocampal cortex at 0–6 months. This strong differential connectivity along the hippocampal long-axis with these medial temporal cortical regions attenuated by 7–12 months as connectivity increased with the weaker axis (i.e. connectivity between anterior hippocampus and parahippocampal gyrus, and between posterior hippocampus and entorhinal cortex increased). In the medial parietal supercluster, there was stronger posterior connectivity than anterior connectivity at 0–6 months that disappeared by 7–12 months as connectivity of the anterior hippocampus with these regions increased. These findings are generally consistent with work that shows that connectivity is greatly increasing within functional networks within the first year of life^31,32, and between the whole hippocampus and neocortex³³. Contrary to our hypotheses, we found that connectivity of anterior-posterior regions of the hippocampus with these canonical episodic memory areas are quite functionally segregated in the first few months of life. With development, segregation lessons as the weaker axis of the hippocampus becomes more integrated into the network.

Intriguingly, we also revealed functional networks commonly associated with cognition beyond episodic memory⁴⁰, based solely on changing connectivity with the hippocampus along its long-axis with development. Connectivity of the hippocampus with the cingulo-opercular network (commonly referred to as the salience network, often involved in tasks requiring sustained cognitive control and vigilance⁴¹) changed linearly with age. At 0–6 months the anterior hippocampus was more connected to this network than the posterior hippocampus, but this pattern gradually changed over the course of development such that by 19–25 months the posterior hippocampus was more connected than the anterior, similar to the pattern described in adults^13,15,17. The dorsal frontoparietal network (also referred to as the dorsal attention network, as it tends to be active during goal-directed tasks involving top-down attention⁴²) did not show a difference in connectivity along the long-axis at 0–6 months, but showed greater connectivity with the posterior hippocampus by 1 year of age, effectively becoming more segregated and adult-like^15,17,36. Finally, there was a collection of areas that included the ventral medial PFC and superior temporal sulcus, which are regions that are involved in social and schematic processing^29,30. We found that connectivity of the anterior hippocampus was changing with these regions in an inverted U pattern, producing strongest anterior > posterior difference at 7–12 months.

The fact that these regions clustered together into identifiable networks based on their developing connectivity along the long-axis of the hippocampus is particularly striking because in adults, these networks are usually defined using independent component analysis or seed-voxel connectivity from cortical areas of the network in question, and typically do not include the hippocampus. The fact that the hippocampus does not couple strongly with these networks at rest in adults does not imply that hippocampal interaction with these networks is not important for cognitive functioning. Indeed, the hippocampus flexibly interacts with neocortical regions during mnemonic functioning depending on memory content and task demands^43,44. It is well-documented that attention and salience play important roles in shaping hippocampal memories, and further, our memories shape what we attend to^45,46. In infants, behavioral work shows that the development of attention and recognition memory are closely associated^47,48. While primary sensory networks are topologically adult-like at birth, a primitive medial frontoparietal network (also known as the default mode network) becomes more similar to adults across the first post-gestational year, followed shortly after by refinement of cingulo-opercular and lateral frontoparietal networks⁴⁹. The medial frontoparietal and dorsal frontoparietal networks don’t become anti-correlated (as in adults) until approximately 1 year of age³¹, which is when the hippocampus starts synchronizing with the medial frontoparietal network at rest^31,50. We speculate that the finding that hippocampal interaction with cingulo-opercular and dorsal frontoparietal networks is developing during the infantile amnesia window reflect maturing interaction between memory, attention, and salience systems in infants, which may be critical for the formation of long-lasting memories. More work is needed to directly test this hypothesis.

Finally, while there were similarities in regions identified as more connected to the anterior or posterior hippocampus in our infants and what has been reported in adults, the connectivity maps of infants aged 19–25 months (the oldest age bin here examined) did not look exactly like those of adults. Notably, in adults the anterior hippocampus is more connected to motor cortex, superior and inferior lateral frontal gyri, and the angular gyrus^13,15,17,35, while by age 19–25 months these regions were either dominated by posterior hippocampal connectivity or equal connectivity between the two axes. Presumably, differential anterior > posterior connectivity will continue to mature through later ages not examined here, by either anterior connections increasing or posterior decreasing. Within the ages examined, both processes were occurring depending on the cortical region examined, and so we did not find strong evidence that one axis of the hippocampus was more mature than the other.

This work is not without limitations. Examining functional imaging data in infants poses many challenges. As infants age there is a decrease in tSNR as motion-related artifact increases. It’s possible that lack of observed hippocampal-cortical connectivity changes across later age bins (Fig. 1) were somewhat masked by lower tSNR. However, we include these variables as regressors of no interest in our models to mitigate differences in connectivity due to these confounds. The fact that we observed different profiles of connectivity (e.g. some regions showed increases in connectivity with development, others decreases or inverted U shaped patterns) increases our confidence that changes in connectivity are generally not driven by linear changes in motion, respiration, or other artifacts (all of which are indexed by tSNR). In addition, the data were collected at two different sites, but site did not interact with the variables of interest in this study. It is also worth noting that the infants were asleep during data collection, which may affect network coupling to some extent. Recent work has found that sleeping infant functional networks are more similar to sleeping adult functional networks than that of awake adults, although this finding is somewhat dependent on the network examined^51,52.

To summarize, we found evidence of some degree of differentiation of anterior-posterior hippocampal connectivity in even the youngest infants examined (0–6 months). Nonetheless, hippocampal connectivity along the long-axis with the neocortex was evidently still maturing over the first two years of life. This maturation of connectivity was not restricted to classic memory areas, rather, cortical networks canonically associated with attention, salience, and social cognition were also identified based on developing connectivity with the hippocampus along its anterior-posterior axis. These findings raise the possibility that infantile amnesia is in part a disorder of immature functional interaction between memory, attention, and salience systems that engender memory formation in adults. Additionally, these results support the notion that infantile amnesia is in part an issue of under-developed hippocampal-cortical connectivity, and further, suggests that connectivity along the long-axis of the hippocampus is a functionally relevant marker of infant development. The development of hippocampal-cortical connectivity during the earliest years of life may provide important clues for understanding infantile amnesia and cognition more broadly.

Participants

Neuroimaging data was acquired as part of the Baby Connectome Project⁵³. A total of 319 participants spanning 694 sessions were available for analysis. Some infants were scanned multiple times across sessions with age. In order to adhere to assumptions regarding statistical independence of data in our models we included data from only one session per participant, choosing the session with the highest whole-brain temporal signal to noise ratio (tSNR) of the resting-state data. Fifty-four participants were excluded due to poor data quality quantified by excessive motion (AFNI’s @1dDiffmag values > 0.3 mm/TR) or low whole-brain tSNR (< 30). An additional 10 participants were excluded due to failed co-registration during data preprocessing that could not be remedied without extensive manual alteration. Finally, as there were few participants older than 25 months (n = 38 spanning 26–69 months of age), these participants were also excluded. The final sample consisted of 212 infants between the ages of < 1 month and 25 months. We divided participants into four age bins spanning 6–7 months in age: 0–6 months (n = 50), 7–12 months (n = 59), 13–18 months (n = 58), and 19–25 months (n = 45). The decision to bin infants into discrete categorical age ranges was made to allow the exploration of the relationship between age and connectivity without making assumptions about the shape of the association (i.e. linear or nonlinear, as required when modelling continuous data). The age bin cutoffs were constructed to allow a comparable number of participants in each age bin, with roughly equal age ranges. A histogram of the specific ages of infants in each age bin can be viewed in Supplementary Fig. 1.

Procedure and data acquisition

Detailed information regarding the imaging protocol is described by Howell and colleagues (2019). All imaging was acquired on 3T Siemens Prisma MRI scanners at the Center for Magnetic Resonance Research (CMRR) at the University of Minnesota and the Biomedical Research Imaging Center (BRIC) at the University of North Carolina at Chapel Hill, using a Siemens 32 channel head coil. We here utilize the T1w (MPRAGE, 0.8mm isotropic voxels, TE = 2.24, TR = 2400/1060), T2w (3D variable flip angle turbo spin-echo sequence, 0.8mm isotropic voxels, TE = 564, TR = 3200), and two resting-state T2*-weighted functional scans (gradient-echo EPI acquisition: Multiband 8, TR = 800ms, TE = 37ms, flip angle = 52 degrees, FOV = 208mm, 104x91 matrix, 2mm isotropic voxels) with anterior-to-posterior and posterior-to-anterior encoding (420 TRs, 5 min 36 sec for each encoding; total EPI duration: 840 TRs, 11 min 12 sec) per participant.

Preprocessing

Data processing was conducted with Freesurfer (RRID: SCR_001847), infant Freesurfer⁵⁴, and AFNI (RRID:SCR_005927). The T1w anatomical data were processed using either Freesurfer (for participants > 10 months of age) or infant Freesurfer (for participants < = 10 months of age) for tissue segmentation. Different versions of Freesurfer were required for different age ranges, as the grey matter white matter contrast is poor in infants less than a year old, often resulting in failed convergence for adult Freesurfer. Anatomical scans and segmented grey matter, white matter, and CSF volumes were linearly aligned to the EPI data using AFNI’s 3dAllineate. In infants > 10 months, warping the T1w data to the EPI data produced a good coregistration. For infants < = 10 months, the T1w scan was first aligned to the T2w scan (which had better grey matter/white matter contrast). The T2w scan was then co-registered with the EPI data, and the warp was applied to the T1w data to align it with the EPI data. The functional data were despiked (3dDespike), adjusted for slice-time acquisition (3dTshift), warped to the mid-point of the P-A and A-P encodings and volume registered using AFNI’s unWarpEPI, and frame-to-frame motion was estimated. The data were then smoothed with a 4mm FWHM kernel (3dmerge) and rescaled to % signal change. We then applied a modified form of the ANATICOR (Jo et al., 2010) denoising approach with a nuisance regression (3dTfitter) containing: 6 motion parameters, local average of WM signal (averaged within a 15 mm radius sphere), average ventricle signal, as well as the top 3 principal components of the time series from all white matter and ventricle voxels (modified aCompCor, Behzadi et al., 2007; Stoddard et al., 2016). All regressors were detrended with a fourth-order polynomial before denoising (3dDetrend), and the same detrending was applied during nuisance regression to the voxel time series. The residual time series were concatenated across the A-P and P-A resting-state runs. Subject-specific anatomical and functional data were first registered to an age-specific MNI template, and subsequently, to the median age T1w MNI template (11–14 months) using nonlinear warps (3dQwarp).

Region of interest definition

The hippocampus was isolated from the Freesurfer or infantFreesurfer subcortical segmentations for each participant and manually corrected where necessary. The hippocampus was then manually segmented into anterior and posterior segments at the uncal notch⁵⁵ to create anterior and posterior masks for each hemisphere for every participant. The uncal notch was identifiable in even the youngest infants. The resulting masks were co-registered to the group analysis space (median age template) using the warps created during preprocessing.

Functional connectivity computation

We conducted a seed-to-voxel functional connectivity analysis using the left and right anterior and posterior hippocampal masks as seeds. Specifically, the average time-course across voxels within each hippocampal region was calculated using AFNI’s 3dmaskave. Next, the Pearson’s correlation between each hippocampal ROI time-course and the time-course of each voxel in the brain was calculated using 3dTcorr1D. The resulting correlation maps (one for each ROI, reflecting whole-brain functional connectivity), were Fisher-z transformed to normalize the data for group analyses.

Group-level contrasts

We ran a linear model using AFNI’s 3dMVM to test for group differences in connectivity across the age bins. Specifically, connectivity was predicted as a function of long-axis (anterior/posterior), age bin (0-6mo/7-12mo/13-18mo/19-25mo), and hemisphere (left/right). We also included whole brain mean tSNR and mean motion (estimated with AFNI’s 3dDiffmag) as regressors of no interest to adjust effects for variability in tSNR and motion across participants. tSNR is useful for inclusion in these analyses as it is a composite of all sources of nuisance variability that impact the BOLD signal and functional connectivity, including unmodeled effects of motion and respiration. We were particularly interested in the interaction between long-axis compartments and age bin. To directly explore the degree of differential anterior-posterior connectivity at each age, we additionally specified a-priori pairwise contrasts comparing anterior > posterior hippocampal connectivity within each age bin. Finally, we contrasted hippocampal connectivity across successive age bins (7-12mo > 0-6mo, 13-18mo > 7-12mo, 19-25mo > 13-18mo) to better understand the main effect of age bin on hippocampal-cortical connectivity. Voxels surviving a threshold FDR-corrected to q < 0.01 corrected across all voxels in the brain were considered statistically significant. We chose this method of correction because it is robust to inflated false positives due to multiple comparisons. To further mitigate false positives, we additionally conservatively implemented a cluster-forming threshold to remove very small clusters of less than 10 contiguous voxels.

Hierarchical clustering analysis

Having identified several neocortical clusters of voxels where connectivity was changing along the long-axis as a function of age bin (quantified by a significant long-axis x age bin interaction), we next sought to describe the patterns driving the interaction for each region identified. To facilitate interpretation, we clustered together neocortical regional clusters evincing a similar direction of interaction to form what we term here, “superclusters”. As we did not know the number of superclusters a-priori, we used a hierarchical clustering approach. We extracted the average connectivity across participants and hemispheres between the anterior and posterior hippocampi and each neocortical cluster where there was a significant interaction for each age bin. We subtracted posterior from anterior hippocampal connectivity to get a measure of the difference in connectivity along the long axis for each neocortical cluster for each age bin. We calculated the difference in connectivity along the long-axis because we were interested specifically in the interaction between long-axis and age bin and did not want the clustering solution to be driven by the main effects of long-axis (as in the case of including anterior and posterior connectivity values separately). The result was a 44 neocortical clusters x 4 age bins matrix of anterior-posterior differential connectivity. We then transformed this matrix into a distance matrix by quantifying the Euclidean distance between each of these observations. Finally, we clustered the distance matrix into superclusters using Ward’s method.

The agglomerative clustering process starts with one cluster for each observation (in this case, the connectivity values between the hippocampus and each neocortical cluster) and combines observations with the smallest distance between them to form new clusters (i.e. superclusters). Ward’s method sums the squared Euclidean distances between each member of a cluster and the cluster centroid, joining clusters which increase the sum of within-cluster square errors the least⁵⁶. A new distance is calculated between new clusters, and the process is repeated until there is only one cluster with all observations in it. We used the agglomeration schedule to determine the optimal number of superclusters, choosing the point in the agglomerative process where the distance between conglomerated clusters jumps, indicating the combination of distant clusters that are different enough that they likely shouldn’t be combined.

Statistical Analyses

Statistical analyses of connectivity or each supercluster were computed in R studio version 2023.12.1.402⁵⁷( http://www.posit.co/). We used a linear mixed effects model to test connectivity as a function of supercluster, long-axis, and age bin with a random intercept for each participant using the nlme package in R (version 3.1.158; https://CRAN.R-project.org/package = nlme)⁵⁸. As model residuals were originally heteroskedastic, we specified the variance structure to allow the variance to vary across levels of the heteroskedastic fixed effects, as identified using Levene’s test for unequal variance as part of the car package in R⁵⁹ (version 3.1.0; https://cran.r-project.org/web/packages/car/index.html). Denominator degrees of freedom and p-values were estimated using the containment method as implemented in nlme. Post-hoc comparisons were interrogated using pairwise t-tests on the estimated marginal means from the omnibus model, using the emmeans package (version 1.7.5, https://cran.r-project.org/web/packages/emmeans/index.html), and were FDR corrected to control for multiple comparisons. To confirm that results from these statistical analyses were not driven by subject-level bias, we repeated our statistical tests on weighted averages computed in a leave-one-out validation approach, as described in the Supplementary Methods.

Cluster validation

We took several steps to validate the hierarchical clustering solution. First, we visually verified that the superclusters were distinct by plotting the first two dimensions of a principal components analysis. Second, we plotted differential anterior-posterior hippocampal connectivity over age bins for each neocortical cluster comprising the superclusters, to visually check that neocortical clusters were being grouped together sensibly based on the interaction between long-axis and age bin. Third, we calculated silhouette statistics to assess overall supercluster quality, and to quantify the degree to which neocortical clusters within each supercluster fit the supercluster they were assigned to versus other superclusters. In order to perform statistics to compare the patterns of each supercluster, it was important to verify that the comparisons would not be biased by subject-level noise as a result of performing the selection and comparisons on the same data (i.e. double-dipping; e.g. Kriegeskorte et al., 2009). To evaluate this, the hierarchical clustering was repeated while leaving each subject out once (212 times). We then tabulated which clusters were grouped together into superclusters across all 212 iterations, which by definition does not depend on subject-level noise. We then constructed “bias-free” estimates of the supercluster patterns using a weighted average of only those pairs of clusters that were grouped together 100% of the time. The bias-free, cross-validated patterns could then be compared to the original patterns based on the entire group of subjects.

Acknowledgements and Data Sharing

We would like to thank Cameron Paranzino for his assistance with preparation of the data. This work was supported by the National Institute of Mental Health (NIMH) Intramural Research Program (ZIA MH-002930), and utilizes data collected as part of the UNC/UMN Baby Connectome Project Consortium. Data used in the preparation of this manuscript were obtained from the NIMH Data Archive (NDA). NDA is a collaborative informatics system created by the National Institutes of Health to provide a national resource to support and accelerate research in mental health. Dataset identifier: 10.15154/w07p-s888. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or of the Submitters submitting original data to NDA.

Conflict of Interest Statement

The authors declare no competing financial interests.

Madsen, H. B. & Kim, J. H. Ontogeny of memory: An update on 40 years of work on infantile amnesia. Behav. Brain Res. 298, 4–14 (2016).
Alberini, C. M. & Travaglia, A. Infantile amnesia: A critical period of learning to learn and remember. J. Neurosci. 37, 5783–5795 (2017).
Bauer, P. J. Constructing a past in infancy: A neuro-developmental account. Trends Cogn. Sci. 10, 175–181 (2006).
Gilmore, J. H. et al. Longitudinal development of cortical and subcortical gray matter from birth to 2 year. Cereb. Cortex 22, 2478–2485 (2012).
Uematsu, A. et al. Developmental Trajectories of Amygdala and Hippocampus from Infancy to Early Adulthood in Healthy Individuals. PLoS ONE 7, e46970 (2012).
Feigley, D. A. & Spear, N. E. Effect of age and punishment condition on long-term retention by the rat of active- and passive-avoidance learning. J. Comp. Physiol. Psychol. 73, 515–526 (1970).
Kirby, R. Aquisition, extinction, and retention of an avoidance response in rats as a function of age. J. Comp. Physiol. Psychol. 56, 158–162 (1963).
Tsai, T.-C., Huang, C.-C. & Hsu, K.-S. Infantile Amnesia Is Related to Developmental Immaturity of the Maintenance Mechanisms for Long-Term Potentiation. Mol. Neurobiol. 56, 907–919 (2019).
Guskjolen, A. et al. Recovery of ‘Lost’ Infant Memories in Mice. Curr. Biol. CB 28, 2283-2290.e3 (2018).
Rovee-Collier, C. K., Sullivan, M. W., Enright, M., Lucas, D. & Fagen, J. W. Reactivation of infant memory. Science 208, 1159–1161 (1980).
Teyler, T. J. & Rudy, J. W. The Hippocampal Indexing Theory and Episodic Memory: Updating the Index. Hippocampus 17, 1158–1169 (2007).
Gilmore, A. W. et al. Dynamic Content Reactivation Supports Naturalistic Autobiographical Recall in Humans. J. Neurosci. 41, 153–166 (2021).
Barnett, A. J., Man, V. & McAndrews, M. P. Parcellation of the hippocampus using resting functional connectivity in temporal lobe epilepsy. Front. Neurol. 10, 1–12 (2019).
Barnett, A. J. et al. Intrinsic connectivity reveals functionally distinct cortico-hippocampal networks in the human brain. PLOS Biol. 19, e3001275 (2021).
Frank, L. E., Bowman, C. R. & Zeithamova, D. Differential Functional Connectivity along the Long Axis of the Hippocampus Aligns with Differential Role in Memory Specificity and Generalization. J. Cogn. Neurosci. 31, 1958–1975 (2019).
Grady, C. L. Meta‐analytic and functional connectivity evidence from functional magnetic resonance imaging for an anterior to posterior gradient of function along the hippocampal axis. Hippocampus 30, 456–471 (2020).
Vogel, J. W. et al. A molecular gradient along the longitudinal axis of the human hippocampus informs large-scale behavioral systems. Nat. Commun. 11, 960 (2020).
Audrain, S. & McAndrews, M. P. Schemas provide a scaffold for neocortical integration of new memories over time. Nat. Commun. 13, 1–16 (2022).
Robin, J. & Moscovitch, M. Details, gist and schema: Hippocampal–neocortical interactions underlying recent and remote episodic and spatial memory. Curr. Opin. Behav. Sci. 17, 114–123 (2017).
Sheldon, S. & Levine, B. The role of the hippocampus in memory and mental construction. Ann. N. Y. Acad. Sci. 1369, 76–92 (2016).
Audrain, S., Gilmore, A. W., Wilson, J. M., Schacter, D. L. & Martin, A. A Role for the Anterior Hippocampus in Autobiographical Memory Construction Regardless of Temporal Distance. J. Neurosci. 42, 6445–6452 (2022).
Ghetti, S. Development of item-space and item-time binding. Curr. Opin. Behav. Sci. 17, 211–216 (2017).
Johnson, E. G., Prabhakar, J., Mooney, L. N. & Ghetti, S. Neuroimaging the sleeping brain: Insight on memory functioning in infants and toddlers. Infant Behav. Dev. 58, 101427 (2020).
Ellis, C. T. et al. Evidence of hippocampal learning in human infants. Curr. Biol. 31, 3358-3364.e4 (2021).
Insausti, R., Cebada-Sánchez, S. & Marcos, P. Postnatal development of the human hippocampal formation. Adv. Anat. Embryol. Cell Biol. 206, 1–86 (2010).
Lavenex, P. & Banta Lavenex, P. Building hippocampal circuits to learn and remember: Insights into the development of human memory. Behav. Brain Res. 254, 8–21 (2013).
Vos de Wael, R. et al. Anatomical and microstructural determinants of hippocampal subfield functional connectome embedding. Proc. Natl. Acad. Sci. 115, 10154–10159 (2018).
Marcus, D. S. et al. Informatics and Data Mining Tools and Strategies for the Human Connectome Project. Front. Neuroinformatics 5, 1–12 (2011).
Adolphs, R. The social brain: neural basis of social knowledge. Annu. Rev. Psychol. 60, 693–716 (2009).
Gotts, S. J. et al. Fractionation of social brain circuits in autism spectrum disorders. Brain J. Neurol. 135, 2711–2725 (2012).
Gao, W. et al. The Synchronization within and Interaction between the Default and Dorsal Attention Networks in Early Infancy. Cereb. Cortex 23, 594–603 (2013).
Gao, W., Alcauter, S., Smith, J. K., Gilmore, J. H. & Lin, W. Development of human brain cortical network architecture during infancy. Brain Struct. Funct. 220, 1173–1186 (2015).
Liu, J. et al. Hippocampal functional connectivity development during the first two years indexes 4-year working memory performance. Cortex 138, 165–177 (2021).
Adnan, A. et al. Distinct hippocampal functional networks revealed by tractography-based parcellation. Brain Struct. Funct. 221, 2999–3012 (2016).
Robinson, J. L., Salibi, N. & Deshpande, G. Functional connectivity of the left and right hippocampi: Evidence for functional lateralization along the long-axis using meta-analytic approaches and ultra-high field functional neuroimaging. NeuroImage 135, 64–78 (2016).
Howell, A. L., Osher, D. E., Li, J. & Saygin, Z. M. The intrinsic neonatal hippocampal network: rsfMRI findings. J. Neurophysiol. 124, 1458–1468 (2020).
Coutureau, E. & Di Scala, G. Entorhinal cortex and cognition. Prog. Neuropsychopharmacol. Biol. Psychiatry 33, 753–761 (2009).
Ritchey, M. & Cooper, R. A. Deconstructing the Posterior Medial Episodic Network. Trends Cogn. Sci. 24, 451–465 (2020).
Rugg, M. D. & Vilberg, K. L. Brain networks underlying episodic memory retrieval. Curr. Opin. Neurobiol. 23, 255–260 (2013).
Uddin, L. Q., Yeo, B. T. T. & Spreng, R. N. Towards a Universal Taxonomy of Macro-scale Functional Human Brain Networks. Brain Topogr. 32, 926–942 (2019).
Uddin, L. Q. Salience Network of the Human Brain. (Elsevier, Netherlands, 2017).
Corbetta, M. & Shulman, G. L. Control of goal-directed and stimulus-driven attention in the brain. Nat. Rev. Neurosci. 3, 201–215 (2002).
Gilboa, A. & Moscovitch, M. No consolidation without representation: Correspondence between neural and psychological representations in recent and remote memory. Neuron 1–17 (2021) doi:10.1016/j.neuron.2021.04.025.
Sheldon, S. & Levine, B. The medial temporal lobe functional connectivity patterns associated with forming different mental representations. Hippocampus 28, 269–280 (2018).
Aly, M. & Turk-Browne, N. B. How hippocampal memory shapes and is shaped by attention. in The Hippocampus from Cells to Systems 369–403 (Springer, Cham, 2017).
Lisman, J. E. & Grace, A. A. The hippocampal-VTA loop: Controlling the entry of info into long-term memory. Neuron 46, 703–713 (2005).
Richards, J. E., Reynolds, G. D. & Courage, M. L. The neural basis of infant attention. Assoc. Psychol. Sci. 19, 41–46 (2010).
Rose, S. A., Feldman, J. F. & Jankowski, J. J. Infant visual recognition memory: Independent contributions of speed and attention. Dev. Psychol. 39, 563–571 (2003).
Gao, W. et al. Functional Network Development During the First Year: Relative Sequence and Socioeconomic Correlations. Cereb. Cortex 25, 2919–2928 (2015).
Gao, W. et al. Evidence on the emergence of the brain’s default network from 2-week-old to 2-year-old healthy pediatric subjects. Proc. Natl. Acad. Sci. 106, 6790–6795 (2009).
Mitra, A. et al. Resting-state fMRI in sleeping infants more closely resembles adult sleep than adult wakefulness. PLOS ONE 12, e0188122 (2017).
Yates, T. S., Ellis, C. T. & Turk-Browne, N. B. Functional networks in the infant brain during sleep and wake states. Preprint at https://doi.org/10.1101/2023.02.15.528718 (2023).
Howell, B. R. et al. The UNC/UMN Baby Connectome Project (BCP): An overview of the study design and protocol development. NeuroImage 185, 891–905 (2019).
Zöllei, L., Iglesias, J. E., Ou, Y., Grant, P. E. & Fischl, B. Infant FreeSurfer: An automated segmentation and surface extraction pipeline for T1-weighted neuroimaging data of infants 0-2 years. NeuroImage 218, 116946 (2020).
Poppenk, J., Evensmoen, H. R., Moscovitch, M. & Nadel, L. Long-axis specialization of the human hippocampus. Trends Cogn. Sci. 17, 230–240 (2013).
Ward, J. H. Hierarchical grouping to optimize an objective function. J. Am. Stat. Assoc. 58, (1963).
Posit team. RStudio: Integrated Development Environment for R. Posit Software (2024).
Pinheiro, J. & Bates, D. Mixed-effects models in S and S-PLUS. xvi, 528 p. (2000).
Fox, J. & Weisberg, S. An R Companion to Applied Regression. (Sage, Thousand Oaks, CA, 2019).

There is NO Competing Interest.

BabbyHipposSupplemental.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

The development of functional connectivity along the hippocampal long-axis in infants

Status:

Version 1

Abstract

Figures

Introduction

Results

Hippocampal coupling with memory-related cortex increases across the first 6–12 months of age

Functional connectivity along the long-axis of the hippocampus is differentiated in even the youngest infants

Maturation of functional connectivity along the hippocampal long-axis reveals canonical cortical networks associated with cognition

Development of hippocampal connectivity with cortical memory regions

Development of hippocampal connectivity with cortical networks associated with attention, salience, and social processing

Cluster validation

Discussion

Online Methods

Participants

Procedure and data acquisition

Preprocessing

Region of interest definition

Functional connectivity computation

Group-level contrasts

Hierarchical clustering analysis

Statistical Analyses

Cluster validation

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1